Strategic Control of 60GHz Millimeter-Wave High-Speed...
Transcript of Strategic Control of 60GHz Millimeter-Wave High-Speed...
Research ArticleStrategic Control of 60GHz Millimeter-Wave High-SpeedWireless Links for Distributed Virtual Reality Platforms
Joongheon Kim1 Jae-Jin Lee2 andWoojoo Lee3
1School of Computer Science and Engineering Chung-Ang University Seoul Republic of Korea2SoC Design Group Electronics and Telecommunications Research Institute (ETRI) Daejeon Republic of Korea3Department of Electronics Engineering Myongji University Yongin Republic of Korea
Correspondence should be addressed to Woojoo Lee spaceetrirekr
Received 25 November 2016 Accepted 5 March 2017 Published 28 March 2017
Academic Editor Konstantinos Demestichas
Copyright copy 2017 Joongheon Kim et alThis is an open access article distributed under the Creative CommonsAttribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
This paper discusses the stochastic and strategic control of 60GHzmillimeter-wave (mmWave)wireless transmission for distributedand mobile virtual reality (VR) applications In VR scenarios establishing wireless connection between VR data-center (calledVR server (VRS)) and head-mounted VR device (called VRD) allows various mobile services Consequently utilizing wirelesstechnologies is obviously beneficial in VR applications In order to transmit massive VR data the 60GHz mmWave wirelesstechnology is considered in this research However transmitting the maximum amount of data introduces maximum powerconsumption in transceiversTherefore this paper proposes a dynamicadaptive algorithm that can control the power allocation inthe 60GHz mmWave transceivers The proposed algorithm dynamically controls the power allocation in order to achieve time-average energy-efficiency for VR data transmission over 60GHz mmWave channels while preserving queue stabilization Thesimulation results show that the proposed algorithm presents desired performance
1 Introduction
As actively discussed nowadays virtual reality (VR) or aug-mented reality (AR) applications have received a lot of atten-tion by industry and academia research organizations [1ndash4]In order to provide mobile services in head-mounted VRdisplays in VR applications establishing wireless connectionsbetween VR video storage (called VR server (VRS)) andhead-mounted VR devices (VRD) is essential The candidatewireless technologies should provide ultra-high-speed datacommunication speeds for transmitting VR video streamingwithout latencies
For massive high-volume and high-definition multime-dia contents delivery millimeter-wave (mmWave) wirelesscommunication technologies are widely discussed nowadays[5ndash7] InmmWavewireless communication research 60GHzwireless technologies are generally and widely considered[8ndash11] because it is only one standardized millimeter-wavewireless technology so called IEEE 80211ad [12]
In this paper the 60GHz wireless channel is exploredfor simultaneous large-scalemassivemultimedia informationdelivery based on following reasons
(i) Multigigabit-per-Second (Multi-Gbps) Data SpeedThe currently existing mmWave wireless schemes areable to support multi-Gbps data transmission speedsowing to their ultrawideband channel bandwidth Fortruly immersive VR experiences VR systems shouldsupport the high-speed data communications whichcan be realized by the mmWave technologies
(ii) High Directionality mmWave wireless transmissionis extremely high directional due to its high carrierfrequencies According to the fact that large-scaleand densely deployed VR users will perform wirelesscommunications relatively high interference occur-rence is expected If the transmission beams arequite narrow the interference impacts will be obvi-ously reduced Therefore the high directionality of
HindawiMobile Information SystemsVolume 2017 Article ID 5040347 10 pageshttpsdoiorg10115520175040347
2 Mobile Information Systems
mmWave propagation is beneficial in terms of theinterference reduction
Based on the 60GHz wireless communication technolo-gies this paper proposes a strategic and stochastic queue con-trol algorithm to minimize the time-average expected powerconsumption (which is equivalent to maximizing the time-average expected energy-efficiency) subject to queue stabi-lization The VRS calculates the amount of transmit powerallocation for transmitting data (from the transmission queuein VRS) to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will beprocessed from the queue within the VRS Then the queueshould be more stable whereas power (or energy) manage-ment is not efficient On the other hand if the amount oftransmit power allocation is relatively small the small num-ber of bits will be processed from the transmission queuewithin the VRS Processing insufficient number of bits resultsin queue overflows in the VRS queue which is not allowedbecause it will lose VR information at the VRSTheVR infor-mation loss should induce usersrsquo unsatisfactory experiencesTherefore an energy-efficient stochastic transmit powerallocation algorithm is required to transmit bits with the con-cept of the joint optimization of energy-efficiency and bufferstability
The last sections of this paper are as follows the referencedistributed VR network platforms is introduced in Section 2The next section introduces 60GHz wireless technologieswith 60GHz IEEE 80211ad standards Based on the introduc-tion the methodologies for estimating maximum communi-cation speeds are presented in Section 3 In Section 4 strate-gic and stochastic control for mmWave transmit powers forthe minimization of time-average expected power con-sumption subject to queue stabilization is proposed basedon 60GHz mmWave wireless channel understanding Therelated intensive simulation results are presented in Section 5while Section 6 concludes this paper
2 Reference Mobile Virtual Reality NetworkModel with 60GHz Millimeter-WaveWireless Links
Current VRDs support 360-degree video applications toallow users to interact with digital environments and objectsFor example according to usersrsquo head motions such as pitch-ing (leaning forwardbackward) yawing (rotating leftright)and rolling (spinning clockwisecounterclockwise) the state-of-the-art VRDs offer corresponding video images to theusers Oculus GearVR [13] is a representative one of suchVRDs in market which nicely provides 4K full HD videoplayback
However unfortunately it is hard to agree that currentVRDs are providing the truly immersive VR experiences thathave been an aspiration of many That is because the trulyimmersive VR systems should have three essential featuresquality responsiveness and mobility [14] The quality meansthe realistic visual portrayals which should be based on high
resolution video images For instance videos at 6K resolution(ie the 6K video streaming service already exists in market)or videos at 8K resolution may be the candidate to providesuch high quality visual portrayalsThe responsivenessmeanshow fast a userrsquos motion is reflected to the video playback ina VRD Due to very high sensitivity of human ocular propri-oception immediate feedback from the motion to the VRDdisplay is necessary whereas the current VR systemsmay notbe able to support such fast responsiveness Meanwhile themobility is regarding a question of what we can do in realworld but cannot in the current VR systems Namely we canmove leftright go forwardbackward and jump updownwhich may cause nothing in the current (typical) VRDs
Extensive research efforts have been invested to overcomethe limitations of the current VR systems thereby meetingthe aforesaid requirements For example Facebook has pro-posed the dynamic streaming technique to provide 6K videostreaming service [15] Furthermore tremendous efforts onnew video codecs to support high quality video with highefficiency make us expect that the high resolution videos willbe no longer an issue in near future For the fast respon-siveness issue embedding the bigger size caches (thanks tosemiconductor technology progress) into VRDs seems tobe a plausible solution in that a VRD then rarely needsto send its motion information to a VRS and wait for thecorresponding video data to be delivered from a VRS For themobility issue researchers have recently started studying onmotion parallax [16] which has been driving the advent of 8Kvideo applications By taking a userrsquos position and directioninto consideration these 8K video applications are beyondthe conventional three degrees of freedom (DoF) and enableany motions to be reflected
Nevertheless the truly immersive VR systems do stillhave amajor challenge the high-speedVR network As risingresolution of video applications including various parallaxesof each single image the video applications require a lot ofdata and delivery of these experiences across the Internetpresents the inevitable challenge on network (ie even 6Kresolution used by the GearVRmay be 20 times the size of 4Kvideos)We have searched on themost possible solutions andeventually determined that the mmWave-based network isthe strongest candidate for the near future VR system
Our proposed model of the VR system includingmmWave-based VR network is described in Figure 1 One ormultiple mmWave antennas are deployed in the system eachof which supports the GHz data transmission The userrsquosrequests for example motion changes and VRD controlsare delivered to the VRS and the VRS performs appropriateactions such as serving different video streaming sending dif-ferent parallax images in the playing video and compressingthe video with different codec The mux logic in the figureshows one of the control ways that the VRS carries out torespond to the corresponding user information When thevideo data is delivered each data is piled up in the buffersTheVRD may have multilayer buffers to realize various controlsin the video playback
Mobile Information Systems 3
VR display (VRD)
VRserver(VRS)
mmWAVEantenna User information
Data
VideoData
Select
Figure 1 Concept of the proposed VR system exploiting the mmWave-based VR network
3 Maximum Communication Speed of60GHz Wireless Systems with IEEE80211ad Standard Features
This section mathematically calculates the maximum speed(ie upper bound) of wireless data transmission over 60GHzmmWave channels and eventually defines the quality of VRvideo streaming through the following steps (i) calculatingthe received signal strength at VRD (ii) obtaining maximumwireless data transmission speed and (iii) defining the qualityof streamingThemaximumwireless data transmission speedwill be derived with two possible ways (i) using Shannoncapacity equation and (ii) using the level of supportablemod-ulation and coding scheme (MCS) defined in IEEE 80211adspecification and its associated wireless data transmissionspeed
31 Calculation of the Received Signal Strength at VRD Thereceived signal strength in 60GHz mmWave wireless chan-nels that depends on distance between VRS and VRD(denoted by 119889) at VRD in a milli-Watt scale 119875VRD
mW (119889) can becalculated as follows
119875VRDmW (119889) = 119891mW (119875VRD
dBm (119889)) (1)
where119891mW(119909) = 1011990910 and119875VRDdBm (119889) that is the received sig-
nal strength in 60GHz mmWave wireless channels depend-ing on distance between VRS and VRD (denoted by 119889) atVRD in a dB scale can be obtained as follows
119875VRDdBm (119889) = 119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi (2)
where119866VRSdBi 119866VRD
dBi and 119875VRSdBm are the transmit antenna gain at
VRS in a dB scale the receive antenna gain at VRD in a dBscale and the transmit power (assumed to be 19 dBm [17]) atVRS in a dB scale respectively 119871(119889) is the path-loss thatdepends on 119889 In the equation 119866VRS
dBi = 24 dBi according to[17] and we assume 119866VRD
dBi = 0 dBi which corresponds to the
omnidirectional receive antenna at VRD Note that assumingomnidirectional receive antenna at VRD is beneficial becauseit reduces beam training time overhead that is essential formobile mmWave services as clearly discussed in [18]
According to the 60GHz IEEE 80211ad line-of-sight(LOS) path-loss [19] the dB scaled 119871(119889) in (2) can be definedas follows
119871 (119889) = 119860 + 20 log10 (119891GHz) + 10119899 log10 (119889) (3)
where 119860 is a specific value for the selected type of antennaand beamforming algorithm that relies on the antennabeamwidth that is 119860 = 325 dB as defined in [19] In addi-tion the path-loss coefficient 119899 is set to 2 and 119891GHz is thecarrier frequency inGHz thereby119891GHz = 60The direct pathsbetween VRS and VRD have no obstacles thus LOS model isconsidered in this research
Based on the calculation result of the received signalstrength in VRD the interfering signals are added to thedesired transmission through the main wireless link fromVRS to VRD Therefore signal-to-interference plus noiseratio (SINR) in a dB scale denoted by 120574dB can be obtained asfollows
120574dB (119889) = 119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW) (4)
where 119891dB(119909) = 10 sdot log10(119909) and 119899mW is a background noisein a milli-Watt scale that can be derived as follows [20]
119899mW = 119891mW (119896119861119879119890 + 10 log10119861 + 119865119899) (5)
In (5) 119896119861119879119890 is a noise power spectral density whichis minus174 dBmHz [20] 119861 is one subchannel bandwidth in60GHz mmWave which is defined as 216GHz in IEEE80211ad [5] and 119865119899 is a noise figure which is assumed to be5 dB [12] 119878119868 in (4) is a set of the wireless transmissions fromVRS toVRD (those are not ourmain considering VRwirelesscommunications) excluding the aforementioned main video
4 Mobile Information Systems
VRserver
Transmitantenna
VRserver
Transmitantenna
Main video streaming link
Interference
(VRS) A (VRS) B
(lowast lowast)Link i
VR display (VRD) A VR display (VRD) B
Figure 2 Interference scenario with directional transmit antennas
streaming that is a set of interference transmissions Inaddition 119868119894mW means the interference in a milli-Watt scale toour main VRD induced by the neighbor 60GHz mmWavewireless transmissions from links 119894 where 119894 isin 119878119868 In a conse-quence the interference from the 60GHz mmWave wirelesslink 119894 can be expressed as follows
119868119894mW = 119891mW (119866119894VRSdBi (120579lowast 120601lowast) + 119875119894VRSdBm minus 119871 (119889119894)) (6)
where 119875119894VRSdBm stands for the transmit power from the VRS of60GHz mmWave wireless link 119894 and 119871(119889119894) is the path-loss in(3) for 119889119894 that is the distance between the VRS of link 119894 andthe VRD of the main video streaming Lastly 119866119894VRSdBi (120579lowast 120601lowast)is the transmit antenna radiation gain from the transmitterof 60GHz mmWave wireless link 119894 to the VRD of main videostreaming It means that 120579lowast and 120601lowast are the azimuthal and ele-vational angular differences between two links (one link ismain video streaming and the other link is wireless link 119894)
Figure 2 provides an example for the interference sce-nario As illustrated in Figure 2 there exist two parallel VRwireless data transmissions One is from the VRS of mainvideo streaming (named VRS 119860 in Figure 2) to its associatedVRD (named VRD 119860 in Figure 2) of main video streamingand the other one is from VRS of wireless link 119894 (named VRS119861 in Figure 2) to its associated VRD of wireless link 119894 (namedVRD 119861 in Figure 2)120579lowast and 120601lowast are the azimuth and elevation angular differ-ences between the link 119894 and main video streaming respec-tively As illustrated in Figure 2 the directional wirelesstransmission from VRS 119861 to VRD 119861 will be one potentialinterference source to the VRD 119860
32 Obtaining Maximum Wireless Data Transmission SpeedIn order to obtain the maximum wireless data transmissionspeed two possible methods may exist (i) using Shannoncapacity equation (refer to Section 321) and (ii) using thelevel of supportable MCS defined in IEEE 80211ad specifica-tion and its associated wireless data transmission speed (referto Section 322)
321 Using Shannon Capacity Equation Based on the SINRcalculated in (4) in Section 31 the theoretical maximumwireless data transmission speed is formulated as
119861 sdot log2 (120574dB (119889)) (7)where 119861 is one subchannel bandwidth which is 216GHz in60GHz mmWave wireless systems
322 Using the Level of Supportable MCS Defined in IEEE80211ad Specification and Its Associated Wireless Data Trans-mission Speed This subsection uses 120574dB(119889) which resultedfrom the previous step in Section 31 The derived 120574dB(119889) willbe compared with the receiver sensitivity defined in IEEE80211ad specification [12]The receiver sensitivity values andtheir associated supportable wireless data transmissionspeeds are summarized in Table 1 [12] However the directcomparison between 119903dB(119889) and the values in Table 1 [12] maynot be possible Instead the values should be converted tosignal-to-noise ratio (SNR) as
SNR119896dB = [Receiver-Sensitivity]119896 minus 119899dBm (8)
where [Receiver-Sensitivity]119896 is a receiver sensitivity of MCSlevel 119896 SNR119896dB is the associated corresponding SNR of the
Mobile Information Systems 5
Table 1 IEEE 80211ad single-carrier MCS table [12]
Receiver Selected MCS indicesSensitivity (Data rates unit Mbps)minus78 dBm MCS0 (275)minus68 dBm MCS1 (385)minus66 dBm MCS2 (770)minus65 dBm MCS3 (9625)minus64 dBm MCS4 (1155)minus63 dBm MCS6 (1540)minus62 dBm MCS7 (1925)minus61 dBm MCS8 (2310)minus59 dBm MCS9 (25025)minus55 dBm MCS10 (3080)minus54 dBm MCS11 (3850)minus53 dBm MCS12 (4620)
receiver sensitivity of MCS level 119896 and 119899dBm is a backgroundnoise in a dB scale as previously presented in (5) When theobtained 120574dB(119889) value in Section 31 is in between SNR119894dB (iethe SNR of MCS index 119894 (lower)) and SNR119895dB (ie the SNRof MCS index 119895 (higher)) the 60GHz mmWave wireless linkof the main video streaming uses the MCS index 119894 Once wedetermine the supportable MCS level the associated wirelessdata transmission speed is obtained from Table 1
33 Quality Estimation of Wireless Video Streaming [12] Inorder to estimate the quality of the wireless video streamingthat depends on the derived maximum wireless data trans-mission speed we use the model presented in [5 12] Thissection summaries the discussion in [5 12] The wirelesstransmission of 1080 p 30 fps (30 frames each of which has1080-by-1920 pixels will be presented in a display per second)in RGB (8 bits for each RedGreenBlue color representationie 8 times 3 = 24 bits) video frames with no compression (nosource coding but only channel coding exists) requires the15-gigabits (Gbps) wireless data transmission speed sinceeach frame has 1920times1080 pixels with 30 frames per secondand each pixel has 24 bits in an RGB format Similarly thewireless data transmission speed of 3Gbps is at least requiredfor the full-quality 1080 60 fps streaming in RGB (ie1080times1920times24times60) In YCbCr 4 2 0 formats the requiredwireless data transmission speed is half of the wireless datatransmission speed in RGB formats that is 075Gbps and15Gbps speeds are required for nonsource-coded real-time1080 p 30 fps and 1080 p 60 fps video streaming asintroduced If themaximumwireless data transmission speedcalculated and obtained in Sections 31 and 32 is less than thefull-quality threshold the quality may degrade One widelysuggested reference model of the quality is as [5 12]
119891119902 (119909) = 1
log120572 (119909max + 1) log120572 (119909 + 1) if 119909 lt 119909max119909max otherwise (9)
where 120572 is a base (1 lt 120572) 119909max is a desired wireless datatransmission speed for the uncompressed streaming (ie
075Gbps in 1080 30 fps in YCbCr 4 2 0 formats 15 Gbpsin 1080 p 60 fps in YCbCr 4 2 0 formats 15 Gbps in1080 p 30 fps in RGB formats and 30Gbps in 1080 p 60 fps in RGB formats) and 119909 is the achievable wireless datatransmission speed derived in Sections 31 and 32
4 Proposed Strategic Control of 60GHzWireless Communications in Mobile VirtualReality Applications
41 Design Rationale For each wireless transmission fromVRS to VRD if the transmission queue in VRS is almostempty the transmission algorithm should not need to allocatea lot of transmit power in order to transmit more bitsfrom the queue Then the algorithm will allocate relativelylittle amount of powers into the transmitter in order towork in the energy-efficient manner On the other hand ifthe transmission queue in VRS is almost occupied byVR data information (almost in the overflow situations)the transmission algorithm should allocate relatively largeamount of powers into the transmitter in order to avoid thequeue-overflow situations Therefore this section proposesa stochastic optimization framework that works for theminimization of time-average expected power consumption(which is equivalent to the maximization of time-averageexpected energy-efficiency) while preserving queue stabi-lization that depends on the queue-backlog sizes in each unittime
42 Strategic Control of 60GHz Wireless Virtual RealityApplications For each VRS V119894 isin V where V is the set ofexisting wireless links between VRSs and VRDs the queuedynamics in unit time 119905 isin 0 1 can be formulated asfollows
119876119894 [119905 + 1] = (119876119894 [119905] minus 120583119894 [119905] + 120582119894 [119905])+ 119876119894 [0] = 0 (10)
where (119886)+ ≜ max119886 0 and 119876119894[119905] 120583119894[119905] and 120582119894[119905] mean thequeue backlog at VRS forallV119894 isin V at 119905 the departure process(ie wireless transmission of VR videos to VRD) from VRSforallV119894 isin V at 119905 and the arrival process (ie VR video chunkplacement) to VRS forallV119894 at 119905 respectively The unit of 119876119894[119905]120583119894[119905] and 120582119894[119905] is bit Here 120583119894[119905] and 120582119894[119905]mean the amountsof transmitted bits from VRS forallV119894 isin V to its associatedVRD at unit time 119905 over 60GHz mmWave wireless channelsand the generated data in VRS forallV119894 isin V to be transmittedto its associated VRD at unit time 119905 over 60GHz mmWavewireless channels respectively Notice that the arrival processis independent of power allocation decision
In order to formulate 120583119894[119905] we consider the case wherebythe maximum wireless data transmission speed is obtainedfrom the Shannon capacity equation Then according to (1)and (2) it is obvious that120583119894 [119905] = 119861 sdot log2 (120574dB (119889)) (11)
= 119861 sdot log2 (119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW)) (12)
6 Mobile Information Systems
Parameters(i) 119881 a parameter for power-delay (queueing delay) tradeoffs(ii) 119861 channel bandwidth (216GHz in 60GHz mmWave wireless systems)(iii) 119899mW background noise(iv) 119866VRS
dBi transmit antenna gain at VRS(v)P a set of possible power allocation candidates(vi) 119878119868 a set of possible interference sources
Stochastic Power Allocation for the Transceiver of VR Server (VRS)119905 = 0 119879max the number of discrete-time operationwhile 119905 le 119879max do
(i) Observes 119876119894[119905] 119889 and 119868119894mW(ii) Finds stochastic optimal power allocation 119875VRS
119894dBm[119905] inP which can minimize
119875VRS119894dBm[119905] minus 119881119861119876119894 [119905] log2 (119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm[119905] minus 119871(119889))
119899mW + sum119894isin119878119868 119868119894mW))
Algorithm 1 Stochastic power allocation for the minimization of time-average expected power consumption subject to queue-stabilizationin the transceiver of virtual reality (VR) servers
= 119861sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi )119899mW + sum119894isin119878119868 119868119894mW
)) (13)
= 119861 sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
dBm minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (14)
Note that 119866VRDdBi is disappeared in (14) because 119866VRD
dBi =0 dBi (again we supposed the omnidirectional receiverantenna setting)
The mathematical program to minimize the summationof the time-average expected power consumption (which isobviously equivalent to the maximization of the summationof time-average expected energy-efficiency) is as follows
min sumV119894isinV
E [119875VRS119894dBm] (15)
where E[119875VRS119894dBm] stands for the time-average expected power
consumption at V119894 isin V and this is equivalent to
min sumV119894isinV
( lim119905rarrinfin
1119905119905minus1sum120591=0
119875VRS119894dBm [120591]) (16)
Note the objective function in our stochastic optimizationframework has the following two constraints (i) a constraintfor queue rate stability and (ii) a constraint for discretizedpower allocation due to radio frequency (RF) hardwaredesign limitations (ie it is impossible to allocation real-number scaled transmit powers) The first can be expressedas
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]]120591 = 0 forallV119894 isin V (17)
and we can rewrite (17) in this formulation
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]] lt infin forallV119894 isin V (18)
The constraint for discretized power allocation can beexpressed as
119875VRS119894dBm [119905] isin P ≜ 119901min 1199011 1199012 119901max (19)
Now let us introduce a new variable Θ(119905) that denotesthe column vector of all queues in VRSes at unit time 119905 anditeratively define the quadratic Lyapunov function 119871[119905] asfollows
119871 [119905] = 12Θ119879 [119905] Θ [119905] = 12 sumV119894isinV
(119876119894 [119905])2 (20)
where Θ119879[119905] denotes the transpose of Θ[119905] Then let usintroduce another new variable Δ[119905] that is a conditionalquadratic Lyapunov function that may be formulated asfollows
Δ [119905] ≜ E [119871 [119905 + 1] minus 119871 [119905] | Θ [119905]] (21)
that is the drift on unit time 119905 The decision-making policyof the dynamic and stochastic power allocation is designedto (i) solve the minimization of time-average expected powerconsumption formulation by observing the current queue-backlog sizes 119876119894[119905] and (ii) iteratively determine the amountof power allocation 119875VRS
119894dBm for maximizing a bound on
E [119875VRS [119905] | Θ [119905]] minus 119881Δ [119905] (22)
where 119875VRS[119905] is the column vector of 119875VRS119894dBm[119905] forallV119894 isin V
and 119881 is the positive constant control parameter setting ofthe dynamicstochastic decision-making policy that affectsthe power-delay (ie queueing delay) tradeoffs
The proposed algorithm involves minimizing a bound onE[119875VRS[119905] | Θ[119905]] minus 119881Δ[119905] and finally this Lyapunov opti-mization theory gives Algorithm 1 to minimize the followingequation
sumV119894isinV
119875VRS119894dBm [119905] + 119881 sum
V119894isinV119876119894 [119905] (120582119894 [119905] minus 120583119894 [119905]) (23)
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
2 Mobile Information Systems
mmWave propagation is beneficial in terms of theinterference reduction
Based on the 60GHz wireless communication technolo-gies this paper proposes a strategic and stochastic queue con-trol algorithm to minimize the time-average expected powerconsumption (which is equivalent to maximizing the time-average expected energy-efficiency) subject to queue stabi-lization The VRS calculates the amount of transmit powerallocation for transmitting data (from the transmission queuein VRS) to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will beprocessed from the queue within the VRS Then the queueshould be more stable whereas power (or energy) manage-ment is not efficient On the other hand if the amount oftransmit power allocation is relatively small the small num-ber of bits will be processed from the transmission queuewithin the VRS Processing insufficient number of bits resultsin queue overflows in the VRS queue which is not allowedbecause it will lose VR information at the VRSTheVR infor-mation loss should induce usersrsquo unsatisfactory experiencesTherefore an energy-efficient stochastic transmit powerallocation algorithm is required to transmit bits with the con-cept of the joint optimization of energy-efficiency and bufferstability
The last sections of this paper are as follows the referencedistributed VR network platforms is introduced in Section 2The next section introduces 60GHz wireless technologieswith 60GHz IEEE 80211ad standards Based on the introduc-tion the methodologies for estimating maximum communi-cation speeds are presented in Section 3 In Section 4 strate-gic and stochastic control for mmWave transmit powers forthe minimization of time-average expected power con-sumption subject to queue stabilization is proposed basedon 60GHz mmWave wireless channel understanding Therelated intensive simulation results are presented in Section 5while Section 6 concludes this paper
2 Reference Mobile Virtual Reality NetworkModel with 60GHz Millimeter-WaveWireless Links
Current VRDs support 360-degree video applications toallow users to interact with digital environments and objectsFor example according to usersrsquo head motions such as pitch-ing (leaning forwardbackward) yawing (rotating leftright)and rolling (spinning clockwisecounterclockwise) the state-of-the-art VRDs offer corresponding video images to theusers Oculus GearVR [13] is a representative one of suchVRDs in market which nicely provides 4K full HD videoplayback
However unfortunately it is hard to agree that currentVRDs are providing the truly immersive VR experiences thathave been an aspiration of many That is because the trulyimmersive VR systems should have three essential featuresquality responsiveness and mobility [14] The quality meansthe realistic visual portrayals which should be based on high
resolution video images For instance videos at 6K resolution(ie the 6K video streaming service already exists in market)or videos at 8K resolution may be the candidate to providesuch high quality visual portrayalsThe responsivenessmeanshow fast a userrsquos motion is reflected to the video playback ina VRD Due to very high sensitivity of human ocular propri-oception immediate feedback from the motion to the VRDdisplay is necessary whereas the current VR systemsmay notbe able to support such fast responsiveness Meanwhile themobility is regarding a question of what we can do in realworld but cannot in the current VR systems Namely we canmove leftright go forwardbackward and jump updownwhich may cause nothing in the current (typical) VRDs
Extensive research efforts have been invested to overcomethe limitations of the current VR systems thereby meetingthe aforesaid requirements For example Facebook has pro-posed the dynamic streaming technique to provide 6K videostreaming service [15] Furthermore tremendous efforts onnew video codecs to support high quality video with highefficiency make us expect that the high resolution videos willbe no longer an issue in near future For the fast respon-siveness issue embedding the bigger size caches (thanks tosemiconductor technology progress) into VRDs seems tobe a plausible solution in that a VRD then rarely needsto send its motion information to a VRS and wait for thecorresponding video data to be delivered from a VRS For themobility issue researchers have recently started studying onmotion parallax [16] which has been driving the advent of 8Kvideo applications By taking a userrsquos position and directioninto consideration these 8K video applications are beyondthe conventional three degrees of freedom (DoF) and enableany motions to be reflected
Nevertheless the truly immersive VR systems do stillhave amajor challenge the high-speedVR network As risingresolution of video applications including various parallaxesof each single image the video applications require a lot ofdata and delivery of these experiences across the Internetpresents the inevitable challenge on network (ie even 6Kresolution used by the GearVRmay be 20 times the size of 4Kvideos)We have searched on themost possible solutions andeventually determined that the mmWave-based network isthe strongest candidate for the near future VR system
Our proposed model of the VR system includingmmWave-based VR network is described in Figure 1 One ormultiple mmWave antennas are deployed in the system eachof which supports the GHz data transmission The userrsquosrequests for example motion changes and VRD controlsare delivered to the VRS and the VRS performs appropriateactions such as serving different video streaming sending dif-ferent parallax images in the playing video and compressingthe video with different codec The mux logic in the figureshows one of the control ways that the VRS carries out torespond to the corresponding user information When thevideo data is delivered each data is piled up in the buffersTheVRD may have multilayer buffers to realize various controlsin the video playback
Mobile Information Systems 3
VR display (VRD)
VRserver(VRS)
mmWAVEantenna User information
Data
VideoData
Select
Figure 1 Concept of the proposed VR system exploiting the mmWave-based VR network
3 Maximum Communication Speed of60GHz Wireless Systems with IEEE80211ad Standard Features
This section mathematically calculates the maximum speed(ie upper bound) of wireless data transmission over 60GHzmmWave channels and eventually defines the quality of VRvideo streaming through the following steps (i) calculatingthe received signal strength at VRD (ii) obtaining maximumwireless data transmission speed and (iii) defining the qualityof streamingThemaximumwireless data transmission speedwill be derived with two possible ways (i) using Shannoncapacity equation and (ii) using the level of supportablemod-ulation and coding scheme (MCS) defined in IEEE 80211adspecification and its associated wireless data transmissionspeed
31 Calculation of the Received Signal Strength at VRD Thereceived signal strength in 60GHz mmWave wireless chan-nels that depends on distance between VRS and VRD(denoted by 119889) at VRD in a milli-Watt scale 119875VRD
mW (119889) can becalculated as follows
119875VRDmW (119889) = 119891mW (119875VRD
dBm (119889)) (1)
where119891mW(119909) = 1011990910 and119875VRDdBm (119889) that is the received sig-
nal strength in 60GHz mmWave wireless channels depend-ing on distance between VRS and VRD (denoted by 119889) atVRD in a dB scale can be obtained as follows
119875VRDdBm (119889) = 119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi (2)
where119866VRSdBi 119866VRD
dBi and 119875VRSdBm are the transmit antenna gain at
VRS in a dB scale the receive antenna gain at VRD in a dBscale and the transmit power (assumed to be 19 dBm [17]) atVRS in a dB scale respectively 119871(119889) is the path-loss thatdepends on 119889 In the equation 119866VRS
dBi = 24 dBi according to[17] and we assume 119866VRD
dBi = 0 dBi which corresponds to the
omnidirectional receive antenna at VRD Note that assumingomnidirectional receive antenna at VRD is beneficial becauseit reduces beam training time overhead that is essential formobile mmWave services as clearly discussed in [18]
According to the 60GHz IEEE 80211ad line-of-sight(LOS) path-loss [19] the dB scaled 119871(119889) in (2) can be definedas follows
119871 (119889) = 119860 + 20 log10 (119891GHz) + 10119899 log10 (119889) (3)
where 119860 is a specific value for the selected type of antennaand beamforming algorithm that relies on the antennabeamwidth that is 119860 = 325 dB as defined in [19] In addi-tion the path-loss coefficient 119899 is set to 2 and 119891GHz is thecarrier frequency inGHz thereby119891GHz = 60The direct pathsbetween VRS and VRD have no obstacles thus LOS model isconsidered in this research
Based on the calculation result of the received signalstrength in VRD the interfering signals are added to thedesired transmission through the main wireless link fromVRS to VRD Therefore signal-to-interference plus noiseratio (SINR) in a dB scale denoted by 120574dB can be obtained asfollows
120574dB (119889) = 119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW) (4)
where 119891dB(119909) = 10 sdot log10(119909) and 119899mW is a background noisein a milli-Watt scale that can be derived as follows [20]
119899mW = 119891mW (119896119861119879119890 + 10 log10119861 + 119865119899) (5)
In (5) 119896119861119879119890 is a noise power spectral density whichis minus174 dBmHz [20] 119861 is one subchannel bandwidth in60GHz mmWave which is defined as 216GHz in IEEE80211ad [5] and 119865119899 is a noise figure which is assumed to be5 dB [12] 119878119868 in (4) is a set of the wireless transmissions fromVRS toVRD (those are not ourmain considering VRwirelesscommunications) excluding the aforementioned main video
4 Mobile Information Systems
VRserver
Transmitantenna
VRserver
Transmitantenna
Main video streaming link
Interference
(VRS) A (VRS) B
(lowast lowast)Link i
VR display (VRD) A VR display (VRD) B
Figure 2 Interference scenario with directional transmit antennas
streaming that is a set of interference transmissions Inaddition 119868119894mW means the interference in a milli-Watt scale toour main VRD induced by the neighbor 60GHz mmWavewireless transmissions from links 119894 where 119894 isin 119878119868 In a conse-quence the interference from the 60GHz mmWave wirelesslink 119894 can be expressed as follows
119868119894mW = 119891mW (119866119894VRSdBi (120579lowast 120601lowast) + 119875119894VRSdBm minus 119871 (119889119894)) (6)
where 119875119894VRSdBm stands for the transmit power from the VRS of60GHz mmWave wireless link 119894 and 119871(119889119894) is the path-loss in(3) for 119889119894 that is the distance between the VRS of link 119894 andthe VRD of the main video streaming Lastly 119866119894VRSdBi (120579lowast 120601lowast)is the transmit antenna radiation gain from the transmitterof 60GHz mmWave wireless link 119894 to the VRD of main videostreaming It means that 120579lowast and 120601lowast are the azimuthal and ele-vational angular differences between two links (one link ismain video streaming and the other link is wireless link 119894)
Figure 2 provides an example for the interference sce-nario As illustrated in Figure 2 there exist two parallel VRwireless data transmissions One is from the VRS of mainvideo streaming (named VRS 119860 in Figure 2) to its associatedVRD (named VRD 119860 in Figure 2) of main video streamingand the other one is from VRS of wireless link 119894 (named VRS119861 in Figure 2) to its associated VRD of wireless link 119894 (namedVRD 119861 in Figure 2)120579lowast and 120601lowast are the azimuth and elevation angular differ-ences between the link 119894 and main video streaming respec-tively As illustrated in Figure 2 the directional wirelesstransmission from VRS 119861 to VRD 119861 will be one potentialinterference source to the VRD 119860
32 Obtaining Maximum Wireless Data Transmission SpeedIn order to obtain the maximum wireless data transmissionspeed two possible methods may exist (i) using Shannoncapacity equation (refer to Section 321) and (ii) using thelevel of supportable MCS defined in IEEE 80211ad specifica-tion and its associated wireless data transmission speed (referto Section 322)
321 Using Shannon Capacity Equation Based on the SINRcalculated in (4) in Section 31 the theoretical maximumwireless data transmission speed is formulated as
119861 sdot log2 (120574dB (119889)) (7)where 119861 is one subchannel bandwidth which is 216GHz in60GHz mmWave wireless systems
322 Using the Level of Supportable MCS Defined in IEEE80211ad Specification and Its Associated Wireless Data Trans-mission Speed This subsection uses 120574dB(119889) which resultedfrom the previous step in Section 31 The derived 120574dB(119889) willbe compared with the receiver sensitivity defined in IEEE80211ad specification [12]The receiver sensitivity values andtheir associated supportable wireless data transmissionspeeds are summarized in Table 1 [12] However the directcomparison between 119903dB(119889) and the values in Table 1 [12] maynot be possible Instead the values should be converted tosignal-to-noise ratio (SNR) as
SNR119896dB = [Receiver-Sensitivity]119896 minus 119899dBm (8)
where [Receiver-Sensitivity]119896 is a receiver sensitivity of MCSlevel 119896 SNR119896dB is the associated corresponding SNR of the
Mobile Information Systems 5
Table 1 IEEE 80211ad single-carrier MCS table [12]
Receiver Selected MCS indicesSensitivity (Data rates unit Mbps)minus78 dBm MCS0 (275)minus68 dBm MCS1 (385)minus66 dBm MCS2 (770)minus65 dBm MCS3 (9625)minus64 dBm MCS4 (1155)minus63 dBm MCS6 (1540)minus62 dBm MCS7 (1925)minus61 dBm MCS8 (2310)minus59 dBm MCS9 (25025)minus55 dBm MCS10 (3080)minus54 dBm MCS11 (3850)minus53 dBm MCS12 (4620)
receiver sensitivity of MCS level 119896 and 119899dBm is a backgroundnoise in a dB scale as previously presented in (5) When theobtained 120574dB(119889) value in Section 31 is in between SNR119894dB (iethe SNR of MCS index 119894 (lower)) and SNR119895dB (ie the SNRof MCS index 119895 (higher)) the 60GHz mmWave wireless linkof the main video streaming uses the MCS index 119894 Once wedetermine the supportable MCS level the associated wirelessdata transmission speed is obtained from Table 1
33 Quality Estimation of Wireless Video Streaming [12] Inorder to estimate the quality of the wireless video streamingthat depends on the derived maximum wireless data trans-mission speed we use the model presented in [5 12] Thissection summaries the discussion in [5 12] The wirelesstransmission of 1080 p 30 fps (30 frames each of which has1080-by-1920 pixels will be presented in a display per second)in RGB (8 bits for each RedGreenBlue color representationie 8 times 3 = 24 bits) video frames with no compression (nosource coding but only channel coding exists) requires the15-gigabits (Gbps) wireless data transmission speed sinceeach frame has 1920times1080 pixels with 30 frames per secondand each pixel has 24 bits in an RGB format Similarly thewireless data transmission speed of 3Gbps is at least requiredfor the full-quality 1080 60 fps streaming in RGB (ie1080times1920times24times60) In YCbCr 4 2 0 formats the requiredwireless data transmission speed is half of the wireless datatransmission speed in RGB formats that is 075Gbps and15Gbps speeds are required for nonsource-coded real-time1080 p 30 fps and 1080 p 60 fps video streaming asintroduced If themaximumwireless data transmission speedcalculated and obtained in Sections 31 and 32 is less than thefull-quality threshold the quality may degrade One widelysuggested reference model of the quality is as [5 12]
119891119902 (119909) = 1
log120572 (119909max + 1) log120572 (119909 + 1) if 119909 lt 119909max119909max otherwise (9)
where 120572 is a base (1 lt 120572) 119909max is a desired wireless datatransmission speed for the uncompressed streaming (ie
075Gbps in 1080 30 fps in YCbCr 4 2 0 formats 15 Gbpsin 1080 p 60 fps in YCbCr 4 2 0 formats 15 Gbps in1080 p 30 fps in RGB formats and 30Gbps in 1080 p 60 fps in RGB formats) and 119909 is the achievable wireless datatransmission speed derived in Sections 31 and 32
4 Proposed Strategic Control of 60GHzWireless Communications in Mobile VirtualReality Applications
41 Design Rationale For each wireless transmission fromVRS to VRD if the transmission queue in VRS is almostempty the transmission algorithm should not need to allocatea lot of transmit power in order to transmit more bitsfrom the queue Then the algorithm will allocate relativelylittle amount of powers into the transmitter in order towork in the energy-efficient manner On the other hand ifthe transmission queue in VRS is almost occupied byVR data information (almost in the overflow situations)the transmission algorithm should allocate relatively largeamount of powers into the transmitter in order to avoid thequeue-overflow situations Therefore this section proposesa stochastic optimization framework that works for theminimization of time-average expected power consumption(which is equivalent to the maximization of time-averageexpected energy-efficiency) while preserving queue stabi-lization that depends on the queue-backlog sizes in each unittime
42 Strategic Control of 60GHz Wireless Virtual RealityApplications For each VRS V119894 isin V where V is the set ofexisting wireless links between VRSs and VRDs the queuedynamics in unit time 119905 isin 0 1 can be formulated asfollows
119876119894 [119905 + 1] = (119876119894 [119905] minus 120583119894 [119905] + 120582119894 [119905])+ 119876119894 [0] = 0 (10)
where (119886)+ ≜ max119886 0 and 119876119894[119905] 120583119894[119905] and 120582119894[119905] mean thequeue backlog at VRS forallV119894 isin V at 119905 the departure process(ie wireless transmission of VR videos to VRD) from VRSforallV119894 isin V at 119905 and the arrival process (ie VR video chunkplacement) to VRS forallV119894 at 119905 respectively The unit of 119876119894[119905]120583119894[119905] and 120582119894[119905] is bit Here 120583119894[119905] and 120582119894[119905]mean the amountsof transmitted bits from VRS forallV119894 isin V to its associatedVRD at unit time 119905 over 60GHz mmWave wireless channelsand the generated data in VRS forallV119894 isin V to be transmittedto its associated VRD at unit time 119905 over 60GHz mmWavewireless channels respectively Notice that the arrival processis independent of power allocation decision
In order to formulate 120583119894[119905] we consider the case wherebythe maximum wireless data transmission speed is obtainedfrom the Shannon capacity equation Then according to (1)and (2) it is obvious that120583119894 [119905] = 119861 sdot log2 (120574dB (119889)) (11)
= 119861 sdot log2 (119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW)) (12)
6 Mobile Information Systems
Parameters(i) 119881 a parameter for power-delay (queueing delay) tradeoffs(ii) 119861 channel bandwidth (216GHz in 60GHz mmWave wireless systems)(iii) 119899mW background noise(iv) 119866VRS
dBi transmit antenna gain at VRS(v)P a set of possible power allocation candidates(vi) 119878119868 a set of possible interference sources
Stochastic Power Allocation for the Transceiver of VR Server (VRS)119905 = 0 119879max the number of discrete-time operationwhile 119905 le 119879max do
(i) Observes 119876119894[119905] 119889 and 119868119894mW(ii) Finds stochastic optimal power allocation 119875VRS
119894dBm[119905] inP which can minimize
119875VRS119894dBm[119905] minus 119881119861119876119894 [119905] log2 (119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm[119905] minus 119871(119889))
119899mW + sum119894isin119878119868 119868119894mW))
Algorithm 1 Stochastic power allocation for the minimization of time-average expected power consumption subject to queue-stabilizationin the transceiver of virtual reality (VR) servers
= 119861sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi )119899mW + sum119894isin119878119868 119868119894mW
)) (13)
= 119861 sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
dBm minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (14)
Note that 119866VRDdBi is disappeared in (14) because 119866VRD
dBi =0 dBi (again we supposed the omnidirectional receiverantenna setting)
The mathematical program to minimize the summationof the time-average expected power consumption (which isobviously equivalent to the maximization of the summationof time-average expected energy-efficiency) is as follows
min sumV119894isinV
E [119875VRS119894dBm] (15)
where E[119875VRS119894dBm] stands for the time-average expected power
consumption at V119894 isin V and this is equivalent to
min sumV119894isinV
( lim119905rarrinfin
1119905119905minus1sum120591=0
119875VRS119894dBm [120591]) (16)
Note the objective function in our stochastic optimizationframework has the following two constraints (i) a constraintfor queue rate stability and (ii) a constraint for discretizedpower allocation due to radio frequency (RF) hardwaredesign limitations (ie it is impossible to allocation real-number scaled transmit powers) The first can be expressedas
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]]120591 = 0 forallV119894 isin V (17)
and we can rewrite (17) in this formulation
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]] lt infin forallV119894 isin V (18)
The constraint for discretized power allocation can beexpressed as
119875VRS119894dBm [119905] isin P ≜ 119901min 1199011 1199012 119901max (19)
Now let us introduce a new variable Θ(119905) that denotesthe column vector of all queues in VRSes at unit time 119905 anditeratively define the quadratic Lyapunov function 119871[119905] asfollows
119871 [119905] = 12Θ119879 [119905] Θ [119905] = 12 sumV119894isinV
(119876119894 [119905])2 (20)
where Θ119879[119905] denotes the transpose of Θ[119905] Then let usintroduce another new variable Δ[119905] that is a conditionalquadratic Lyapunov function that may be formulated asfollows
Δ [119905] ≜ E [119871 [119905 + 1] minus 119871 [119905] | Θ [119905]] (21)
that is the drift on unit time 119905 The decision-making policyof the dynamic and stochastic power allocation is designedto (i) solve the minimization of time-average expected powerconsumption formulation by observing the current queue-backlog sizes 119876119894[119905] and (ii) iteratively determine the amountof power allocation 119875VRS
119894dBm for maximizing a bound on
E [119875VRS [119905] | Θ [119905]] minus 119881Δ [119905] (22)
where 119875VRS[119905] is the column vector of 119875VRS119894dBm[119905] forallV119894 isin V
and 119881 is the positive constant control parameter setting ofthe dynamicstochastic decision-making policy that affectsthe power-delay (ie queueing delay) tradeoffs
The proposed algorithm involves minimizing a bound onE[119875VRS[119905] | Θ[119905]] minus 119881Δ[119905] and finally this Lyapunov opti-mization theory gives Algorithm 1 to minimize the followingequation
sumV119894isinV
119875VRS119894dBm [119905] + 119881 sum
V119894isinV119876119894 [119905] (120582119894 [119905] minus 120583119894 [119905]) (23)
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 3
VR display (VRD)
VRserver(VRS)
mmWAVEantenna User information
Data
VideoData
Select
Figure 1 Concept of the proposed VR system exploiting the mmWave-based VR network
3 Maximum Communication Speed of60GHz Wireless Systems with IEEE80211ad Standard Features
This section mathematically calculates the maximum speed(ie upper bound) of wireless data transmission over 60GHzmmWave channels and eventually defines the quality of VRvideo streaming through the following steps (i) calculatingthe received signal strength at VRD (ii) obtaining maximumwireless data transmission speed and (iii) defining the qualityof streamingThemaximumwireless data transmission speedwill be derived with two possible ways (i) using Shannoncapacity equation and (ii) using the level of supportablemod-ulation and coding scheme (MCS) defined in IEEE 80211adspecification and its associated wireless data transmissionspeed
31 Calculation of the Received Signal Strength at VRD Thereceived signal strength in 60GHz mmWave wireless chan-nels that depends on distance between VRS and VRD(denoted by 119889) at VRD in a milli-Watt scale 119875VRD
mW (119889) can becalculated as follows
119875VRDmW (119889) = 119891mW (119875VRD
dBm (119889)) (1)
where119891mW(119909) = 1011990910 and119875VRDdBm (119889) that is the received sig-
nal strength in 60GHz mmWave wireless channels depend-ing on distance between VRS and VRD (denoted by 119889) atVRD in a dB scale can be obtained as follows
119875VRDdBm (119889) = 119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi (2)
where119866VRSdBi 119866VRD
dBi and 119875VRSdBm are the transmit antenna gain at
VRS in a dB scale the receive antenna gain at VRD in a dBscale and the transmit power (assumed to be 19 dBm [17]) atVRS in a dB scale respectively 119871(119889) is the path-loss thatdepends on 119889 In the equation 119866VRS
dBi = 24 dBi according to[17] and we assume 119866VRD
dBi = 0 dBi which corresponds to the
omnidirectional receive antenna at VRD Note that assumingomnidirectional receive antenna at VRD is beneficial becauseit reduces beam training time overhead that is essential formobile mmWave services as clearly discussed in [18]
According to the 60GHz IEEE 80211ad line-of-sight(LOS) path-loss [19] the dB scaled 119871(119889) in (2) can be definedas follows
119871 (119889) = 119860 + 20 log10 (119891GHz) + 10119899 log10 (119889) (3)
where 119860 is a specific value for the selected type of antennaand beamforming algorithm that relies on the antennabeamwidth that is 119860 = 325 dB as defined in [19] In addi-tion the path-loss coefficient 119899 is set to 2 and 119891GHz is thecarrier frequency inGHz thereby119891GHz = 60The direct pathsbetween VRS and VRD have no obstacles thus LOS model isconsidered in this research
Based on the calculation result of the received signalstrength in VRD the interfering signals are added to thedesired transmission through the main wireless link fromVRS to VRD Therefore signal-to-interference plus noiseratio (SINR) in a dB scale denoted by 120574dB can be obtained asfollows
120574dB (119889) = 119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW) (4)
where 119891dB(119909) = 10 sdot log10(119909) and 119899mW is a background noisein a milli-Watt scale that can be derived as follows [20]
119899mW = 119891mW (119896119861119879119890 + 10 log10119861 + 119865119899) (5)
In (5) 119896119861119879119890 is a noise power spectral density whichis minus174 dBmHz [20] 119861 is one subchannel bandwidth in60GHz mmWave which is defined as 216GHz in IEEE80211ad [5] and 119865119899 is a noise figure which is assumed to be5 dB [12] 119878119868 in (4) is a set of the wireless transmissions fromVRS toVRD (those are not ourmain considering VRwirelesscommunications) excluding the aforementioned main video
4 Mobile Information Systems
VRserver
Transmitantenna
VRserver
Transmitantenna
Main video streaming link
Interference
(VRS) A (VRS) B
(lowast lowast)Link i
VR display (VRD) A VR display (VRD) B
Figure 2 Interference scenario with directional transmit antennas
streaming that is a set of interference transmissions Inaddition 119868119894mW means the interference in a milli-Watt scale toour main VRD induced by the neighbor 60GHz mmWavewireless transmissions from links 119894 where 119894 isin 119878119868 In a conse-quence the interference from the 60GHz mmWave wirelesslink 119894 can be expressed as follows
119868119894mW = 119891mW (119866119894VRSdBi (120579lowast 120601lowast) + 119875119894VRSdBm minus 119871 (119889119894)) (6)
where 119875119894VRSdBm stands for the transmit power from the VRS of60GHz mmWave wireless link 119894 and 119871(119889119894) is the path-loss in(3) for 119889119894 that is the distance between the VRS of link 119894 andthe VRD of the main video streaming Lastly 119866119894VRSdBi (120579lowast 120601lowast)is the transmit antenna radiation gain from the transmitterof 60GHz mmWave wireless link 119894 to the VRD of main videostreaming It means that 120579lowast and 120601lowast are the azimuthal and ele-vational angular differences between two links (one link ismain video streaming and the other link is wireless link 119894)
Figure 2 provides an example for the interference sce-nario As illustrated in Figure 2 there exist two parallel VRwireless data transmissions One is from the VRS of mainvideo streaming (named VRS 119860 in Figure 2) to its associatedVRD (named VRD 119860 in Figure 2) of main video streamingand the other one is from VRS of wireless link 119894 (named VRS119861 in Figure 2) to its associated VRD of wireless link 119894 (namedVRD 119861 in Figure 2)120579lowast and 120601lowast are the azimuth and elevation angular differ-ences between the link 119894 and main video streaming respec-tively As illustrated in Figure 2 the directional wirelesstransmission from VRS 119861 to VRD 119861 will be one potentialinterference source to the VRD 119860
32 Obtaining Maximum Wireless Data Transmission SpeedIn order to obtain the maximum wireless data transmissionspeed two possible methods may exist (i) using Shannoncapacity equation (refer to Section 321) and (ii) using thelevel of supportable MCS defined in IEEE 80211ad specifica-tion and its associated wireless data transmission speed (referto Section 322)
321 Using Shannon Capacity Equation Based on the SINRcalculated in (4) in Section 31 the theoretical maximumwireless data transmission speed is formulated as
119861 sdot log2 (120574dB (119889)) (7)where 119861 is one subchannel bandwidth which is 216GHz in60GHz mmWave wireless systems
322 Using the Level of Supportable MCS Defined in IEEE80211ad Specification and Its Associated Wireless Data Trans-mission Speed This subsection uses 120574dB(119889) which resultedfrom the previous step in Section 31 The derived 120574dB(119889) willbe compared with the receiver sensitivity defined in IEEE80211ad specification [12]The receiver sensitivity values andtheir associated supportable wireless data transmissionspeeds are summarized in Table 1 [12] However the directcomparison between 119903dB(119889) and the values in Table 1 [12] maynot be possible Instead the values should be converted tosignal-to-noise ratio (SNR) as
SNR119896dB = [Receiver-Sensitivity]119896 minus 119899dBm (8)
where [Receiver-Sensitivity]119896 is a receiver sensitivity of MCSlevel 119896 SNR119896dB is the associated corresponding SNR of the
Mobile Information Systems 5
Table 1 IEEE 80211ad single-carrier MCS table [12]
Receiver Selected MCS indicesSensitivity (Data rates unit Mbps)minus78 dBm MCS0 (275)minus68 dBm MCS1 (385)minus66 dBm MCS2 (770)minus65 dBm MCS3 (9625)minus64 dBm MCS4 (1155)minus63 dBm MCS6 (1540)minus62 dBm MCS7 (1925)minus61 dBm MCS8 (2310)minus59 dBm MCS9 (25025)minus55 dBm MCS10 (3080)minus54 dBm MCS11 (3850)minus53 dBm MCS12 (4620)
receiver sensitivity of MCS level 119896 and 119899dBm is a backgroundnoise in a dB scale as previously presented in (5) When theobtained 120574dB(119889) value in Section 31 is in between SNR119894dB (iethe SNR of MCS index 119894 (lower)) and SNR119895dB (ie the SNRof MCS index 119895 (higher)) the 60GHz mmWave wireless linkof the main video streaming uses the MCS index 119894 Once wedetermine the supportable MCS level the associated wirelessdata transmission speed is obtained from Table 1
33 Quality Estimation of Wireless Video Streaming [12] Inorder to estimate the quality of the wireless video streamingthat depends on the derived maximum wireless data trans-mission speed we use the model presented in [5 12] Thissection summaries the discussion in [5 12] The wirelesstransmission of 1080 p 30 fps (30 frames each of which has1080-by-1920 pixels will be presented in a display per second)in RGB (8 bits for each RedGreenBlue color representationie 8 times 3 = 24 bits) video frames with no compression (nosource coding but only channel coding exists) requires the15-gigabits (Gbps) wireless data transmission speed sinceeach frame has 1920times1080 pixels with 30 frames per secondand each pixel has 24 bits in an RGB format Similarly thewireless data transmission speed of 3Gbps is at least requiredfor the full-quality 1080 60 fps streaming in RGB (ie1080times1920times24times60) In YCbCr 4 2 0 formats the requiredwireless data transmission speed is half of the wireless datatransmission speed in RGB formats that is 075Gbps and15Gbps speeds are required for nonsource-coded real-time1080 p 30 fps and 1080 p 60 fps video streaming asintroduced If themaximumwireless data transmission speedcalculated and obtained in Sections 31 and 32 is less than thefull-quality threshold the quality may degrade One widelysuggested reference model of the quality is as [5 12]
119891119902 (119909) = 1
log120572 (119909max + 1) log120572 (119909 + 1) if 119909 lt 119909max119909max otherwise (9)
where 120572 is a base (1 lt 120572) 119909max is a desired wireless datatransmission speed for the uncompressed streaming (ie
075Gbps in 1080 30 fps in YCbCr 4 2 0 formats 15 Gbpsin 1080 p 60 fps in YCbCr 4 2 0 formats 15 Gbps in1080 p 30 fps in RGB formats and 30Gbps in 1080 p 60 fps in RGB formats) and 119909 is the achievable wireless datatransmission speed derived in Sections 31 and 32
4 Proposed Strategic Control of 60GHzWireless Communications in Mobile VirtualReality Applications
41 Design Rationale For each wireless transmission fromVRS to VRD if the transmission queue in VRS is almostempty the transmission algorithm should not need to allocatea lot of transmit power in order to transmit more bitsfrom the queue Then the algorithm will allocate relativelylittle amount of powers into the transmitter in order towork in the energy-efficient manner On the other hand ifthe transmission queue in VRS is almost occupied byVR data information (almost in the overflow situations)the transmission algorithm should allocate relatively largeamount of powers into the transmitter in order to avoid thequeue-overflow situations Therefore this section proposesa stochastic optimization framework that works for theminimization of time-average expected power consumption(which is equivalent to the maximization of time-averageexpected energy-efficiency) while preserving queue stabi-lization that depends on the queue-backlog sizes in each unittime
42 Strategic Control of 60GHz Wireless Virtual RealityApplications For each VRS V119894 isin V where V is the set ofexisting wireless links between VRSs and VRDs the queuedynamics in unit time 119905 isin 0 1 can be formulated asfollows
119876119894 [119905 + 1] = (119876119894 [119905] minus 120583119894 [119905] + 120582119894 [119905])+ 119876119894 [0] = 0 (10)
where (119886)+ ≜ max119886 0 and 119876119894[119905] 120583119894[119905] and 120582119894[119905] mean thequeue backlog at VRS forallV119894 isin V at 119905 the departure process(ie wireless transmission of VR videos to VRD) from VRSforallV119894 isin V at 119905 and the arrival process (ie VR video chunkplacement) to VRS forallV119894 at 119905 respectively The unit of 119876119894[119905]120583119894[119905] and 120582119894[119905] is bit Here 120583119894[119905] and 120582119894[119905]mean the amountsof transmitted bits from VRS forallV119894 isin V to its associatedVRD at unit time 119905 over 60GHz mmWave wireless channelsand the generated data in VRS forallV119894 isin V to be transmittedto its associated VRD at unit time 119905 over 60GHz mmWavewireless channels respectively Notice that the arrival processis independent of power allocation decision
In order to formulate 120583119894[119905] we consider the case wherebythe maximum wireless data transmission speed is obtainedfrom the Shannon capacity equation Then according to (1)and (2) it is obvious that120583119894 [119905] = 119861 sdot log2 (120574dB (119889)) (11)
= 119861 sdot log2 (119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW)) (12)
6 Mobile Information Systems
Parameters(i) 119881 a parameter for power-delay (queueing delay) tradeoffs(ii) 119861 channel bandwidth (216GHz in 60GHz mmWave wireless systems)(iii) 119899mW background noise(iv) 119866VRS
dBi transmit antenna gain at VRS(v)P a set of possible power allocation candidates(vi) 119878119868 a set of possible interference sources
Stochastic Power Allocation for the Transceiver of VR Server (VRS)119905 = 0 119879max the number of discrete-time operationwhile 119905 le 119879max do
(i) Observes 119876119894[119905] 119889 and 119868119894mW(ii) Finds stochastic optimal power allocation 119875VRS
119894dBm[119905] inP which can minimize
119875VRS119894dBm[119905] minus 119881119861119876119894 [119905] log2 (119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm[119905] minus 119871(119889))
119899mW + sum119894isin119878119868 119868119894mW))
Algorithm 1 Stochastic power allocation for the minimization of time-average expected power consumption subject to queue-stabilizationin the transceiver of virtual reality (VR) servers
= 119861sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi )119899mW + sum119894isin119878119868 119868119894mW
)) (13)
= 119861 sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
dBm minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (14)
Note that 119866VRDdBi is disappeared in (14) because 119866VRD
dBi =0 dBi (again we supposed the omnidirectional receiverantenna setting)
The mathematical program to minimize the summationof the time-average expected power consumption (which isobviously equivalent to the maximization of the summationof time-average expected energy-efficiency) is as follows
min sumV119894isinV
E [119875VRS119894dBm] (15)
where E[119875VRS119894dBm] stands for the time-average expected power
consumption at V119894 isin V and this is equivalent to
min sumV119894isinV
( lim119905rarrinfin
1119905119905minus1sum120591=0
119875VRS119894dBm [120591]) (16)
Note the objective function in our stochastic optimizationframework has the following two constraints (i) a constraintfor queue rate stability and (ii) a constraint for discretizedpower allocation due to radio frequency (RF) hardwaredesign limitations (ie it is impossible to allocation real-number scaled transmit powers) The first can be expressedas
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]]120591 = 0 forallV119894 isin V (17)
and we can rewrite (17) in this formulation
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]] lt infin forallV119894 isin V (18)
The constraint for discretized power allocation can beexpressed as
119875VRS119894dBm [119905] isin P ≜ 119901min 1199011 1199012 119901max (19)
Now let us introduce a new variable Θ(119905) that denotesthe column vector of all queues in VRSes at unit time 119905 anditeratively define the quadratic Lyapunov function 119871[119905] asfollows
119871 [119905] = 12Θ119879 [119905] Θ [119905] = 12 sumV119894isinV
(119876119894 [119905])2 (20)
where Θ119879[119905] denotes the transpose of Θ[119905] Then let usintroduce another new variable Δ[119905] that is a conditionalquadratic Lyapunov function that may be formulated asfollows
Δ [119905] ≜ E [119871 [119905 + 1] minus 119871 [119905] | Θ [119905]] (21)
that is the drift on unit time 119905 The decision-making policyof the dynamic and stochastic power allocation is designedto (i) solve the minimization of time-average expected powerconsumption formulation by observing the current queue-backlog sizes 119876119894[119905] and (ii) iteratively determine the amountof power allocation 119875VRS
119894dBm for maximizing a bound on
E [119875VRS [119905] | Θ [119905]] minus 119881Δ [119905] (22)
where 119875VRS[119905] is the column vector of 119875VRS119894dBm[119905] forallV119894 isin V
and 119881 is the positive constant control parameter setting ofthe dynamicstochastic decision-making policy that affectsthe power-delay (ie queueing delay) tradeoffs
The proposed algorithm involves minimizing a bound onE[119875VRS[119905] | Θ[119905]] minus 119881Δ[119905] and finally this Lyapunov opti-mization theory gives Algorithm 1 to minimize the followingequation
sumV119894isinV
119875VRS119894dBm [119905] + 119881 sum
V119894isinV119876119894 [119905] (120582119894 [119905] minus 120583119894 [119905]) (23)
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
4 Mobile Information Systems
VRserver
Transmitantenna
VRserver
Transmitantenna
Main video streaming link
Interference
(VRS) A (VRS) B
(lowast lowast)Link i
VR display (VRD) A VR display (VRD) B
Figure 2 Interference scenario with directional transmit antennas
streaming that is a set of interference transmissions Inaddition 119868119894mW means the interference in a milli-Watt scale toour main VRD induced by the neighbor 60GHz mmWavewireless transmissions from links 119894 where 119894 isin 119878119868 In a conse-quence the interference from the 60GHz mmWave wirelesslink 119894 can be expressed as follows
119868119894mW = 119891mW (119866119894VRSdBi (120579lowast 120601lowast) + 119875119894VRSdBm minus 119871 (119889119894)) (6)
where 119875119894VRSdBm stands for the transmit power from the VRS of60GHz mmWave wireless link 119894 and 119871(119889119894) is the path-loss in(3) for 119889119894 that is the distance between the VRS of link 119894 andthe VRD of the main video streaming Lastly 119866119894VRSdBi (120579lowast 120601lowast)is the transmit antenna radiation gain from the transmitterof 60GHz mmWave wireless link 119894 to the VRD of main videostreaming It means that 120579lowast and 120601lowast are the azimuthal and ele-vational angular differences between two links (one link ismain video streaming and the other link is wireless link 119894)
Figure 2 provides an example for the interference sce-nario As illustrated in Figure 2 there exist two parallel VRwireless data transmissions One is from the VRS of mainvideo streaming (named VRS 119860 in Figure 2) to its associatedVRD (named VRD 119860 in Figure 2) of main video streamingand the other one is from VRS of wireless link 119894 (named VRS119861 in Figure 2) to its associated VRD of wireless link 119894 (namedVRD 119861 in Figure 2)120579lowast and 120601lowast are the azimuth and elevation angular differ-ences between the link 119894 and main video streaming respec-tively As illustrated in Figure 2 the directional wirelesstransmission from VRS 119861 to VRD 119861 will be one potentialinterference source to the VRD 119860
32 Obtaining Maximum Wireless Data Transmission SpeedIn order to obtain the maximum wireless data transmissionspeed two possible methods may exist (i) using Shannoncapacity equation (refer to Section 321) and (ii) using thelevel of supportable MCS defined in IEEE 80211ad specifica-tion and its associated wireless data transmission speed (referto Section 322)
321 Using Shannon Capacity Equation Based on the SINRcalculated in (4) in Section 31 the theoretical maximumwireless data transmission speed is formulated as
119861 sdot log2 (120574dB (119889)) (7)where 119861 is one subchannel bandwidth which is 216GHz in60GHz mmWave wireless systems
322 Using the Level of Supportable MCS Defined in IEEE80211ad Specification and Its Associated Wireless Data Trans-mission Speed This subsection uses 120574dB(119889) which resultedfrom the previous step in Section 31 The derived 120574dB(119889) willbe compared with the receiver sensitivity defined in IEEE80211ad specification [12]The receiver sensitivity values andtheir associated supportable wireless data transmissionspeeds are summarized in Table 1 [12] However the directcomparison between 119903dB(119889) and the values in Table 1 [12] maynot be possible Instead the values should be converted tosignal-to-noise ratio (SNR) as
SNR119896dB = [Receiver-Sensitivity]119896 minus 119899dBm (8)
where [Receiver-Sensitivity]119896 is a receiver sensitivity of MCSlevel 119896 SNR119896dB is the associated corresponding SNR of the
Mobile Information Systems 5
Table 1 IEEE 80211ad single-carrier MCS table [12]
Receiver Selected MCS indicesSensitivity (Data rates unit Mbps)minus78 dBm MCS0 (275)minus68 dBm MCS1 (385)minus66 dBm MCS2 (770)minus65 dBm MCS3 (9625)minus64 dBm MCS4 (1155)minus63 dBm MCS6 (1540)minus62 dBm MCS7 (1925)minus61 dBm MCS8 (2310)minus59 dBm MCS9 (25025)minus55 dBm MCS10 (3080)minus54 dBm MCS11 (3850)minus53 dBm MCS12 (4620)
receiver sensitivity of MCS level 119896 and 119899dBm is a backgroundnoise in a dB scale as previously presented in (5) When theobtained 120574dB(119889) value in Section 31 is in between SNR119894dB (iethe SNR of MCS index 119894 (lower)) and SNR119895dB (ie the SNRof MCS index 119895 (higher)) the 60GHz mmWave wireless linkof the main video streaming uses the MCS index 119894 Once wedetermine the supportable MCS level the associated wirelessdata transmission speed is obtained from Table 1
33 Quality Estimation of Wireless Video Streaming [12] Inorder to estimate the quality of the wireless video streamingthat depends on the derived maximum wireless data trans-mission speed we use the model presented in [5 12] Thissection summaries the discussion in [5 12] The wirelesstransmission of 1080 p 30 fps (30 frames each of which has1080-by-1920 pixels will be presented in a display per second)in RGB (8 bits for each RedGreenBlue color representationie 8 times 3 = 24 bits) video frames with no compression (nosource coding but only channel coding exists) requires the15-gigabits (Gbps) wireless data transmission speed sinceeach frame has 1920times1080 pixels with 30 frames per secondand each pixel has 24 bits in an RGB format Similarly thewireless data transmission speed of 3Gbps is at least requiredfor the full-quality 1080 60 fps streaming in RGB (ie1080times1920times24times60) In YCbCr 4 2 0 formats the requiredwireless data transmission speed is half of the wireless datatransmission speed in RGB formats that is 075Gbps and15Gbps speeds are required for nonsource-coded real-time1080 p 30 fps and 1080 p 60 fps video streaming asintroduced If themaximumwireless data transmission speedcalculated and obtained in Sections 31 and 32 is less than thefull-quality threshold the quality may degrade One widelysuggested reference model of the quality is as [5 12]
119891119902 (119909) = 1
log120572 (119909max + 1) log120572 (119909 + 1) if 119909 lt 119909max119909max otherwise (9)
where 120572 is a base (1 lt 120572) 119909max is a desired wireless datatransmission speed for the uncompressed streaming (ie
075Gbps in 1080 30 fps in YCbCr 4 2 0 formats 15 Gbpsin 1080 p 60 fps in YCbCr 4 2 0 formats 15 Gbps in1080 p 30 fps in RGB formats and 30Gbps in 1080 p 60 fps in RGB formats) and 119909 is the achievable wireless datatransmission speed derived in Sections 31 and 32
4 Proposed Strategic Control of 60GHzWireless Communications in Mobile VirtualReality Applications
41 Design Rationale For each wireless transmission fromVRS to VRD if the transmission queue in VRS is almostempty the transmission algorithm should not need to allocatea lot of transmit power in order to transmit more bitsfrom the queue Then the algorithm will allocate relativelylittle amount of powers into the transmitter in order towork in the energy-efficient manner On the other hand ifthe transmission queue in VRS is almost occupied byVR data information (almost in the overflow situations)the transmission algorithm should allocate relatively largeamount of powers into the transmitter in order to avoid thequeue-overflow situations Therefore this section proposesa stochastic optimization framework that works for theminimization of time-average expected power consumption(which is equivalent to the maximization of time-averageexpected energy-efficiency) while preserving queue stabi-lization that depends on the queue-backlog sizes in each unittime
42 Strategic Control of 60GHz Wireless Virtual RealityApplications For each VRS V119894 isin V where V is the set ofexisting wireless links between VRSs and VRDs the queuedynamics in unit time 119905 isin 0 1 can be formulated asfollows
119876119894 [119905 + 1] = (119876119894 [119905] minus 120583119894 [119905] + 120582119894 [119905])+ 119876119894 [0] = 0 (10)
where (119886)+ ≜ max119886 0 and 119876119894[119905] 120583119894[119905] and 120582119894[119905] mean thequeue backlog at VRS forallV119894 isin V at 119905 the departure process(ie wireless transmission of VR videos to VRD) from VRSforallV119894 isin V at 119905 and the arrival process (ie VR video chunkplacement) to VRS forallV119894 at 119905 respectively The unit of 119876119894[119905]120583119894[119905] and 120582119894[119905] is bit Here 120583119894[119905] and 120582119894[119905]mean the amountsof transmitted bits from VRS forallV119894 isin V to its associatedVRD at unit time 119905 over 60GHz mmWave wireless channelsand the generated data in VRS forallV119894 isin V to be transmittedto its associated VRD at unit time 119905 over 60GHz mmWavewireless channels respectively Notice that the arrival processis independent of power allocation decision
In order to formulate 120583119894[119905] we consider the case wherebythe maximum wireless data transmission speed is obtainedfrom the Shannon capacity equation Then according to (1)and (2) it is obvious that120583119894 [119905] = 119861 sdot log2 (120574dB (119889)) (11)
= 119861 sdot log2 (119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW)) (12)
6 Mobile Information Systems
Parameters(i) 119881 a parameter for power-delay (queueing delay) tradeoffs(ii) 119861 channel bandwidth (216GHz in 60GHz mmWave wireless systems)(iii) 119899mW background noise(iv) 119866VRS
dBi transmit antenna gain at VRS(v)P a set of possible power allocation candidates(vi) 119878119868 a set of possible interference sources
Stochastic Power Allocation for the Transceiver of VR Server (VRS)119905 = 0 119879max the number of discrete-time operationwhile 119905 le 119879max do
(i) Observes 119876119894[119905] 119889 and 119868119894mW(ii) Finds stochastic optimal power allocation 119875VRS
119894dBm[119905] inP which can minimize
119875VRS119894dBm[119905] minus 119881119861119876119894 [119905] log2 (119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm[119905] minus 119871(119889))
119899mW + sum119894isin119878119868 119868119894mW))
Algorithm 1 Stochastic power allocation for the minimization of time-average expected power consumption subject to queue-stabilizationin the transceiver of virtual reality (VR) servers
= 119861sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi )119899mW + sum119894isin119878119868 119868119894mW
)) (13)
= 119861 sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
dBm minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (14)
Note that 119866VRDdBi is disappeared in (14) because 119866VRD
dBi =0 dBi (again we supposed the omnidirectional receiverantenna setting)
The mathematical program to minimize the summationof the time-average expected power consumption (which isobviously equivalent to the maximization of the summationof time-average expected energy-efficiency) is as follows
min sumV119894isinV
E [119875VRS119894dBm] (15)
where E[119875VRS119894dBm] stands for the time-average expected power
consumption at V119894 isin V and this is equivalent to
min sumV119894isinV
( lim119905rarrinfin
1119905119905minus1sum120591=0
119875VRS119894dBm [120591]) (16)
Note the objective function in our stochastic optimizationframework has the following two constraints (i) a constraintfor queue rate stability and (ii) a constraint for discretizedpower allocation due to radio frequency (RF) hardwaredesign limitations (ie it is impossible to allocation real-number scaled transmit powers) The first can be expressedas
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]]120591 = 0 forallV119894 isin V (17)
and we can rewrite (17) in this formulation
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]] lt infin forallV119894 isin V (18)
The constraint for discretized power allocation can beexpressed as
119875VRS119894dBm [119905] isin P ≜ 119901min 1199011 1199012 119901max (19)
Now let us introduce a new variable Θ(119905) that denotesthe column vector of all queues in VRSes at unit time 119905 anditeratively define the quadratic Lyapunov function 119871[119905] asfollows
119871 [119905] = 12Θ119879 [119905] Θ [119905] = 12 sumV119894isinV
(119876119894 [119905])2 (20)
where Θ119879[119905] denotes the transpose of Θ[119905] Then let usintroduce another new variable Δ[119905] that is a conditionalquadratic Lyapunov function that may be formulated asfollows
Δ [119905] ≜ E [119871 [119905 + 1] minus 119871 [119905] | Θ [119905]] (21)
that is the drift on unit time 119905 The decision-making policyof the dynamic and stochastic power allocation is designedto (i) solve the minimization of time-average expected powerconsumption formulation by observing the current queue-backlog sizes 119876119894[119905] and (ii) iteratively determine the amountof power allocation 119875VRS
119894dBm for maximizing a bound on
E [119875VRS [119905] | Θ [119905]] minus 119881Δ [119905] (22)
where 119875VRS[119905] is the column vector of 119875VRS119894dBm[119905] forallV119894 isin V
and 119881 is the positive constant control parameter setting ofthe dynamicstochastic decision-making policy that affectsthe power-delay (ie queueing delay) tradeoffs
The proposed algorithm involves minimizing a bound onE[119875VRS[119905] | Θ[119905]] minus 119881Δ[119905] and finally this Lyapunov opti-mization theory gives Algorithm 1 to minimize the followingequation
sumV119894isinV
119875VRS119894dBm [119905] + 119881 sum
V119894isinV119876119894 [119905] (120582119894 [119905] minus 120583119894 [119905]) (23)
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 5
Table 1 IEEE 80211ad single-carrier MCS table [12]
Receiver Selected MCS indicesSensitivity (Data rates unit Mbps)minus78 dBm MCS0 (275)minus68 dBm MCS1 (385)minus66 dBm MCS2 (770)minus65 dBm MCS3 (9625)minus64 dBm MCS4 (1155)minus63 dBm MCS6 (1540)minus62 dBm MCS7 (1925)minus61 dBm MCS8 (2310)minus59 dBm MCS9 (25025)minus55 dBm MCS10 (3080)minus54 dBm MCS11 (3850)minus53 dBm MCS12 (4620)
receiver sensitivity of MCS level 119896 and 119899dBm is a backgroundnoise in a dB scale as previously presented in (5) When theobtained 120574dB(119889) value in Section 31 is in between SNR119894dB (iethe SNR of MCS index 119894 (lower)) and SNR119895dB (ie the SNRof MCS index 119895 (higher)) the 60GHz mmWave wireless linkof the main video streaming uses the MCS index 119894 Once wedetermine the supportable MCS level the associated wirelessdata transmission speed is obtained from Table 1
33 Quality Estimation of Wireless Video Streaming [12] Inorder to estimate the quality of the wireless video streamingthat depends on the derived maximum wireless data trans-mission speed we use the model presented in [5 12] Thissection summaries the discussion in [5 12] The wirelesstransmission of 1080 p 30 fps (30 frames each of which has1080-by-1920 pixels will be presented in a display per second)in RGB (8 bits for each RedGreenBlue color representationie 8 times 3 = 24 bits) video frames with no compression (nosource coding but only channel coding exists) requires the15-gigabits (Gbps) wireless data transmission speed sinceeach frame has 1920times1080 pixels with 30 frames per secondand each pixel has 24 bits in an RGB format Similarly thewireless data transmission speed of 3Gbps is at least requiredfor the full-quality 1080 60 fps streaming in RGB (ie1080times1920times24times60) In YCbCr 4 2 0 formats the requiredwireless data transmission speed is half of the wireless datatransmission speed in RGB formats that is 075Gbps and15Gbps speeds are required for nonsource-coded real-time1080 p 30 fps and 1080 p 60 fps video streaming asintroduced If themaximumwireless data transmission speedcalculated and obtained in Sections 31 and 32 is less than thefull-quality threshold the quality may degrade One widelysuggested reference model of the quality is as [5 12]
119891119902 (119909) = 1
log120572 (119909max + 1) log120572 (119909 + 1) if 119909 lt 119909max119909max otherwise (9)
where 120572 is a base (1 lt 120572) 119909max is a desired wireless datatransmission speed for the uncompressed streaming (ie
075Gbps in 1080 30 fps in YCbCr 4 2 0 formats 15 Gbpsin 1080 p 60 fps in YCbCr 4 2 0 formats 15 Gbps in1080 p 30 fps in RGB formats and 30Gbps in 1080 p 60 fps in RGB formats) and 119909 is the achievable wireless datatransmission speed derived in Sections 31 and 32
4 Proposed Strategic Control of 60GHzWireless Communications in Mobile VirtualReality Applications
41 Design Rationale For each wireless transmission fromVRS to VRD if the transmission queue in VRS is almostempty the transmission algorithm should not need to allocatea lot of transmit power in order to transmit more bitsfrom the queue Then the algorithm will allocate relativelylittle amount of powers into the transmitter in order towork in the energy-efficient manner On the other hand ifthe transmission queue in VRS is almost occupied byVR data information (almost in the overflow situations)the transmission algorithm should allocate relatively largeamount of powers into the transmitter in order to avoid thequeue-overflow situations Therefore this section proposesa stochastic optimization framework that works for theminimization of time-average expected power consumption(which is equivalent to the maximization of time-averageexpected energy-efficiency) while preserving queue stabi-lization that depends on the queue-backlog sizes in each unittime
42 Strategic Control of 60GHz Wireless Virtual RealityApplications For each VRS V119894 isin V where V is the set ofexisting wireless links between VRSs and VRDs the queuedynamics in unit time 119905 isin 0 1 can be formulated asfollows
119876119894 [119905 + 1] = (119876119894 [119905] minus 120583119894 [119905] + 120582119894 [119905])+ 119876119894 [0] = 0 (10)
where (119886)+ ≜ max119886 0 and 119876119894[119905] 120583119894[119905] and 120582119894[119905] mean thequeue backlog at VRS forallV119894 isin V at 119905 the departure process(ie wireless transmission of VR videos to VRD) from VRSforallV119894 isin V at 119905 and the arrival process (ie VR video chunkplacement) to VRS forallV119894 at 119905 respectively The unit of 119876119894[119905]120583119894[119905] and 120582119894[119905] is bit Here 120583119894[119905] and 120582119894[119905]mean the amountsof transmitted bits from VRS forallV119894 isin V to its associatedVRD at unit time 119905 over 60GHz mmWave wireless channelsand the generated data in VRS forallV119894 isin V to be transmittedto its associated VRD at unit time 119905 over 60GHz mmWavewireless channels respectively Notice that the arrival processis independent of power allocation decision
In order to formulate 120583119894[119905] we consider the case wherebythe maximum wireless data transmission speed is obtainedfrom the Shannon capacity equation Then according to (1)and (2) it is obvious that120583119894 [119905] = 119861 sdot log2 (120574dB (119889)) (11)
= 119861 sdot log2 (119891dB ( 119875VRDmW (119889)
119899mW + sum119894isin119878119868 119868119894mW)) (12)
6 Mobile Information Systems
Parameters(i) 119881 a parameter for power-delay (queueing delay) tradeoffs(ii) 119861 channel bandwidth (216GHz in 60GHz mmWave wireless systems)(iii) 119899mW background noise(iv) 119866VRS
dBi transmit antenna gain at VRS(v)P a set of possible power allocation candidates(vi) 119878119868 a set of possible interference sources
Stochastic Power Allocation for the Transceiver of VR Server (VRS)119905 = 0 119879max the number of discrete-time operationwhile 119905 le 119879max do
(i) Observes 119876119894[119905] 119889 and 119868119894mW(ii) Finds stochastic optimal power allocation 119875VRS
119894dBm[119905] inP which can minimize
119875VRS119894dBm[119905] minus 119881119861119876119894 [119905] log2 (119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm[119905] minus 119871(119889))
119899mW + sum119894isin119878119868 119868119894mW))
Algorithm 1 Stochastic power allocation for the minimization of time-average expected power consumption subject to queue-stabilizationin the transceiver of virtual reality (VR) servers
= 119861sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi )119899mW + sum119894isin119878119868 119868119894mW
)) (13)
= 119861 sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
dBm minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (14)
Note that 119866VRDdBi is disappeared in (14) because 119866VRD
dBi =0 dBi (again we supposed the omnidirectional receiverantenna setting)
The mathematical program to minimize the summationof the time-average expected power consumption (which isobviously equivalent to the maximization of the summationof time-average expected energy-efficiency) is as follows
min sumV119894isinV
E [119875VRS119894dBm] (15)
where E[119875VRS119894dBm] stands for the time-average expected power
consumption at V119894 isin V and this is equivalent to
min sumV119894isinV
( lim119905rarrinfin
1119905119905minus1sum120591=0
119875VRS119894dBm [120591]) (16)
Note the objective function in our stochastic optimizationframework has the following two constraints (i) a constraintfor queue rate stability and (ii) a constraint for discretizedpower allocation due to radio frequency (RF) hardwaredesign limitations (ie it is impossible to allocation real-number scaled transmit powers) The first can be expressedas
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]]120591 = 0 forallV119894 isin V (17)
and we can rewrite (17) in this formulation
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]] lt infin forallV119894 isin V (18)
The constraint for discretized power allocation can beexpressed as
119875VRS119894dBm [119905] isin P ≜ 119901min 1199011 1199012 119901max (19)
Now let us introduce a new variable Θ(119905) that denotesthe column vector of all queues in VRSes at unit time 119905 anditeratively define the quadratic Lyapunov function 119871[119905] asfollows
119871 [119905] = 12Θ119879 [119905] Θ [119905] = 12 sumV119894isinV
(119876119894 [119905])2 (20)
where Θ119879[119905] denotes the transpose of Θ[119905] Then let usintroduce another new variable Δ[119905] that is a conditionalquadratic Lyapunov function that may be formulated asfollows
Δ [119905] ≜ E [119871 [119905 + 1] minus 119871 [119905] | Θ [119905]] (21)
that is the drift on unit time 119905 The decision-making policyof the dynamic and stochastic power allocation is designedto (i) solve the minimization of time-average expected powerconsumption formulation by observing the current queue-backlog sizes 119876119894[119905] and (ii) iteratively determine the amountof power allocation 119875VRS
119894dBm for maximizing a bound on
E [119875VRS [119905] | Θ [119905]] minus 119881Δ [119905] (22)
where 119875VRS[119905] is the column vector of 119875VRS119894dBm[119905] forallV119894 isin V
and 119881 is the positive constant control parameter setting ofthe dynamicstochastic decision-making policy that affectsthe power-delay (ie queueing delay) tradeoffs
The proposed algorithm involves minimizing a bound onE[119875VRS[119905] | Θ[119905]] minus 119881Δ[119905] and finally this Lyapunov opti-mization theory gives Algorithm 1 to minimize the followingequation
sumV119894isinV
119875VRS119894dBm [119905] + 119881 sum
V119894isinV119876119894 [119905] (120582119894 [119905] minus 120583119894 [119905]) (23)
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
6 Mobile Information Systems
Parameters(i) 119881 a parameter for power-delay (queueing delay) tradeoffs(ii) 119861 channel bandwidth (216GHz in 60GHz mmWave wireless systems)(iii) 119899mW background noise(iv) 119866VRS
dBi transmit antenna gain at VRS(v)P a set of possible power allocation candidates(vi) 119878119868 a set of possible interference sources
Stochastic Power Allocation for the Transceiver of VR Server (VRS)119905 = 0 119879max the number of discrete-time operationwhile 119905 le 119879max do
(i) Observes 119876119894[119905] 119889 and 119868119894mW(ii) Finds stochastic optimal power allocation 119875VRS
119894dBm[119905] inP which can minimize
119875VRS119894dBm[119905] minus 119881119861119876119894 [119905] log2 (119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm[119905] minus 119871(119889))
119899mW + sum119894isin119878119868 119868119894mW))
Algorithm 1 Stochastic power allocation for the minimization of time-average expected power consumption subject to queue-stabilizationin the transceiver of virtual reality (VR) servers
= 119861sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRSdBm minus 119871 (119889) + 119866VRD
dBi )119899mW + sum119894isin119878119868 119868119894mW
)) (13)
= 119861 sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
dBm minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (14)
Note that 119866VRDdBi is disappeared in (14) because 119866VRD
dBi =0 dBi (again we supposed the omnidirectional receiverantenna setting)
The mathematical program to minimize the summationof the time-average expected power consumption (which isobviously equivalent to the maximization of the summationof time-average expected energy-efficiency) is as follows
min sumV119894isinV
E [119875VRS119894dBm] (15)
where E[119875VRS119894dBm] stands for the time-average expected power
consumption at V119894 isin V and this is equivalent to
min sumV119894isinV
( lim119905rarrinfin
1119905119905minus1sum120591=0
119875VRS119894dBm [120591]) (16)
Note the objective function in our stochastic optimizationframework has the following two constraints (i) a constraintfor queue rate stability and (ii) a constraint for discretizedpower allocation due to radio frequency (RF) hardwaredesign limitations (ie it is impossible to allocation real-number scaled transmit powers) The first can be expressedas
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]]120591 = 0 forallV119894 isin V (17)
and we can rewrite (17) in this formulation
lim119905rarrinfin
1119905119905minus1sum120591=0
E [119876119894 [120591]] lt infin forallV119894 isin V (18)
The constraint for discretized power allocation can beexpressed as
119875VRS119894dBm [119905] isin P ≜ 119901min 1199011 1199012 119901max (19)
Now let us introduce a new variable Θ(119905) that denotesthe column vector of all queues in VRSes at unit time 119905 anditeratively define the quadratic Lyapunov function 119871[119905] asfollows
119871 [119905] = 12Θ119879 [119905] Θ [119905] = 12 sumV119894isinV
(119876119894 [119905])2 (20)
where Θ119879[119905] denotes the transpose of Θ[119905] Then let usintroduce another new variable Δ[119905] that is a conditionalquadratic Lyapunov function that may be formulated asfollows
Δ [119905] ≜ E [119871 [119905 + 1] minus 119871 [119905] | Θ [119905]] (21)
that is the drift on unit time 119905 The decision-making policyof the dynamic and stochastic power allocation is designedto (i) solve the minimization of time-average expected powerconsumption formulation by observing the current queue-backlog sizes 119876119894[119905] and (ii) iteratively determine the amountof power allocation 119875VRS
119894dBm for maximizing a bound on
E [119875VRS [119905] | Θ [119905]] minus 119881Δ [119905] (22)
where 119875VRS[119905] is the column vector of 119875VRS119894dBm[119905] forallV119894 isin V
and 119881 is the positive constant control parameter setting ofthe dynamicstochastic decision-making policy that affectsthe power-delay (ie queueing delay) tradeoffs
The proposed algorithm involves minimizing a bound onE[119875VRS[119905] | Θ[119905]] minus 119881Δ[119905] and finally this Lyapunov opti-mization theory gives Algorithm 1 to minimize the followingequation
sumV119894isinV
119875VRS119894dBm [119905] + 119881 sum
V119894isinV119876119894 [119905] (120582119894 [119905] minus 120583119894 [119905]) (23)
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 7
In (23) 120582119894[119905] is independent of power allocation and thusit is not controllable by power allocation Therefore it can beignored that is
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sum
V119894isinV119876119894 [119905] 120583119894 [119905] (24)
According to the fact that 120583119894[119905] can be derived by (14)finally (24) is reformulated as follows
sumV119894isinV
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot sum
V119894isinV119876119894 [119905]
sdot log2(119891dB (119891mW (119866VRSdBi + 119875VRS
119894dBm [119905] minus 119871 (119889))119899mW + sum119894isin119878119868 119868119894mW
)) (25)
where 119875VRS119894dBm[119905] is the transmit power allocation at VRS in a
dB scale at unit time 119905 in V119894 isin VIn (25) it is clear that the given equation (25) is separable
that is the separated minimization of the objective functionof each VRS forallV119894 isin V introduces the minimization of (25)Therefore each VRS forallV119894 isin V minimizes its own objectivefunction as follows
119875VRS119894dBm [119905] minus 119881 sdot 119861 sdot 119876119894 [119905]sdot log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (26)
From this given (26) we have to find the amount of powerallocation that minimizes (26) Therefore our final closed-form solution is as follows
119875lowastVRS119894dBm [119905] larr997888 arg min119875VRS119894dBm[119905]isinP
119875VRS119894dBm [119905] minus 119881119861119876119894 [119905] log2(119891dB (119891mW (119866VRS
dBi + 119875VRS119894dBm [119905] minus 119871 (119889))
119899mW + sum119894isin119878119868 119868119894mW)) (27)
The entire stochastic procedure is also described inAlgorithm 1
5 Performance Evaluation
This section consists of (i) simulation parameter settings(refer to Section 51) and (ii) performance evaluation interms of queue dynamics and energy-efficiency (refer toSection 52) respectively
51 Simulation Settings For evaluating the performance ofour proposed queue-stable time-average expected powerconsumption minimization algorithm the following simula-tion parameters are used
(i) Transmit antenna gain 15 dBi [17](ii) Transmit power allocation setP in (19) [17]
P ≜ 119901min = 10 dBm 1199011 = 13 dBm 1199012= 16 dBm 119901max = 19 dBm (28)
(iii) The number of interfering sources 3(iv) 119881 settings (three different values)
119881≜ 1198811 = 1 times 10minus23 1198812 = 5 times 10minus23 1198813 = 5 times 10minus24 (29)
For the simulation study in this paper we have threedifferent simulation results those are conducted with threedifferent 119881 settings that is 119881 = 1198811 = 1 times 10minus23 (for Figures3(a) and 3(b))119881 = 1198812 = 5times10minus23 (for Figures 3(c) and 3(d))and 119881 = 1198813 = 5 times 10minus24 (for Figures 3(e) and 3(f))
52 Simulation Results In this section we tracked the queuedynamics and energy consumption behaviors as plotted inFigure 3 As shown in Figure 3 the proposed strategiccontrol algorithm shows stable performance If the queue-backlog size reaches certain levels the proposed stochasticand strategic algorithm tries to stabilize the queue backlogsAs presented in Figures 3(a) 3(c) and 3(e) the queue-backlogsizes are stabilized near 15 times 1013 bits 05 times 1013 bits and 3 times1013 bits respectively Comparing to the performance of theproposed algorithmwith119881 = 1198811 = 1times10minus23 the performanceof the proposed algorithm with 119881 = 1198812 = 5 times 10minus23 takesmore queue stability because the average queue-backlog sizebecomes more smaller It means that the proposed algorithmwith 119881 = 1198812 = 5 times 10minus23 pursues more queue stabilizationIn addition the proposed algorithm with 119881 = 1198813 = 5 times 10minus24allows more queue-backlog sizes (ie allowing more queue-ing delays) comparing to the proposed algorithm with 119881 =1198811 = 1 times 10minus23
In Figures 3(b) 3(d) and 3(f) energy consumptionbehaviors with strategic stochastic control are simulatedand plotted with various 119881 values As presented in Figures3(b) 3(d) and 3(f) more queue stability takes more energyconsumption (especially refer to Figures 3(c) and 3(d)) Thisis reasonable due to the fact that more energy consumptionfor more transmit power allocation definitely leads to thestability of queues due to the fact that it will process morebits from the queue via increased SNR The average energyconsumption values and queue stability points with variousthree 119881 settings are presented in Table 2
As shown in Table 2 the tradeoff between energy-efficiency and queue stability is observed with various 119881settings In this case due to the fact that energy-efficiency andqueue stabilization have tradeoff relationship more queue
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
8 Mobile Information Systems
35
Proposed strategic controlUpper bound
Lower bound
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
Que
ue b
ackl
og (u
nit
bits)
times1013
(a) Queue dynamics with 119881 = 1198811 = 1 times 10minus23
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(b) Energy consumption with 119881 = 1198811 = 1 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
500 1000 15000Time (unit time slots)
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
(c) Queue dynamics with 119881 = 1198812 = 5 times 10minus23
300
350
400
450
500
550
600
650En
ergy
cons
umpt
ion
(uni
t m
illi-W
att)
500 1000 15000Time (unit time slots)
(d) Energy consumption with 119881 = 1198812 = 5 times 10minus23
Proposed strategic controlUpper bound
Lower bound
times1013
0
05
1
15
2
25
3
35
Que
ue b
ackl
og (u
nit
bits)
500 1000 15000Time (unit time slots)
(e) Queue dynamics with 119881 = 1198813 = 5 times 10minus24
500 1000 15000Time (unit time slots)
300
350
400
450
500
550
600
650
Ener
gy co
nsum
ptio
n (u
nit
mill
i-Wat
t)
(f) Energy consumption with 119881 = 1198813 = 5 times 10minus24
Figure 3 Queue dynamics and energy consumption behaviors for our proposed stochastic algorithm with various 119881 settings
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mobile Information Systems 9
Table 2 Average energy consumption and queue stability point in each time slot with various 119881 settings
119881 1198811 1198812 1198813Energy consumption (unit milli-Watt) 5889934 6238235 5503866Queue stability point (unit bits) asymp15 times1013 asymp05 times1013 asymp3 times1013
stabilization takes more energy Therefore the proposedalgorithmwith119881 = 1198812 = 5times10minus23 is less energy-efficientThisresult states that the proposed algorithm with 119881 = 1198812 = 5 times10minus23 ismore suitable for real-time (or time critical) VR videocontentsHowever this (setting119881 as big as possible) definitelyintroduces more energy consumption that is setting119881 as bigas possible is not always good In this case the system needsto consider additional energy provisioningmechanisms (egmore lithiumbattery allocation and self-chargingmechanisminvestigation in hardware platforms)
Notice that our considering stochastic network optimiza-tion and control theory formulated in (24) shows that higher119881 value setting provides more weights on queue stabilizationThis theoretical discussion is verified with this simulationresult
6 Concluding Remarks and Future Work
This paper proposes a strategic stochastic control algorithmthat is for the minimization of time-average expected powerconsumption subject to queue stability in distributed VRnetwork platforms In distributed VR network platforms theVRS contains VR contents that should be transmitted to thehead-mounted VRD For the wireless transmission 60GHzmmWave wireless communication technologies are usedowing to their high-speed massive data transmission (iemulti-Gbps data speed) On top of this 60GHz wirelesslink from VRS to its associated VRD a stochastic queue-stable control algorithm is proposed to minimize the time-average expected power consumption The VRS calculatesthe amount of transmit power allocation for transmitting bitsfrom VRS to its associated VRD If the amount of transmitpower allocation at VRS is quite high more bits will be pro-cessed from the queue This control eventually stabilizes thequeues whereas power (or energy) management is not opti-mal On the other hand the small number of bits will be pro-cessed from the VRS queue if the amount of transmit powerallocation is not enough Then the VRS queue should beunstabilized which should introduce the queue overflows inthe VRS Therefore an energy-efficient stochastic transmitpower control algorithm is necessary to transmit bits fromthe VRS queue under the design rationale of the jointoptimization of energy-efficiency and buffer stability Our sim-ulation results demonstrate that our proposed control algo-rithm achieves desired performance
As one of major future research directions realistic andautomatic119881 value setting strategies will be studied for the realimplementation of the proposed stochastic control algorithmin wireless VR platforms One of preliminary results in termsof adaptive 119881 setting is presented in [21]
Conflicts of Interest
The authors declare that there are no conflicts of interestregarding the publication of this paper
Acknowledgments
This work was supported by the National Research Founda-tion of Korea (NRF Korea) under Grant 2016R1C1B1015406and ETRI grant funded by the Korean government (16ZS1210NZV 120583-Grain Architecture for Ultra-Low Energy Processor)The related contents were presented by Joongheon Kim atElectronics and Telecommunications Research Institute(ETRI) Daejeon Korea on November 2016 with the titleldquoStochastic Control of 60GHz Links for Distributed VirtualReality Network Platformsrdquo
References
[1] F Ozkan ldquoIntroducing VR first unlocking the door to a futurein virtual realityrdquo IEEE Consumer Electronics Magazine vol 5no 4 pp 25ndash26 2016
[2] P Lelyveld ldquoVirtual reality primerwith an emphasis on camera-captured VRrdquo SMPTE Motion Imaging Journal vol 124 no 6pp 78ndash85 2015
[3] J P Schulze ldquoAdvanced monitoring techniques for data centersusing virtual realityrdquo SMPTE Motion Imaging Journal vol 119no 5 pp 43ndash46 2010
[4] S Kuriakose and U Lahiri ldquoUnderstanding the psycho-physiological implications of interaction with a virtual reality-based system in adolescents with autism a feasibility studyrdquoIEEE Transactions on Neural Systems and Rehabilitation Engi-neering vol 23 no 4 pp 665ndash675 2015
[5] J Kim Y Tian S Mangold and A F Molisch ldquoJoint scalablecoding and routing for 60 GHz real-time live HD video stream-ing applicationsrdquo IEEETransactions on Broadcasting vol 59 no3 pp 500ndash512 2013
[6] W Roh J-Y Seol J Park et al ldquoMillimeter-wave beamformingas an enabling technology for 5G cellular communications the-oretical feasibility and prototype resultsrdquo IEEECommunicationsMagazine vol 52 no 2 pp 106ndash113 2014
[7] T S Rappaport F Gutierrez E Ben-Dor J N Murdock YQiao and J I Tamir ldquoBroadband millimeter-wave propagationmeasurements and models using adaptive-beam antennas foroutdoor Urban cellular communicationsrdquo IEEE Transactions onAntennas and Propagation vol 61 no 4 pp 1850ndash1859 2013
[8] T S Rappaport J N Murdock and F Gutierrez ldquoState ofthe art in 60-GHz integrated circuits and systems for wirelesscommunicationsrdquo Proceedings of the IEEE vol 99 no 8 pp1390ndash1436 2011
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
10 Mobile Information Systems
[9] K Venugopal and R W Heath ldquoMillimeter wave networkedwearables in dense indoor environmentsrdquo IEEE Access vol 4pp 1205ndash1221 2016
[10] N Chen S Sun M Kadoch and B Rong ldquoSDN controlledmmWave massive MIMO hybrid precoding for 5G heteroge-neous mobile systemsrdquo Mobile Information Systems vol 2016Article ID 9767065 10 pages 2016
[11] H Kim M Ahn S Hong S Lee and S Lee ldquoWearable devicecontrol platform technology for network application devel-opmentrdquo Mobile Information Systems vol 2016 Article ID3038515 2016
[12] J Kim S Kwon and G Choi ldquoPerformance of video streamingin infrastructure-to-vehicle telematic platforms with 60-GHzradiation and IEEE 80211ad basebandrdquo IEEE Transactions onVehicular Technology vol 65 no 12 pp 10111ndash10115 2016
[13] Oculus GearVR httpswww3oculuscomen-usgear-vr[14] E Cuervo and D Chu ldquoPoster mobile virtual reality for
head-mounted displays with interactive streaming video andlikelihood-based foveationrdquo in Proceedings of the ACM Interna-tional Conference on Mobile Systems Applications and Services(MobiSys rsquo16) Singapore June 2016
[15] Facebook httpscodefacebookcomposts194023157622878gear-vr-to-get-dynamic-streaming-for-360-video
[16] P Debevec G Downing M Bolas H-Y Peng and J UrbachldquoSpherical light field environment capture for virtual realityusing a motorized pantilt head and offset camerardquo in Pro-ceedings of the International Conference on Computer Graphicsand Interactive Techniques (SIGGRAPH rsquo15) Los Angeles CalifUSA August 2015
[17] J Kim L Xian R Arefi and A S Sadri ldquo60GHz frequencysharing study between fixed service systems and small-cell sys-tems with modular antenna arraysrdquo in Proceedings of the IEEEGlobal Communications Conference (GLOBECOM) Workshopon Millimeter-Wave Backhaul and Access From Propagation toPrototyping (mmWave) San Diego Calif USA December 2015
[18] J Kim and A F Molisch ldquoFast millimeter-wave beam trainingwith receive beamformingrdquo Journal of Communications andNetworks vol 16 no 5 pp 512ndash522 2014
[19] A Maltsev E Perahia R Maslennikov A Lomayev A Kho-ryaev and A Sevastyanov ldquoPath Loss Model Development forTGad Channel Modelsrdquo IEEE 80211-090553r1 May 2009
[20] A F Molisch Wireless Communications Wiley-IEEE NewYork NY USA 2011
[21] J Kim ldquoEnergy-efficient dynamic packet downloading formed-ical IoT platformsrdquo IEEE Transactions on Industrial Informaticsvol 11 no 6 pp 1653ndash1659 2015
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Submit your manuscripts athttpswwwhindawicom
Computer Games Technology
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Distributed Sensor Networks
International Journal of
Advances in
FuzzySystems
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014
International Journal of
ReconfigurableComputing
Hindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
Applied Computational Intelligence and Soft Computing
thinspAdvancesthinspinthinsp
Artificial Intelligence
HindawithinspPublishingthinspCorporationhttpwwwhindawicom Volumethinsp2014
Advances inSoftware EngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Electrical and Computer Engineering
Journal of
Hindawi Publishing Corporation
httpwwwhindawicom Volume 2014
Advances in
Multimedia
International Journal of
Biomedical Imaging
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Advances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 201
RoboticsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Computational Intelligence and Neuroscience
Industrial EngineeringJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Human-ComputerInteraction
Advances in
Computer EngineeringAdvances in
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014