Convergence of Heterogeneous Wireless Networks for 5G-and-Beyond Communications...

Wireless Communications and Mobile Computing

Convergence of Heterogeneous Wireless Networks for 5G-and-Beyond Communications: Applications, Architecture, and Resource Management

Lead Guest Editor: Mostafa Z. ChowdhuryGuest Editors: Md Jahidur Rahman, Gabriel-Miro Muntean, Phuc V. Trinh, and Juan C. Cano

Convergence of HeterogeneousWireless Networks for 5G-and-BeyondCommunications: Applications, Architecture,and Resource Management

Wireless Communications and Mobile Computing

Convergence of HeterogeneousWireless Networks for 5G-and-BeyondCommunications: Applications, Architecture,and Resource Management

Lead Guest Editor: Mostafa Z. ChowdhuryGuest Editors: Md Jahidur Rahman, Gabriel-Miro Muntean,Phuc V. Trinh, and Juan C. Cano

Copyright © 2019 Hindawi. All rights reserved.

This is a special issue published in “Wireless Communications and Mobile Computing.” All articles are open access articles distributedunder the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, pro-vided the original work is properly cited.

Editorial Board

Javier Aguiar, SpainGhufran Ahmed, PakistanWessam Ajib, CanadaMuhammad Alam, ChinaEva Antonino-Daviu, SpainShlomi Arnon, IsraelLeyre Azpilicueta, MexicoPaolo Barsocchi, ItalyAlessandro Bazzi, ItalyZdenek Becvar, Czech RepublicFrancesco Benedetto, ItalyOlivier Berder, FranceAna M. Bernardos, SpainMauro Biagi, ItalyDario Bruneo, ItalyJun Cai, CanadaZhipeng Cai, USAClaudia Campolo, ItalyGerardo Canfora, ItalyRolando Carrasco, UKVicente Casares-Giner, SpainLuis Castedo, SpainIoannis Chatzigiannakis, ItalyLin Chen, FranceYu Chen, USAHui Cheng, UKErnestina Cianca, ItalyRiccardo Colella, ItalyMario Collotta, ItalyMassimo Condoluci, SwedenDaniel G. Costa, BrazilBernard Cousin, FranceTelmo Reis Cunha, PortugalIgor Curcio, FinlandLaurie Cuthbert, MacauDonatella Darsena, ItalyPham Tien Dat, JapanAndré de Almeida, BrazilAntonio De Domenico, FranceAntonio de la Oliva, SpainGianluca De Marco, ItalyLuca De Nardis, ItalyLiang Dong, USAMohammed El-Hajjar, UK

Oscar Esparza, SpainMaria Fazio, ItalyMauro Femminella, ItalyManuel Fernandez-Veiga, SpainGianluigi Ferrari, ItalyIlario Filippini, ItalyJesus Fontecha, SpainLuca Foschini, ItalyA. G. Fragkiadakis, GreeceSabrina Gaito, ItalyÓscar García, SpainManuel García Sánchez, SpainL. J. García Villalba, SpainJosé A. García-Naya, SpainMiguel Garcia-Pineda, SpainA.-J. García-Sánchez, SpainPiedad Garrido, SpainVincent Gauthier, FranceCarlo Giannelli, ItalyCarles Gomez, SpainJuan A. Gómez-Pulido, SpainKe Guan, ChinaAntonio Guerrieri, ItalyDaojing He, ChinaPaul Honeine, FranceSergio Ilarri, SpainAntonio Jara, SwitzerlandXiaohong Jiang, JapanMinho Jo, Republic of KoreaShigeru Kashihara, JapanDimitrios Katsaros, GreeceMinseok Kim, JapanMario Kolberg, UKNikos Komninos, UKJuan A. L. Riquelme, SpainPavlos I. Lazaridis, UKTuan Anh Le, UKXianfu Lei, ChinaHoa Le-Minh, UKJaime Lloret, SpainMiguel López-Benítez, UKMartín López-Nores, SpainJavier D. S. Lorente, SpainTony T. Luo, Singapore

Maode Ma, SingaporeImadeldin Mahgoub, USAPietro Manzoni, SpainÁlvaro Marco, SpainGustavo Marfia, ItalyFrancisco J. Martinez, SpainDavide Mattera, ItalyMichael McGuire, CanadaNathalie Mitton, FranceKlaus Moessner, UKAntonella Molinaro, ItalySimone Morosi, ItalyKumudu S. Munasinghe, AustraliaEnrico Natalizio, FranceKeivan Navaie, UKThomas Newe, IrelandWing Kwan Ng, AustraliaTuan M. Nguyen, VietnamPetros Nicopolitidis, GreeceGiovanni Pau, ItalyRafael Pérez-Jiménez, SpainMatteo Petracca, ItalyNada Y. Philip, UKMarco Picone, ItalyDaniele Pinchera, ItalyGiuseppe Piro, ItalyVicent Pla, SpainJavier Prieto, SpainRüdiger C. Pryss, GermanySujan Rajbhandari, UKRajib Rana, AustraliaLuca Reggiani, ItalyDaniel G. Reina, SpainJose Santa, SpainStefano Savazzi, ItalyHans Schotten, GermanyPatrick Seeling, USAMuhammad Z. Shakir, UKMohammad Shojafar, ItalyGiovanni Stea, ItalyEnrique Stevens-Navarro, MexicoZhou Su, JapanLuis Suarez, RussiaVille Syrjälä, Finland

Hwee Pink Tan, SingaporePierre-Martin Tardif, CanadaMauro Tortonesi, ItalyFederico Tramarin, ItalyReza Monir Vaghefi, USA

Juan F. Valenzuela-Valdés, SpainAline C. Viana, FranceEnrico M. Vitucci, ItalyHonggang Wang, USAJie Yang, USA

Sherali Zeadally, USAJie Zhang, UKMeiling Zhu, UK

Convergence of Heterogeneous Wireless Networks for 5G-and-Beyond Communications: Applications,Architecture, and Resource ManagementMostafa Zaman Chowdhury , Md Jahidur Rahman , Gabriel-Miro Muntean , Phuc V. Trinh ,and Juan Carlos CanoEditorial (2 pages), Article ID 2578784, Volume 2019 (2019)

Efficient Content Delivery for Mobile Communications in Converged NetworksMahfuzur Rahman Bosunia and Seong-Ho JeongResearch Article (12 pages), Article ID 3170694, Volume 2019 (2019)

Fuzzy Based Network Assignment and Link-Switching Analysis in Hybrid OCC/LiFi SystemMoh. Khalid Hasan , Mostafa Zaman Chowdhury , Md. Shahjalal , and Yeong Min JangResearch Article (15 pages), Article ID 2870518, Volume 2018 (2019)

Exploiting Opportunistic Scheduling for Physical-Layer Security in Multitwo User NOMANetworksKyusung Shim and Beongku AnResearch Article (12 pages), Article ID 2797824, Volume 2018 (2019)

An Implementation Approach and Performance Analysis of Image Sensor BasedMultilateral IndoorLocalization and Navigation SystemMd. Shahjalal , Md. Tanvir Hossan , Moh. Khalid Hasan , Mostafa Zaman Chowdhury ,Nam Tuan Le , and Yeong Min JangResearch Article (13 pages), Article ID 7680780, Volume 2018 (2019)

Threshold Secret Sharing Transmission against Passive Eavesdropping in MIMOWireless NetworksJungho Myung, Keunyoung Kim, and Taehong KimResearch Article (7 pages), Article ID 4143061, Volume 2018 (2019)

Network-Assisted Optimal Datalink Selection Scheme for Heterogeneous Aeronautical NetworkDongli Wang , Guoce Huang, Shufu Dong, Yequn Wang, Jian Liu, and Weiting GaoResearch Article (13 pages), Article ID 9349824, Volume 2018 (2019)

http://orcid.org/0000-0003-1487-086Xhttp://orcid.org/0000-0002-9559-2023http://orcid.org/0000-0002-9332-4770http://orcid.org/0000-0001-8087-6739http://orcid.org/0000-0002-0038-0539http://orcid.org/0000-0003-0838-4400http://orcid.org/0000-0002-6095-2535http://orcid.org/0000-0002-7773-3523http://orcid.org/0000-0003-1487-086Xhttp://orcid.org/0000-0002-4876-6860http://orcid.org/0000-0002-9963-303Xhttp://orcid.org/0000-0003-4851-0811http://orcid.org/0000-0002-0587-3754http://orcid.org/0000-0002-4876-6860http://orcid.org/0000-0001-7171-6602http://orcid.org/0000-0002-7773-3523http://orcid.org/0000-0003-1487-086Xhttp://orcid.org/0000-0003-3117-8613http://orcid.org/0000-0002-9963-303Xhttp://orcid.org/0000-0001-6246-6218http://orcid.org/0000-0003-2064-9272

EditorialConvergence of Heterogeneous Wireless Networks for5G-and-Beyond Communications: Applications, Architecture,and Resource Management

Mostafa Zaman Chowdhury ,1,2 Md Jahidur Rahman ,3 Gabriel-Miro Muntean ,4

Phuc V. Trinh ,5 and Juan Carlos Cano 6

1Kookmin University, Seoul, Republic of Korea2Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh3Qualcomm Technologies Inc., San Diego, CA, USA4Dublin City University (DCU), Dublin, Ireland5National Institute of Information and Communications Technology (NICT), Tokyo, Japan6Technical University of Valencia, Camı́ de Vera, Spain

Correspondence should be addressed to Mostafa Zaman Chowdhury; [email protected]

Received 25 December 2018; Accepted 25 December 2018; Published 14 January 2019

Copyright © 2019 Mostafa Zaman Chowdhury et al. This is an open access article distributed under the Creative CommonsAttribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work isproperly cited.

Evolving fifth-generation- (5G-) and-beyond communica-tion networks are envisioned to provide services withmassiveconnectivity, ultrahigh data-rate, ultralow latency, muchimproved security, very low energy consumption, and highquality of experience. 5G-and-beyond communication sys-tems not only will be more advanced but also are expectedto be more complex in comparison with legacy systems. Toachieve the goals of 5G-and-beyond communication systems,convergence of the heterogeneous wireless technologies hasemerged as one of the key solutions.This entails convergenceof not only the radio frequency (RF) technologies, but alsothe optical and RF/optical wireless technologies. The opticalspec trum is considered as an emerging solution for thedevelopment of future high capacity optical wireless commu-nication (OWC) networks. It offers unique advantages, suchas huge unregulated optical spectrum and inherent security.Therefore, future networks are anticipated to adopt amultitierRF/optical architecture comprising macrocells, microcells,different types of licensed small cells, optical attocells, OWCnetworks, and relays. The future 5G-and-beyond systems,instead of being a single wireless access network, will bea “network of networks.” The seamless integration among

heterogeneous wireless, optical and RF/optical wireless net-works, demands paradigm shifts in such a way that differ-ent networks collaborate with each other so as to achievethe desired goals of the 5G-and-beyond communications.In order to attain full convergence of the heterogeneousnetworks, many technical issues need to be resolved.

The motivation behind this special issue has been tosolicit cutting-edge research relevant to applications, archi-tecture, and resource management of heterogeneous wirelessnetworks for 5G-and-beyond communications. This specialissue invited papers that address such issues. Following arigorous review process (including a second review round),six outstanding papers have been finally selected for inclusionin the special issue. The accepted papers cover a wide rangeof research subjects in the broader area of convergence ofheterogeneous wireless networks to meet the demand of 5G-and-beyond communications systems.

The paper “Network-Assisted Optimal Datalink Selec-tion Scheme for Heterogeneous Aeronautical Network” by D.Wang et al. focuses on datalink selection mechanism inheterogeneous aeronautical network. The authors proposeda priority distinction selection algorithm by constructing

HindawiWireless Communications and Mobile ComputingVolume 2019, Article ID 2578784, 2 pageshttps://doi.org/10.1155/2019/2578784

http://orcid.org/0000-0003-1487-086Xhttp://orcid.org/0000-0002-9559-2023http://orcid.org/0000-0002-9332-4770http://orcid.org/0000-0001-8087-6739http://orcid.org/0000-0002-0038-0539https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://doi.org/10.1155/2019/2578784

2 Wireless Communications and Mobile Computing

multiuser multiobjective optimization problem to maximizethe number of users accessing their optimal datalinks andminimize the modification of the users’ access requests.

The paper by J. Myung et al. entitled “Threshold SecretSharing Transmission against Passive Eavesdropping inMIMOWireless Networks” proposes a threshold secret sharingscheme for secure communications in multiple input andmultiple output wireless networks. In their novel scheme,the base station divides the secret data into a number ofparts using a polynomial-based approach and transmits thedivided data to the legitimate user by beamforming withmultiple spatial dimensions.

The paper “An Implementation Approach and Perfor-mance Analysis of Image Sensor Based Multilateral IndoorLocalization and Navigation System” by M. Shahjalal et al.investigates the implementation issues for indoor mobilerobot localization and navigation systems. The authors pro-posed an indoor navigation and positioning combined algo-rithm and further evaluate its performance for the feasibilityof real-implementation.They developed an Android applica-tion to support data acquisition from multiple simultaneoustransmitter links.

The paper by K. Shim et al. entitled “Exploiting Oppor-tunistic Scheduling for Physical-Layer Security in MultitwoUser NOMA Networks” addresses the opportunistic schedul-ing inmultitwo user nonorthogonalmultiple access (NOMA)systems consisting of one base station, multiple near users,multiple far users, and one eavesdropper. The authors intro-duced a user selection scheme, called best-secure-near-userbest-secure-far-user scheme to improve the secrecy perfor-mance. Additionally, the authors proposed a descent-basedsearch method to find the optimal values of the powerallocation coefficients that can minimize the total secrecyoutage probability.

The paper “Fuzzy Based Network Assignment and Link-Switching Analysis in Hybrid OCC/LiFi System” by M. Khalidet al. proposes a hybrid optical camera communicationsand light fidelity architecture to improve the quality-of-service (QoS) of users.The authors present a network assign-ment mechanism for such hybrid systems. A dynamic link-switching technique is proposed which includes switchingprovisioning based on user mobility and detailed networkswitching flow analysis. Fuzzy logic is used to develop theirproposed mechanism. A time-division multiple access is alsoadopted to ensure fairness in time resource allocation whileserving multiple users using the same light-emitting diode inthe hybrid system.

The paper by M. R. Bosunia and S.-Ho Jeong entitled“Efficient Content Delivery for Mobile Communications inConverged Networks” proposes a content-centric networkingbased content delivery mechanism for 4G and 5G heteroge-neous converge networks. The authors described a mobilitymanagement scheme to support the content diversity andnetwork diversity by leveraging the abundant computationresources in the mobile network. In addition, this paperanalyzes the existing approaches with respect to mobility andevaluates the performance of their seamless content deliverymechanisms in terms of content transfer time, throughput,and data transmission success ratio.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

The editors would like to thank all the authors who submittedtheir excellent research articles to this special issue. Also,we would like to express our deepest gratitude to all thereviewers for providing their valuable and timely feedbackduring the review process, which helped to improve thequality of this special issue. The work of Mostafa ZamanChowdhury was supported by the Korea Research Fellow-ship Program through the National Research Foundation ofKorea (NRF) funded by the Ministry of Science and ICT(2016H1D3A1938180).

Mostafa Zaman ChowdhuryMd Jahidur Rahman

Gabriel-Miro MunteanPhuc V. Trinh

Juan Carlos Cano

Research ArticleEfficient Content Delivery for Mobile Communications inConverged Networks

Mahfuzur Rahman Bosunia 1 and Seong-Ho Jeong 2

1Bangladesh Bank, Dhaka, Bangladesh2Dept. of Information and Communications Engineering, Hankuk University of Foreign Studies, Seoul, Republic of Korea

Correspondence should be addressed to Seong-Ho Jeong; [email protected]

Received 3 August 2018; Accepted 4 November 2018; Published 14 January 2019

Guest Editor: Mostafa Zaman Chowdhury

Copyright © 2019 Mahfuzur Rahman Bosunia and Seong-Ho Jeong. This is an open access article distributed under the CreativeCommons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided theoriginal work is properly cited.

The mobile Internet is already playing a key role in people’s daily lives worldwide, resulting in the dramatic growth in the numberof mobile devices.The size of the Internet and the amount of the traffic are being expanded rapidly, which poses various challenges.In particular, the Internet and mobile communications are entering a new era that demands faster communication services anduninterrupted content delivery. A new paradigm called content-centric networking (CCN) is considered as an appropriate way forefficient content delivery. In this paper, we propose a CCN-based efficient content delivery mechanism in the 4G network and alsoin the upcoming 5G network where various heterogeneous networks are converged. We also propose a novel mobility managementscheme to support the content diversity and network diversity by leveraging the abundant computational resources in the mobilenetwork. In addition, we analyze the existing approaches with respect to mobility and evaluate the performance of our seamlesscontent delivery mechanism in terms of content transfer time, throughput, and data transmission success ratio. Simulation resultsare also presented to show that the content-centric wireless network with our mobility management scheme can improve the datadelivery services significantly compared to the existing schemes.

1. Introduction

Recently, as the use of mobile devices such as smart phonesand tablets has increased exponentially, the number of ser-vices that connect mobile devices to the Internet has alsoincreased. Accordingly, the volume of the mobile content hasincreased enormously, and it is a big challenge to supportfast content delivery as well as seamless mobility in thewireless network. The content-centric networking (CCN) [1,2] architecture can be useful to resolve this issue.

The current Internet is configured for host-to-host com-munication, and therefore it is not suitable for the futureInternet which will deal with various contents and provideseamless mobility for moving users. The main focus of CCNis to provide more efficient, faster, and secured delivery ofa content rather than to establish the communication pathto the content source. CCN provides name-based routingwithout exploiting the content/device address. CCN operatesusing two simple messages for content transmission: first one

is the Interest packet and the other is the Data packet. TheInterest packet contains the request for a desired content, andthe request consists of different attributes such as contentname, content type, and content version. The Data packetcontains the original data with the content name, securityrelated information, and several other attributes, e.g., hopdistance and content source description.

Figure 1 shows the CCN forwarding module whichis equipped with three functional and operative elementsfor content-based routing: Content Store (CS), ForwardingInformation Base (FIB), and Pending Interest Table (PIT).The CS is the physical storage of the content and stores theidentical name of the published content. The FIB preservesthe content routing information for mapping between thecontent name and the next hop towards the content source.The PIT keeps track of and records the status of all receivedInterest packets in order to satisfy later when the content ison hand. Any node that wants to publish a content spreadsthe content name to the nearby nodes to make it available


http://orcid.org/0000-0003-0838-4400http://orcid.org/0000-0002-6095-2535https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://doi.org/10.1155/2019/3170694


Name Data

/hufs.com/cclab/ccn/paper.ts/v1/s1 ………

Prefix Requesting Face(s)

/hufs.com/cclab/ccn/paper.ts/v1/s1 0

Prefix Face list

/hufs.com 0,1

Content Store

Pending Interest Table (PIT)

Forward Information Base (FIB)

Ptr Type

C

P

F

Index

Face 0

Face 1

Face 2

C = Content StoreP = PITF = FIB

Application

Figure 1: A forwarding module for CCN.

and accessible to the content consumers. A node sends anInterest packet to retrieve a content based on the routinginformation available in the FIB. When a node receives theInterest packet, it enforces a lookup inside the CS to checkthe content availability. If the requested content is availablethen the content is delivered back to the content requester. Ifthe content is not available inside the CS, a pending Interestis recorded in the PIT and the Interest packet is forwarded viathe face guided by the FIB.

The LTE-based 4G mobile network establishes a networkarchitecture to provide faster data retrieval and seamlessmobility management. Even though LTE promises to be afaster and more efficient data delivery network, its communi-cation architecture is still centralized and host oriented; thusvarious performance degradation issues, e.g., high bandwidthconsumption, cross domain traffic, high delay for the longcommunication path, and waste of network resources, mayneed to be resolved. A new paradigm called CCN is con-sidered as an appropriate way of efficient content delivery asmentioned above. The integration of CCN and LTE can beused to resolve the issues by providing timely and fast deliveryof the data with efficient resource utilization.

In this paper, we propose a CCN-based efficient contentdelivery mechanism in the 4G network and also in theupcoming 5G network where various heterogeneous net-works are converged. We also propose an efficient mobilitymanagement mechanism to address the content diversity andnetwork diversity by leveraging the abundant computationalresources in the LTE-based 4G network. Our proposedmobility management scheme introduces a context-awarehandover prediction mechanism to deal with the heterogene-ity of wireless content providers and consumers. Contextawareness provides a significant effect to guarantee datacontinuity in the mobile environment. We also introducea content similarity matching mechanism to provide thedata continuity when mobility increases the chance of data

unavailability. The rest of the paper is organized as follows.Section 2 describes the recentwork related to content deliveryinCCN. Section 3 presents a novel approach to integrateCCNwith LTE and some mechanisms for efficient and seamlesscontent delivery in the LTE network in detail. Section 4describes implementation details and performance issues,and finally, Section 5 concludes the paper.

2. Related Work

Seamless content delivery andmobility management in wire-less networks are important research issues which have beenfocused a lot under various scenarios. A lot of works havebeen proposed so far in this area. The LTE network is limitedto the measured value of the received signal strength whichis a factor that triggers handover in the wireless network. Inreality, handover involves content delivery rate requirementsfor a particular application, packet loss, end-device demandfor high/low data rate, and many more.

There are various mobility-related mechanisms and pro-tocols across all the layers in the TCP/IP protocol stackfor providing seamless content transmission. For example,Mobile IP [3] provides mobility support at the Networklayer; the Stream Control Transmission Protocol (SCTP) [3]and the Datagram Congestion Control Protocol (DCCP) [3]provide seamless mobility support at the transport layer.Dynamic DNS (DDNS) [3] and Session Initiation Protocol(SIP) [3] are examples to provide mobility management atthe application layer. The efficiency of these mechanisms islimited due to the acting of IP address as a locator as well asan identifier and also due to the cross layer communicationbetween different layers. The separation of the locator fromits identifier is proposed in Host Identity Protocol (HIP) [3]and Locator Identifier Separation Protocol (LISP) [4], butthey cannot come out from the host-to-host communicationscenario.

Wireless Communications and Mobile Computing 3

The future Internet architecture based onInformation-Centric Networking (ICN) such as CCN, Name DataNetworking (NDN), Data-Oriented Network Architecture(DONA), and Network of Information (NetInf) uses in-network content caching to improve the efficiency of contenttransmission, reduce the network traffic and content accesslatencies, alleviate the present communication bottlenecks,and support ubiquitous access and efficient mobility manage-ment. DONA [5] introduces integrated name resolution andcontent-based routing schemes by replacing the concept ofDNS in the traditional TCP/IP based Internet architecture.DONA uses a flat, self-certifying name of a content andregisters the content name and the location in a domainserver called Resolution Handlers (RHs). RHs are structuredinto a BGP topology of the network and content lookupsare performed by querying a consumer to its local RH. Ifno reference of the requested content is found, the query isforwarded up the tree until a content source is found; an out-of-band delivery path is then established by the source (overIP). DONA reduces the applicability due to the dependencyon RHs like DNS and location-based communication. Theamount of delay and overhead may also be a big concern forDONA.

Network of Information (NetInf) [6] follows the similarmechanisms as DONA and also provides content deliveryusing a name resolution (NR) service. To handle content anddevice mobility issues and provide content-based routing,NetInf uses Multiple Distributed Hash Table (MDHT) [7]and Late Locator Construction (LLC) [8] schemes. MDHTand LLC try to make the content management and nameresolution simple and provide in-network content caching.PERSUIT [9] uses three key components, Rendezvous,Topology, and Routing, to provide seamless content deliveryand mobility management in a publish/subscribe architec-ture. Bloom filter based source routing is used to transfercontent through the network [10]. However, NetInf reducesits efficiency due to the dependency on the NR which isresponsible for content registration, content updates, andcontent-based route establishment. It requires a re-bindingsimilar to DONA. The mechanism for handling mobility inNetInfmay vary according to the chosen content locator; thusthe implementation can be complex.The cost of updating therouting information for PERSUIT is very high. The packetloss may be significant in the case of high mobility.

The CCN architecture [11, 12] decouples the content fromthe location and device and distributes the content usingcontent name-based routing. In CCN, consumer mobility ismanaged inherently by its receiver driven nature. There isno need to update the routing information due to mobilityfrom a consumer’s perspective; the consumer just retransmitsInterest packets if the content is not available yet. Even thoughCCN supports consumer mobility inherently, it faces longdelays to re-issue the Interest after rebinding to a newnetworkand cannot provide seamless content transmission. Contentprovider mobility is still an open issue in CCN; there areseveral issues to be resolved such as update of routing andlocation information, repeated transmission of Interest/Datapackets, and undesirable content delivery delays due tomobility.

A proxy-based approach [13] proposes a publisher mobil-ity support protocol in CCN and a fast FIB update mecha-nism. It introduces the mobility entry in FIB for mobile andtemporal destination and also defines the Home Router (HR)or proxy to announce the original entry of the publisher to thenetwork and establish a tunnel between the previous pointof attachment and a new point of attachment to reduce thepackets loss due to mobility. Clustered CCN [14] introducesthe cluster concept to support mobility which can be viewedas a hierarchical mobility management scheme to support theextensive mobile domain. It forms a cluster with a clusterhead that manages all the responsibilities of its memberslike Interest processing and mobility tracking. With theassistance of the cluster or proxy, it can reduce the Interestdissemination and content distribution, but the overhead andcomplexity of this approach are high due to its centralized andhierarchical nature.

A converged network is useful to exploit content diversityand device heterogeneity by disseminating contents throughseveral networks, e.g., Wi-Fi, broadcast networks or cellularnetworks [15, 16]. Despite a large variability of contentrequests to several routes and several content sources forefficient transmission of user requests and content storing,content requests are typically satisfied by the nearby devicesor networks [17–19]. The separation among content delivery,content storage, and content and device mobility operationsmay reduce the performance efficiency of the network andincreases the operational complexity of the network.

Fetching the content before handover was proposed in[20] to support producer mobility in name-based routing.Software defined controller for CCN [21] proposed amobilitymanagement mechanism for allowing packet forwardingand intermediate routing on the device mobility. SoftwareDefined Mobile Network [22] was proposed to improvethe content delivery efficiency by optimizing caching in theLTE network in which Software Defined Networking (SDN)mechanisms were integrated with the Mobility ManagementEntity (MME). It allows dynamic relocation of contents inany intermediate node. This mechanism was compared andevaluated in [23] by using simulation. The simulation resultsshowed that the in-node content caching reduced trafficload and improved content delivery efficiency. Even thoughintermediate content storing ensures faster content access,the virtual tunnel-based content redirection in the LTEnetwork increases the overhead and reduces the transmissionefficiency.

A mobility direction prediction mechanism was pro-posed in [24] for reducing the number of handovers anddata losses in the LTE network. However the scope of thiswork is very limited due to theTCP/IP-based communicationnature.The future mobile networks demand the mobility andportability of devices and data or contents in an autonomousand adaptive way to provide seamless content delivery, tobe connected to several access networks simultaneously, andmaintain the high quality of content transmission withoutany interruption even in the highly mobile environment. Theproposed mechanism is an enhancement to our previouswork [25], which makes it possible to directly fetch and storecontents in any appropriate node before handover to enable


UE2 eNodeB SGW/PGWUE1

4. Interest

12. Interest

1. Establish connection 2. UE identifier

3. Interest

Content Source

5. Interest

6. Data7. Data

8. Data

9. Terminate connection

10. Establish connection

13. Data14. Terminate connection

11. UE identifier

Figure 2: Content delivery procedure in the content-centric mobile network (e.g., LTE network).

faster content retrieval for reducing content transmissiondelays and preventing the repeated transmission of Inter-est/Data packetsto avoid the network congestion.

3. A CCN-based Mechanism forSeamless and Efficient Content Delivery inthe Mobile Network

Mobility in the Internet means that either the consumer orthe provider is moving away from its point of attachment toanother or both of them are moving away together. Mobilitysupport is a service such that the mobility of the node shouldnot result in any loss of data or extended periods of discon-nection. It is still an issue how CCN will be integrated withthe LTE network. We simplify here a seamless data deliveryprocedure in the content-centric LTE network. The eNodeB,SGW/PGW, and MME can support the CCN function andprotocols. The detailed content delivery procedure in thecontent-centric LTE network is described below and alsoshown in Figure 2.

(i) Step 1: UE1 establishes a connection to an eNodeB.(ii) Step 2: the eNodeB sends the UE1 information con-

taining the node identifier and the interface identifierto the SGW/PGW.

(iii) Step 3: UE1 sends the Interest packet via the eNodeB.When the eNodeB receives the Interest packet, itperforms according to the basic operation of the CCNnode.

(iv) Step 4: the SGW/PGW receives the Interest packetfrom the eNodeB. After searching its CS and PIT, ifno matched content is found, it checks its FIB.

(v) Step 5: the SGW/PGW’s FIB sends the Interest packetvia the face of the content provider.

(vi) Step 6: when a content provider receives the Interestpacket, it searches its CS. When the matched contentis found, it sends the Data packet of the content outvia the incoming face.

(vii) Steps 7, 8: when the SGW/PGW receives the Datapacket, it forwards it to the eNodeB and UE1 basedon their PIT.

(viii) Step 9: UE1 terminates the connection.(ix) Step 10: UE2 establishes a connection to the eNodeB.(x) Step 11: the eNodeB sends the UE2 information con-

taining the node identifier and the interface identifierto the SGW/PGW.

(xi) Step 12: UE2 sends the Interest packet to the eNodeB.(xii) Step 13: when the eNodeB receives the Interest packet,

it performs according to the basic operation of theCCN node. Since it is cached in the previous eNodeB,it sends the Data packet to UE2.

(xiii) Step 14: UE2 terminates the existing connection.

3.1. Device Mobility Prediction. Device mobility or consumermobility allows consumers to change their point of attach-ment without disrupting the connectivity. We have defined amathematical model which takes into account the preferenceof end devices, e.g., UEs, when selecting a base stationfor seamless and fast content retrieval and content deliveryservices. Let 𝑥 be a value for a single criterion and 𝛼 be thesteepness. 𝑥𝑚𝑖𝑛 ≤ 𝑥𝑚 ≤ 𝑥𝑚𝑎𝑥 where 𝑥𝑚 is the midpoint of the


variation range. These variations can be defined as [26] usingthe single criteria utility function as shown in

𝑢 (𝑥) =

{{{{{{{{{{{{{{{{{

0 𝑖𝑓 𝑥 ≤ 𝑥𝑚𝑖𝑛1

1 + 𝑒𝛼(𝑥𝑚−𝑥)/(𝑥−𝑥𝑚𝑖𝑛) 𝑖𝑓 𝑥𝑚𝑖𝑛 < 𝑥 ≤ 𝑥𝑚1 − 11 + 𝑒𝛽(𝑥−𝑥𝑚)/(𝑥𝑚𝑎𝑥−𝑥) 𝑖𝑓 𝑥𝑚 < 𝑥 ≤ 𝑥𝑚𝑎𝑥1 𝑖𝑓 𝑥 ≥ 𝑥𝑚𝑎𝑥

(1)

where

𝛽 = 𝛼 (𝑥𝑚𝑎𝑥 − 𝑥𝑚)𝑥𝑚 − 𝑥𝑚𝑖𝑛(2)

and 𝛼 > 0 is the tuned steepness parameter. The proposedutility function satisfies the following properties: 𝑢(𝑥) =0 ∀𝑥 ≤ 𝑥𝑚𝑖𝑛, 𝑢(𝑥) = 1 ∀𝑥 ≥ 𝑥𝑚𝑎𝑥, and 𝑢(𝑥𝑚) = 0.5.The point of attachment selection in the wireless networkingenvironment is based on the aggregation of different utilityfunctions for decision processes. Hence, we define here amulticriteria utility function that is able to integrate the enddevices’ different choice metrics to select a best point ofattachment. Let 𝑅 = 𝑅1 . . . 𝑅𝑛 be a set of potential alternatives(e.g., possible different eNodeBs) and each alternative can bedescribed as a different descriptor or attributes (e.g., receivedsignal strength, mobility direction, and load in terms ofdata transfer rate) 𝑥 = 𝑥1 ∗ . . . ∗ 𝑥𝑛, and each alternativeattribute being described as a utility function 𝑢(𝑥𝑛), thesimple weighted average of different alternatives A(R1 . . .Rn)is used to maximize the selection probability of the besteNodeB as follows:

𝐴𝑅 =𝑙

∑𝑖=1

𝑤𝑖𝑢 (𝑥𝑖) (3)

where 𝑤𝑖 is a weight that reflects the content receiver’s pref-erence. Weights are assigned according to the UE’s expectedcriteria.

(i) Received Signal Strength. Received signal strength (RSS)is one of the most popular parameters to take a handoverdecision. By monitoring the RSS, it is easily determinedwhether the UE should connect to a new eNodeB or not.The UE reports the received RSS value for all the neighboreNodeBs to the serving eNodeB. The eNodeB that takes ahandover decision uses the RSS values of each eNodeB in asingle criteria utility function as shown in (1).

(ii)Mobility Direction.The proposedmechanism also uses themoving direction prediction of each UE tomake the decisionof the movement towards an eNodeB. We assume that eacheNodeB is aware of the position of the 2-hop neighboreNodeBs and each UE is aware of its own position. Assumingthe serving eNodeB position is (𝑋𝑒, 𝑌𝑒), the position of theUE is (𝑋𝑢, 𝑌𝑢), and the candidate eNodeB position is (𝑋𝑛 , 𝑌𝑛),using the Pythagorean Theorem, it is possible to estimatedistance between a UE and a candidate eNodeB as shown in

d = √(𝑋𝑛 − 𝑋𝑢)2 + (𝑌𝑛 − 𝑌𝑢)2 = √(Δ𝑋)2 + (Δ𝑌)2 (4)

Therefore, the probability of a UE being in a coverage areaof an eNodeB is

𝑃𝑟 {𝑑 ≤ 𝑅𝑟} = 𝑃𝑟 {√(Δ𝑋)2 + (Δ𝑌)2 ≤ 𝑅𝑟} (5)

So 𝑑 is normalized using the coverage 𝑅𝑟 as follows:

𝑑𝑅 = 𝑅𝑟 − 𝑑𝑅𝑟 (6)

Since the velocity is a vector and a UE moves to differentdirection, it is reasonable to predict the direction or angle ofthe UE towards a candidate SBS using the vector formula.Themoving angle of the UE from the associated eNodeB to a newcandidate eNodeB can be estimated as shown in

𝜃 = cos−1 𝑋𝑢𝑋𝑛 + 𝑌𝑢𝑌𝑛√𝑋2𝑢 + 𝑌2𝑢 ∗ √𝑋2𝑛 + 𝑌2𝑛

(7)

It is considered that the 120∘ angle is the acceptable angletowards a eNodeB as in [24]. So it is considered as an offset of𝜇= ±60∘ to normalize the 𝜃 value as shown in the followingformula

𝜃𝑅 = 𝜇 − 𝜃𝑅𝜇 (8)

Then 𝜃 and 𝑑 are used to estimate the movement predic-tion 𝑃𝑚 as shown in the following formula

𝑃𝑚 = (1 − 𝛼) ∗ 𝑑𝑅 + 𝛼 ∗ 𝜃𝑅 (9)The eNodeB that takes a handover decision uses the

movement prediction values of (9) of each eNodeB in a singlecriteria utility function as shown in (1).

(iii) Load. In order to take the accurate context based decisionfor handover, the load of a candidate eNodeBwas additionallyconsidered. In some cases, the UE can attach to an eNodeBwhich has a greater RSS value but might be overloaded interms of connected UEs. In other words, based only theRSS and the number of associated UEs that are currentlyassociated with an eNodeB, the handover decision mayexperience the degradation of a performance. Each eNodeBtransmits its current work load to its two hop neighboreNodeBs. The eNodeB that takes a handover decision usesthe work load value of each eNodeB in a single criteria utilityfunction as shown in (1).

3.2. Candidate eNodeB Selection. The handover decision ismade in the serving eNodeB. The UE reports the mea-surement of RSS values of all the candidate eNodeBs andits own position information to the associated eNodeB.Each eNodeB sends its load estimation to 2-hop neighboreNodeBs. Then the associated eNodeB uses (1) to make theutility estimation of each alternative of the eNodeB. For eachcandidate eNodeB, three different utility values are estimatedand combined in the aggregated metric function as shown in(3). The serving eNodeB selects the eNodeB which has thehighest aggregated metric value.


eNodeB1


eNodeB2

mobility

6. Data

7. Data

16. Configure

MME

eNodeB3


UE Provider

10. Buffer4. Intere

st

8. Data

13. Data

18. D

ata

3. Interest

9. Interest to MME

5. Interest12. Interest

17. Interest

11. Interest

15. UE identifier

2. UE identifie

r

UE

Figure 3: Seamless content retrieval in the highly mobile environment.

3.3. Seamless Content Retrieval Using MME. Consumer mo-bility allows consumers to change their point of attachmentwithout disrupting connectivity. The point of attachmentselection in the wireless networking environment is basedon an aggregation of different utility functions for decisionprocesses. Hence, we define here multicriteria that are ableto integrate the end devices’ different choice metrics to selectthe best point of attachment. This paper proposes a soft-handover approach where a new connection is establishedwith the new eNodeB before breaking the current connection.This is almost similar to the follow-me service that is themodern trend of mobile communications. In the proposedapproach, UEs send the measurement reports to the servingeNodeB. Then the eNodeB follows the procedure mentionin Section 3.1 to decide whether the UE will move to a neweNodeB or not. If the serving eNodeB decides the necessityfor a new eNodeB to continue the seamless content retrievalof the UE, it forwards the Interest and related informationto the new eNodeB; then the new eNodeB forwards theInterest to the most appropriate content provider to retrievethe content. If the content retrieval is successful, the UEreleases the connection with the old eNodeB and continuesthe content transfer using the new eNodeB.Themessage flowfor seamless content retrieval of the UE is shown in Figure 3.

The detailed operational procedure for seamless contentretrieval of the UE in the content-centric LTE network isdescribed below.

(i) Step 1: an UE establishes a new connection toeNodeB1.

(ii) Step 2: eNodeB1 sends details about the UE, e.g., UEidentifier, interface identifier to MME.

(iii) Step 3: the UE sends an Interest message via eNodeB1.When eNodeB1 receives the Interest, it follows thesame procedure performed by a CCN node.

(iv) Step 4: eNodeB2 receives the Interest message fromeNodeB1 and follows the same procedure performedby a CCN node. After doing look-up on its CS andPIT, it forwards the Interest to the mobile contentsource (UE Provider).

(v) Step 5: after forwarding the Interest message to thecontent source, it adds PIT entry to forward contentin future.

(vi) Step 6: when the content source receives the Interestmessage, it looks up its CS. When the matchingcontent is found, it replies back with the Data messageas a response through the arrival interface of theInterest message.

(vii) Step 7, 8: eNodeB2 and eNodeB1 forward Data packetsto the UE.

(viii) Step 9: during the ongoing content transfer, theeNodeB1 estimates the mobility prediction anddecides whether it will move from eNodeB1 or notusing (3), as in Section 3.1. If the serving eNodeB1finds the best candidate for content transmission,the eNodeB1 sends the chunk Interest to the MME


Regular buffer

Separate buffer

Figure 4: Separate buffer to store and forward data.

with preceding the Interest with the best candidateeNodeB.

(ix) Step 10: eNodeB1 receives the Data packet but cannottransmit it to the UE efficiently. In our proposedmechanism,wemaintain a separate buffer as shown inFigure 4 to store and forward theData packets for latertransmission. If an eNodeB tries to buffer the packetsin the main queue, buffer overflow may hamperthe normal operation of the eNodeB. To solve theproblem, we maintain an extra buffer so that all theData packets for upcoming handover can be storedseparately by stamping as buffered packets. To avoidbuffer overflow it checks the current queue statususing the Exponentially Weighted Moving Average(EWMA) formula as shown in the following formula

𝑄𝑎V𝑔 = (1 − 𝛼) ∗ 𝑄𝑎V𝑔 + 𝑄𝑐𝑢𝑟𝑟 ∗ 𝛼 (10)

where 𝛼 is a weight factor and Q𝑐𝑢𝑟𝑟 is the currentqueue size. eNodeB1 can buffer packets if the currentstatus falls below or is equal to a minimum thresholdcalled 𝑄𝑇ℎ calculated as shown in the followingformula

𝑄𝑇ℎ = w ∗ 𝑄𝑠𝑖𝑧𝑒 (11)

where 𝑤 is a weight factor. Based on the extra bufferoccupancy, if the value of Q𝑎𝑣𝑔 is ≤Q𝑇ℎ, eNodeB1discards the Data packet. MME sends the Interestmessage to eNodeB1 to transfer the buffered Datamessage to the desired eNodeB where the UE wantsto move, e.g., eNodeB3.

(x) Step 11: MME forwards the received Interest messageto the desired eNodeB where the UE wants to move,e.g., eNodeB3.

(xi) Step 12: eNodeB3 receives the Interest message andforwards the Interest message to the eNodeB2. It alsoreceives the buffered Data messages sent by eNodeB1.

(xii) Step 13: after receiving the Interest message fromeNodeB3, eNodeB2 sends the Data message as aresponse to the Interest message.

(xiii) Step 14: The UE terminates the connection to theeNodeB1 and uses the same physical handover oper-ation like LTE and makes a connection to eNodeB3.It is a UE initiated handover, and the UE establishes anew connection to eNodeB3.

(xiv) Step 15: eNodeB3 sends the UE details, e.g., UEidentifier, interface identifier to MME.

(xv) Step 16: since the UE’s identifier is already registeredin the MME, MME identifies that the UE is movedfrom eNodeB1 to eNodeB3. MME sends the Interestmessage to eNodeB1 and eNodeB2 to reconfigure theprevious path.

(xvi) Steps 17, 18: the UE sends the Interest to the neweNodeB, e.g., eNodeB3, and continues to receive thecontent seamlessly using the new eNodeB.

3.4. Seamless Content Delivery Using MME. Provider mobil-ity allows sources to relocate without disrupting contentavailability. In order to reduce handover latency and the costof the provider mobility in CCN, we propose a new mech-anism that can allow soft-handover approach where a newconnection is established to a new eNodeB before breakingthe old connection. Once a handover occurs, the producerwill update its prefix to match the new location (e.g., when aproducer named /prefix moves from eNodeB1 to eNodeB2,the producer’s name will change from /eNodeB1/prefix to/eNodeB2/prefix).

If the producer changes its attachment point, i.e., eNodeB,its location name becomes invalid and Interests from the UEand eNodeB will no longer reach the content source. As soonas it is assigned a new location name at the new eNodeB,the eNodeB and MME update the binding information.Consumers exploit CCN’s multipath forwarding to handlehandovers. Due to mobility, if the content name is changedand the producer receives the old named Interest message,then it can use similarity matching mechanism to satisfythe Interest. Let a content be denoted by 𝑑 which consistsof naming components or attributes, e.g., location and typedenoted by V𝑑. Thus the content name is represented byd = (V1𝑑, V2𝑑, . . . , V𝑚𝑑). Then for similarity matching, thefollowing formula shown in (12) is used to calculate thesimilarity, 𝑆1,2, between the content item 𝑑1 and the contentitem 𝑑2.

𝑆1,2 =∑𝑚𝑖=1𝑤𝑖 ∗ 𝐵 (V𝑑1𝑖 , V𝑑2𝑖 )

∑𝑚𝑖=1𝑤𝑖(12)

where 𝑚 is the number of qualitative attributes that presentthe content, e.g., movie, video, size, and length, 𝑤𝑖 is theweight for each attribute based on its significance, and B (i,m)is a similarity function returning 1 if V𝑑1𝑖 = V

𝑑2𝑖 and 0 otherwise.

Using the similarity value obtained from (9) based on therequested content and available content and also based on thesignificance of the data, the producer can determine whetherthe Interest was satisfied or not. The detailed operationalprocedure for seamless data delivery of a provider UE in thecontent-centric LTE network is described below.Themessageflow regarding seamless data delivery of the provider UE isshown in Figure 5.

(i) Step 1: UE (provider) establishes a connection toeNodeB1 and registers its content name to eNodeB1.


eNodeB1

1. Establish connection &

Prefix Registration

eNodeB2

eNodeB3

2. UE identifier 7. Data3. Interest

4. Interest

8 . Data

13. Int

erest

9. Establish n ew connection

& prefix registration

10. UE identifier

UE Consumer

11. Configure

14. Data

MME

11. Configure

12. Interest

15. Data

5. Intere

st

UE (Provider Mobility)

6. Data

Figure 5: Seamless content delivery with content provider mobility.

(ii) Step 2: eNodeB1 sends the UE the details, e.g., UEidentifier, interface identifier to MME.

(iii) Step 3. UE (consumer) sends Interest message viaeNodeB2. When eNodeB2 receives the Interest, itfollows the same procedure performed by a CCNnode.

(iv) Step 4: eNodeB1 receives the Interest packet fromeNodeB2 and follows the same procedure performedby a CCN node. After doing look-up on its CS andPIT, it forwards the Interest to the mobile contentsource.

(v) Step 5: after forwarding the Interest packet to thecontent source, it adds PIT entry to forward data inthe future.

(vi) Step 6: when the content source receives the Interestpacket, it looks up its CS. When the matching contentis found, UE replies back with the Data packet as aresponse through the arrival interface of the Interestpacket.

(vii) Step 7, 8: eNodeB1 and eNodeB2 forward the Datapacket to the UE.

(viii) Step 9: during the ongoing content transfer, theeNodeB1 estimates the mobility prediction using (3)and decides whether UE will move from eNodeB1or not. If the serving eNodeB1 finds any best can-didate for content transmission, it uses the samephysical handover operation like LTE and triggers a

connection to eNodeB3. UE establishes connection toeNodeB3.

(ix) Step 10: eNodeB3 sends the UE details, e.g., UE iden-tifier, interface identifier to MME.

(x) Step 11: Since the UE’s identifier is already registeredin the MME, MME identifies that the UE is movedfrom eNodeB1 to eNodeB3. MME sends the Interestpacket to eNodeB1 and eNodeB2 to reconfigure theirpath.

(xi) Steps 12, 13: eNodeB1 forwards the Interest packetto eNodeB3. When eNodeB3 receives the Interest,it follows the same procedure performed by a CCNnode. For content matching, it can use the similaritymatching equation (9). After doing look-up on its CSand PIT, it forwards the Interest to the mobile contentsource.

(xii) Step 13: UE (Content Producer) receives the oldnamed Interest packet; it uses the similarity matchingequation (9) to satisfy the Interest that ismatchedwithany appropriate content.

(xiii) Steps 14, 15: after receiving the Interest fromeNodeB3,eNodeB2 sends the Data packet as a response to theInterest packet to eNodeB1.

4. Performance Evaluation

This section presents simulation results in order to demon-strate that the content-centric LTE network is well suited to


Another Network

Content Provider

Content Consumer

eNodeB

PGW/MME

Figure 6: A CCN-based mobile network topology for simulation.

today’s communication trend. This section also analyzes theperformance of the content-centric LTE network with ourproposedmobilitymanagement scheme and compares it withthe prediction-based LTE [24] and mobility management forCCN [21].

The simulationswere performed using CCNx, LENA/NS-3, Direct Code Execution (DCE) on VMware, and Ubuntu12.04 environment. We used a simulation topology as shownin Figure 6. The size of the content which is transferredbetween the mobile producer and mobile consumer is 1.1Mbytes. The number of eNodeBs was considered to be threeto show the mobility scenario of the UEs. The number ofUEs covered by each eNodeB varies from 1 to 10 to show theeffectiveness of our proposed mechanisms and also to showthe content delivery efficiency in the low load and high loadenvironment. The UEs may work as a content provider or acontent consumer.

We showed the performance for different number ofcontent providers who publish the video files and differentnumber of consumers. Each content provider publishesdifferent content files after 1-minute interval in the wholesimulation time, and contents are requested randomly fromthe different consumers. We selected a half of all UEs to workas consumers and the other half as producers at each eNodeB.To show the efficiency of the proposed mobility managementmechanism, we varied the mobility speed of the UE. The UEwas placed in the boundary of its serving eNodeB to create thehandover scenario. The UEs and eNodeBs were distributed

uniformly using the grid position allocator. The simulationused a random walk mobility model, and the moving speedof UEs varied in the range from 0 km/h (stationary) to 60km/h. When the number of the UEs and the number of theeNodeBs increase, the number of content providers and thenumber of generated contents also increase and the networkbecomes heavily loaded. For the simulation simplicity andreducing the processing complexity, we assigned each weightof (3) with the equal value of 1/3. We ran the simulation 10times for each simulation configuration and took the averagevalue of the results.

We evaluated the efficiency of the proposed approach andapplicability by showing the average content transfer time, theaverage throughput observed by each consumer over time,and the average content delivery success ratio as performanceparameters.Wemeasured the average content transfer time𝑇using the following formula:

𝑇 = ∑𝑛𝑖=1 𝑇𝑖,𝑓 − 𝑇𝑖,𝑠

𝑛 (13)

where 𝑛 is the total number of UEs which are involved inreceiving the content, 𝑇𝑖,𝑠 is the time at which the UEimakes arequests to retrieve the content, and 𝑇𝑖,𝑓 is the time at whichthe UE receives the requested content. Average throughputis measured as the average number of data bytes received byall the consumer UEs per second. Data transmission successratio is the ratio of the total number of data packets received


Prediction-based LTE Mobility Management for CCN Proposed Content Delivery Mechanism

8 12 16 204No of UEs (producer, consumer)

1.01.52.02.53.03.54.04.55.05.56.06.57.07.58.08.59.09.5

Con

tent

Tra

nsfe

r Tim

e (se

c)

Figure 7: Average content transfer time.

by the all consumer UEs to the total number of Data packetssent by all the producer UEs.

4.1. Content Transfer Time. Figure 7 shows the averagecontent transfer time observed by UEs. The average contenttransfer time of our proposed CCN-based mobility manage-ment mechanism has been evaluated and compared with theprediction-based LTE and CCN-basedmobility managementmechanism in the LTE network.

The content transfer time of our proposed approach isshorter than the content transfer time of others because ofits efficient soft handover based mobility management mech-anism for both consumers and producers. The introductionof the extra buffer reduces the chance of long transmissiondelay, queuing delay, propagation delay, and processing delayat intermediate nodes in case of high mobility scenarios. Incase of high mobility cases, our proposed mechanism usesthe make-before-break approach when changing the routingpath if needed. To avoid the high cost of tunnel setup, we usethe same approach in case of consumer mobility. The mainreason is that a content is retrieved in the candidate eNodeBbefore the original handover occurred. In the proposedmechanism, a node acquires the content from the edgenetwork whereas in the other content communication a nodeacquires the content from a remote network. Therefore, therouting path, cost, and latency in the proposed mechanismare smaller.

4.2. Average Throughput. Figure 8 depicts the averagethroughput of the prediction-based LTE, CCN-based mobil-ity management mechanism, and our proposed mechanismwith the varying number of producers and consumers inrandom mobility scenarios. The throughput of our proposedmechanism is better and stable as it is able to detectand differentiate losses due to congestion, link failure, and

Prediction-based LTE_4UE Prediction-based LTE_12UE Mobility Management for CCN_4UE Mobility Management for CCN_12UE Proposed Content Delivery Mechanism_4UE Proposed Content Delivery Mechanism_12UE

0.3

0.4

0.5

0.6

0.7

0.8

�ro

ughp

ut (M

Bps)

20 30 40 50 60 70 80 9010Time (sec)

Figure 8: Average throughput observed inside the consumer.

mobility. Our in-network buffering capability does not affectthe normal operation of the network in case of high mobility,so the throughput rate is always consistent in case of ourapproach. However, the content delivery rates are hardlyaffected by mobility and tend to be stable, as shown inFigure 8.

4.3. Data Transmission Success Ratio. We also measured theperformance of reachability and continuity of ourmechanismin terms of data transmission success ratio, which implieshow much data were received correctly by the consumerin the random mobility scenario. The proposed contentsimilarity approach increases the content availability whenmobility changes the content location. Also the buffer-ing capability, fast path switch, and handover predictionreduce the packet loss rate. The simulation result showeda significant improvement in this case as illustrated inFigure 9.

5. Conclusion

In this paper, we proposed a novel content delivery mech-anism and a mobility management scheme for the evolvedcommunication architecture such as 4G/5G to make the bal-ance between the content diversity and network diversity. Wethen analyzed the performance of the proposed schemes withthe LTE network in the mobile environment. By presentingdifferent simulation results, we showed that the proposedschemes can be used as a possible solution for faster contenttransmission and seamless content delivery in the mobileenvironment. It is possible to provide accelerated, reliable,resource-efficient, and cost-effective communication, whichwill also be helpful for 5G.


Prediction-based LTE Mobility Management for CCN Proposed Content Delivery Mechanism

0.65

0.70

0.75

0.80

0.85

0.90

Succ

ess R

atio

8 12 16 204No of UEs (producer, consumer)

Figure 9: Data transmission success ratio.

Data Availability

Thesimulation parameters and results and other relevant dataused to support the findings of this study are included withinthe article.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

This research was financially supported by the Ministry ofTrade, Industry and Energy (MOTIE) and Korea Institutefor Advancement of Technology (KIAT) through the Interna-tional Cooperative R&D program. This work was supportedby Hankuk University of Foreign Studies Research Fund of2018.

References

[1] V. Jacobson, D. K. Smetters, J. D. Thornton, M. F. Plass, N. H.Briggs, and R. L. Braynard, “Networking named content,” inProceedings of the 5th ACMConference on Emerging NetworkingExperiments and Technologies (CoNEXT ’09), pp. 1–12, Decem-ber 2009.

[2] M. R. Bosunia, A. Kim, D. P. Jeong, C. Park, and S.-H.Jeong, “Enhanced multimedia data delivery based on content-centric networking in wireless networks,” Applied Mathematics& Information Sciences, vol. 9, no. 2, pp. 579–589, 2015.

[3] D. Le, X. Fu, and D. Hogrefe, “A review of mobility supportparadigms for the internet,” IEEE Communications Surveys &Tutorials, vol. 8, no. 1, pp. 38–51, 2006.

[4] D. Farinacci, V. Fuller, D. Meyer, and D. Lewis, “Locator/idseparation protocol (lisp),” IETF - draft-ietf-lisp-23, 2012.

[5] T. Koponen, M. Chawla, B.-G. Chun et al., “A data oriented(and beyond) network architecture,”Computer CommunicationReview, vol. 37, no. 4, pp. 181–192, 2007.

[6] C. Dannewitz, D. Kutscher, B. Ohlman, S. Farrell, B. Ahl-gren, and H. Karl, “Network of information (NetInf)-Aninformation-centric networking architecture,” Computer Com-munications, vol. 36, no. 7, pp. 721–735, 2013.

[7] M. D’Ambrosio, P. Fasano, M. Marchisio, V. Vercellone, andM. Ullio, “Providing data dissemination services in the futureinternet,” in Proceedings of the IEEE Global TelecommunicationsConference (GLOBECOM ’08), pp. 1–6, November 2008.

[8] A. Eriksson and B. Ohlman, “Dynamic internetworking basedon late locator construction,” in Proceedings of the IEEE GlobalInternet Symposium (GI ’07), pp. 67–72, May 2007.

[9] D. Trossen and G. Parisis, “Designing and realizing aninformation-centric internet,” IEEECommunicationsMagazine,vol. 50, no. 7, pp. 60–67, 2012.

[10] P. Jokela, A. Zahemszky, C. E. Rothenberg, S. Arianfar, andP. Nikander, “LIPSIN: Line speed publish/subscribe inter-networking,” in Proceedings of the ACM SIGCOMM 2009Conference on Data Communication (SIGCOMM ’09), pp. 195–206, August 2009.

[11] S. DiBenedetto, C. Papadopoulos, and D. Massey, “Routingpolicies in named data networking,” in Proceedings of the ACMSIGCOMMWorkshop on Information-Centric Networking (ICN’11), pp. 38–43, August 2011.

[12] G. Carofiglio, G. Morabito, L. Muscariello, I. Solis, and M.Varvello, “From content delivery today to information centricnetworking,” Computer Networks, vol. 57, no. 16, pp. 3116–3127,2013.

[13] J. Lee and D. Kim, “Proxy-assisted content sharing usingcontent centric networking (CCN) for resource-limited mobileconsumer devices,” IEEE Transactions on Consumer Electronics,vol. 57, no. 2, pp. 477–483, 2011.

[14] S. Muhammad, K. Kwangsoo, and C. Seungoh, “Cluster-basedMobility support in Content-centric Networking,” ResearchNotes in Information Science (RNIS), vol. 14, pp. 441–444, 2013.

[15] H. Feng, Z. Chen, and H. Liu, “Performance analysis ofpush-based converged networks with limited storage,” IEEETransactions on Wireless Communications, vol. 15, no. 12, pp.8154–8168, 2016.

[16] I. Loumiotis, P. Kosmides, E. Adamopoulou, K. Demestichas,and M. Theologou, “Dynamic allocation of backhaul resourcesin converged wireless-optical networks,” IEEE Journal onSelected Areas in Communications, vol. 35, no. 2, pp. 280–287,2017.

[17] A. Araldo, G. Dan, and D. Rossi, “Caching encrypted contentvia stochastic cache partitioning,” IEEE/ACM Transactions onNetworking, vol. 26, no. 1, pp. 548–561, 2018.

[18] L. Rui, S. Yang, and H. Huang, “A producer mobility supportscheme for real-time multimedia delivery in named data net-working,” Multimedia Tools and Applications, vol. 77, no. 4, pp.4811–4826, 2018.

[19] R. Tourani, S. Misra, and T. Mick, “IC-MCN: An architecturefor an information-centric mobile converged network,” IEEECommunications Magazine, vol. 54, no. 9, pp. 43–49, 2016.

[20] H. Farahat and H. Hassanein, “Optimal caching for producermobility support in Named Data Networks,” in Proceedings ofthe IEEE International Conference onCommunications (ICC ’16),May 2016.


[21] F. Ren, Y. Qin, H. Zhou, and Y. Xu, “Mobility managementscheme based on software defined controller for content-centricnetworking,” in Proceedings of the INFOCOM InternationalWorkshop onMobilityManagement in theNetworks of the FutureWorld (INFOCOM WKSHPS ’16), pp. 193–198, San Francisco,Calif, USA, April 2016.

[22] J. Costa-Requena, M. Kimmerlin, J. Manner, and R. Kantola,“SDN optimized caching in LTE mobile networks,” in Proceed-ings of the 5th International Conference on Information andCommunication Technology Convergence (ICTC ’14), pp. 128–132, October 2014.

[23] N. Zhang, M. Kimmerlin, J. Costa-Requena, and H. Hämmäin-en, “Cost efficiency of mobile in-network caching,” Interna-tional Journal of Network Management, vol. 26, no. 1, pp. 44–55,2016.

[24] H. Wang, S. Kao, C. Hsiao, and F. Chang, “A moving direc-tion prediction-assisted handover scheme in LTE networks,”EURASIP Journal onWireless Communications andNetworking,2014.

[25] R. M. Bosunia and S.-H. Jeong, “Efficient mobility support forcontent delivery in mobile communications,” in Proceedings ofthe International Conference on Information Networking (ICOIN’17), pp. 118–121, 2017.

[26] P. Vincke,Multicriteria Decision Aid, Wiley, 1992.

Research ArticleFuzzy Based Network Assignment and Link-Switching Analysisin Hybrid OCC/LiFi System

Moh. Khalid Hasan , Mostafa Zaman Chowdhury ,Md. Shahjalal , and YeongMin Jang

Department of Electronics Engineering, Kookmin University, Seoul 02707, Republic of Korea

Correspondence should be addressed to Yeong Min Jang; [email protected]

Received 3 August 2018; Accepted 4 November 2018; Published 19 November 2018

Academic Editor: Laurie Cuthbert

Copyright © 2018 Moh. Khalid Hasan et al. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.

In recent times, optical wireless communications (OWC) have become attractive research interest in mobile communication forits inexpensiveness and high-speed data transmission capability and it is already recognized as complementary to radio-frequency(RF) based technologies. Light fidelity (LiFi) and optical camera communication (OCC) are two promising OWC technologies thatuse a photo detector (PD) and a camera, respectively, to receive optical pulses. These communication systems can be implementedin all kinds of environments using existing light-emitting diode (LED) infrastructures to transmit data. However, both networkinglayers suffer from several limitations. An excellent solution to overcoming these limitations is the integration of OCC and LiFi. Inthis paper, we propose a hybrid OCC and LiFi architecture to improve the quality-of-service (QoS) of users. A network assignmentmechanism is developed for the hybrid system. A dynamic link-switching technique for efficient handover management betweennetworks is proposed afterward which includes switching provisioning based on usermobility and detailed network switching flowanalysis. Fuzzy logic (FL) is used to develop the proposed mechanisms. A time-division multiple access (TDMA) based approach,called round-robin scheduling (RRS), is also adopted to ensure fairness in time resource allocation while serving multiple usersusing the same LED in the hybrid system. Furthermore, simulation results are presented taking different practical applicationscenarios into consideration.The performance analysis of the network assignment mechanism, which is provided at the end of thepaper, demonstrates the importance and feasibility of the proposed scheme.

1. Introduction

Communication currently relies on the radio-frequency (RF)spectrum, which is overcrowded and strictly regulated [1].Because of several factors including interference, limitedresources, and human safety it is obvious that RF basedtechnologies will not be sufficient to manage the massivefuture data traffic. Wireless communication using the opticalspectrum has been regarded as a congruent solution to thespectrum congestion of RF based technologies [2–6]. Inparticular, the optical wireless technology, especially visiblelight communication (VLC), has added a new dimension inthe world of mobile communications for its huge unregulatedspectrum (up to 800 THz [7]), cost effectiveness, energy effi-ciency, and high security [3, 8, 9]. Moreover, current indoorand outdoor environments are currently heavily congested

with light-emitting diode (LED) based lighting infrastruc-tures, enabling VLC to be exploited as a complementarytechnology to RF.

Light fidelity (LiFi) is a subset of OWC technology inwhich a photo detector (PD) receives the variation in theintensity of light, which carries data bits encoded from thelight source [3, 10, 11]. A PD can detect high-speed LEDflickering, a capability that enables LiFi to support highdata rates. An extensive improvement in bandwidth reuse isobserved for LiFi technologies, resulting in excellent spectralefficiency. Because of these benefits provided by LiFi, severalarchitectures integrating LiFi and RF have been already pro-posed to enhance the quality-of-service (QoS) of users; thesearchitectures include those that manage resource allocation[12–15], dynamic handover [16, 17], energy harvesting [18],delay analysis [19], and channel assignment [20].


http://orcid.org/0000-0002-7773-3523http://orcid.org/0000-0003-1487-086Xhttp://orcid.org/0000-0002-4876-6860http://orcid.org/0000-0002-9963-303Xhttps://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://doi.org/10.1155/2018/2870518


However, LiFi cannot be efficiently utilized in daylightbecause it suffers from extensive interference generated bysunlight [3]. In indoor environments, it can suffer from thesame problem resulting from neighboring lighting infras-tructures. LiFi has a low signal-to-interference-plus-noiseratio (SINR) because it is heavily affected by the inter-ferences generated by adjacent light sources. In addition,the communication distance that can be obtained usingLiFi is comparatively short with respect to other existingtechnologies. These limitations inspire further research onthe optimum potentiality of LiFi in practical environments.

Optical camera communication (OCC) is a recentlyintroduced VLC technique that uses an image sensor toreceive optical signals [21–26]. The exponential growthin camera-mounted smart devices has enabled OCC tobe utilized in innovative application scenarios, such asindoor/outdoor positioning [27, 28], localized advertising[29], digital signage, and vehicle-to-vehicle (V2V) or vehicle-to-infrastructure (V2I) communications [30–32]. OCC hasadded significant user flexibility with the use of smartphonecameras to receive data from LEDs. Furthermore, OCCis highly stable in terms of variations in communicationdistance. Because of the limited angle-of-view (AoV) ofcameras, OCC is less affected by interferences generated fromneighboring LEDs.However, similar to LiFi, OCC has severallimitations. For example, because of the meager samplingrates of current commercial cameras, OCC offers a low datarate, which particularly decreases the user QoS.

Until now, a hybrid OCC and LiFi model had not beendeveloped. In this study, a hybrid networking architectureintegrating OCC and LiFi is proposed to enhance the userQoS. The network is assigned to users through utilizationof fuzzy logic (FL). FL is a convenient approach to mapan input to an output and is provided on the basis ofseveral truth scores ranging from 0 to 1 [33]. This method isflexible and intuitive without far-reaching complexity, whichare characteristics that lead us to choose this approach.We propose a new network assignment mechanism insidethe LED cell (the entire coverage area of the LED) forusers. Fuzzy inputs are chosen using the parameters thatdetermine the quality of both networks. The fuzzy rules aregenerated by considering real-world application scenariosof users. The center-of-gravity (CoG) method is used todefuzzify the inputs and obtain mark allocations for eachuser. Furthermore, we develop a FL-based vertical link-switching mechanism between the networking layers, asboth of the networks support user mobility. We brieflydiscuss the switching probability and corresponding networkswitching flow analysis of the hybrid system. Round-robinscheduling (RRS) [34], an existing time-division multipleaccess (TDMA) approach, is adopted to ensure fairness inresource allocation among users.

The remainder of the paper is organized as follows:Section 2 provides a system overview and an analysis onchannel parameters, which includes theoretical representa-tions of SINR for both technologies. The FL-based networkassignment mechanism, including a discussion on user QoS,is explained in Section 3. Sections 4 and 5 describe the link-switching strategy and the network switching flow process,

respectively. The performance of the assignment mechanismis evaluated in Section 6, which also includes a discussionof the outage probability and QoS performance. A briefsummary of our work is provided in Section 7. Finally,Section 8 presents future research possibilities related to ourproposed hybrid infrastructure.

2. System Overview

2.1. Hybrid System Architecture. In this study, a hybridOCC/LiFi networking layer is considered. Taking usermobil-ity into account, this hybrid system can serve multiple users.Therefore, the hybrid system is suitable for any roamingor stationary user. A particular LED is configured by twoparallel LED-driving circuitries. Although both technologiesuse the same optical spectrum, no interference will begenerated because the TDMA based RRSmethod is exploitedto allocate time resources when there are multiple users.A generalized block diagram of our proposed architectureis shown in Figure 1. The PD can receive high-rate LEDflickering, whereas a camera cannot. Current commercialcameras are configured with low frame rates (in most cases,30–50 frames per second). This configuration particularlyreduces the modulation bandwidth of OCC [35, 36]. It is alsoworth noting that the LEDflickering must not be observed byhuman eyes (equivalent to a threshold of approximately 100Hz [37]).

2.2. OCC Channel Model. For a VLC system, the route foroptical signal transmission has two components: line-of-sight (LOS) and non-line-of-sight (NLOS). Because of thenature of camera pixels, region-of-interest (RoI) mechanismsare applied for OCC, by which the reflection componentof the transmitted signal is spatially separated from theLOS component [23]. An indoor hybrid system with thetransmitter and receiver presented at Tx and Rx, respectively,is illustrated in Figure 2. The LED cell represents the entirecoverage area of the LED.

The LOS channel for optical signal transmission is mod-eled by Lambertian radiant intensity, which is represented bythe following equation [38]:

𝑅𝑜 (𝛼) = (𝑚𝑙 + 1) cos𝑚𝑙 (𝛼𝑖𝑟)2𝜋 (1)

where 𝛼𝑖𝑟 signifies the angle of irradiance of the LED. 𝑚𝑙 isthe Lambertian emission index, which originates from theradiation angle Ψ1/2, called the radiation semiangle of theLED;𝑚𝑙 is defined as

𝑚𝑙 = −logcosΨ1/22 (2)We assume that the Euclidean distance between Tx and

Rx is 𝑑𝑎,𝑏, which is calculated from the horizontal distance𝑑𝑏,𝑥 and the vertical distance 𝑑𝑎,ℎ (𝑑𝑎,𝑏 = √𝑑2𝑎,ℎ + 𝑑2𝑏,𝑥). Theoverall DC channel gain for OCC is formulated as [12]

𝐻𝐼𝑆𝑡,𝑟 = 𝑔𝑜𝑝 cos (𝛼𝑖𝑛) Δ 𝑜𝑐𝑐𝑅𝑜 (𝛼)𝐴𝑐𝑑2𝑎,𝑏 (3)


Input data

Encoder Modulator LED driver

A/D

Output data

AmplificationPD

Framesampling

OCC transmitter OCC receiver

LED

D/A Modulator LED driver

Demodulator

Demodulator

Decoder

ImageSensor

LiFi transmitter LiFi receiver

Figure 1: Basic block diagram of the hybrid OCC/LiFi architecture.

LED

LiFi dongle

FOV of PDAoV of Camera

Smartphone camera

Gateway db,x

da,ℎ − ir

info

da,b

２Ｒ

ao

４Ｒ

LED cell

Figure 2: Data transmission model for the hybrid network.

where 𝛼𝑖𝑛 implies the corresponding angle of incidence,𝑔𝑜𝑝 represents the gain of the optical filter, and Δ 𝑜𝑐𝑐 is arectangular function whose value implies that the channelhas no gain if the LED remains outside of the angle-of-view(AoV) of camera receiver. If 𝛽𝑎𝑜V is the AoV of the camera,then Δ 𝑜𝑐𝑐 is represented as

Δ 𝑜𝑐𝑐 = {{{0, 𝛼𝑖𝑛 ≥ 𝛽𝑎𝑜V1, 𝛼𝑖𝑛 < 𝛽𝑎𝑜V (4)

𝐴𝑐 is the area of the entire image of the LED projected inthe image sensor. It is often signified by the number of pixelsoccupied by the image. If𝜌 denotes the pixel edge length, thenthe projected area is

𝐴𝑐 = 𝐴 𝑙𝑓𝑜2

𝜌2𝑑2𝑎,𝑏

(5)

where 𝑓𝑜 denotes the focal length of the camera and 𝐴 𝑙represents the physical area of the LED.

There is a minimum area of the projected image in theimage sensor, below which the transmitted data cannot be

decoded.The power received by the image sensor in this caseis termed as the threshold power and expressed as

𝑃𝐼𝑆𝑡ℎ = argmin𝑃𝐼𝑆𝑟= argmin[𝑔𝑜𝑝 cos (𝛼𝑖𝑛) Δ 𝑜𝑐𝑐𝑅𝑜 (𝛼) 𝐴𝑐𝑃𝑡𝑑2

𝑎,𝑏

] (6)

where 𝑃𝐼𝑆𝑟 is the power received by the image sensor and 𝑃𝑡denotes the optical power transmitted by the LED.

Most existing commercial cameras offer a low AoV. Asa result, the LOS components of neighboring LEDs do notreach inside the camera’sAoV.Moreover, asmentioned above,introducing RoI signaling techniques significantly reducesthe effect of the reflected components. Thus, OCC offers anexcellent SINR, which is represented as

𝑆𝐼𝑁𝑅𝑜𝑐𝑐 = (𝜁𝑐𝑃𝑡𝐻𝐼𝑆𝑡,𝑟)2

∑𝑁𝑖=0 (𝜁𝑐𝑃𝑡𝐻𝑜𝑐𝑐𝑖,𝑟 )2 + 𝑁𝑜𝑓𝑟(7)

where 𝜁𝑐 denotes the optical-to-electrical conversion effi-ciency at the image sensor, 𝑁𝑜 is the spectral density of the


noise power, 𝑓𝑟 is the sampling rate of the camera, 𝑁 isthe number of interfering transmitters, and 𝐻𝑜𝑐𝑐𝑖,𝑟 is the DCgain from these transmitters. The channel capacity can beexpressed by the Shannon capacity formula [23], which is

𝐶𝑜𝑐𝑐 = 𝑓𝑟𝑊𝑠log2 (1 + 𝑆𝐼𝑁𝑅𝑜𝑐𝑐) (8)where𝑊𝑠 represents the number of data symbols transmittedto the pixels within each image frame.

2.3. LiFi Channel Model. The NLOS part of the transmittedsignal is disregarded in terms of LiFi because our basebandmodulation bandwidth 𝐵 is 20 MHz, which does not exceedthe maximum allowable value [16, 39]. Thus, the LOS trans-mission model for LiFi is represented as

𝐻𝑃𝐷𝑡,𝑟 = 𝑔𝑜𝑝𝑔𝑐𝑜𝑛 cos (𝛼𝑖𝑛) Δ 𝑙𝑖𝑓𝑖𝑅𝑜 (𝛼)𝐴𝑝𝑑2𝑎,𝑏

(9)

where 𝐴𝑝 denotes the physical area of the PD sensitive tolight and 𝑔𝑐𝑜𝑛 is the gain of the optical concentrator, whichis a function of the refractive index and field-of-view (FoV)of PD.The rectangular function Δ 𝑙𝑖𝑓𝑖 is expressed as

Δ 𝑙𝑖𝑓𝑖 = {{{0, 𝛼𝑖𝑛 ≥ 𝛽𝑓𝑜V1, 𝛼𝑖𝑛 < 𝛽𝑓𝑜V (10)

where 𝛽𝑓𝑜V denotes the PD FoV. The PD should receive acertain amount of power to generate a minimum electricalcurrent in order to decode the actual sent data bits. Thethreshold power of LiFi is denoted as

𝑃𝑃𝐷𝑡ℎ = argmin𝑃𝑃𝐷𝑟= argmin[𝑔𝑜𝑝𝑔𝑐𝑜𝑛 cos (𝛼𝑖𝑛) Δ 𝑙𝑖𝑓𝑖𝑅𝑜 (𝛼) 𝐴𝑝𝑃𝑡𝑑2

𝑎,𝑏

] (11)

where 𝑃𝑃𝐷𝑟 denotes the total amount of power received bythe PD. LiFi uses an intensity based modulation scheme; assuch, LiFi is affected by the interference generated by neigh-boring LEDs and other background lights. This interferenceultimately results in reducing the SINR to a great extent,as LED infrastructures are commonly developed for indoorenvironments. Several studies [16, 40] have investigated theSINR in terms of LiFi, which can be expressed as

𝑆𝐼𝑁𝑅𝑙𝑖𝑓𝑖 = ((𝜁𝑝√𝑃𝑒/𝑃𝑡) 𝑃𝑡𝐻𝑃𝐷𝑡,𝑟 )2

∑𝑁𝑖=0 ((𝜁𝑝√𝑃𝑒/𝑃𝑡) 𝑃𝑡𝐻𝑃𝐷𝑖,𝑟 )2 + 𝑁𝑜𝐵(12)

where 𝜁𝑝 is the optical-to-electrical conversion efficiency atthe PD and 𝑃𝑒 is the amount of electrical power convertedafter receiving the optical signals. The LiFi channel capacitycan also be calculated from the Shannon capacity formula,which is

𝐶𝑙𝑖𝑓𝑖 = 𝐵 log2 (1 + 𝑆𝐼𝑁𝑅𝑙𝑖𝑓𝑖) (13)

3. FL-Based Network Assignment

Inside the hybrid network, the network is selected accordingto the type of service and quality that the user requires. FL isdispensed to assign a particular user to a network. Instead ofmaking decisions for choosing a network in a hybrid systemin terms of Boolean logic (only true or false values), the FL-based assignment considers truth values of variables rangingfrom0 to 1 [33, 41–43].We apply theMamdani fuzzy inferencesystem to evaluate our proposed scheme; this system includesthree principal steps: fuzzification of input variables, rulesevaluation, and defuzzification.

Fuzzification refers to the process of transforming thecrisp inputs into degrees of functional blocks through usingthe different types of fuzzifiers, called membership func-tions. A fuzzy set is graphically represented by membershipfunctions. For example, a triangular function is presented inFigure 3(a) and described as

𝜇 (𝑥; 𝑎𝑇, 𝑏𝑇) ={{{{{{{{{

0, 𝑥 ≤ 𝑎𝑇 [𝑅𝑒𝑑 𝑙𝑖𝑛𝑒]𝑥 − 𝑎𝑇𝑏𝑇 − 𝑎𝑇 , 𝑎𝑇 ≤ 𝑥 ≤ 𝑏𝑇 [𝐵𝑙𝑢𝑒 𝑙𝑖𝑛𝑒]1, 𝑥 ≥ 𝑏𝑇 [𝐺𝑟𝑒𝑒𝑛 𝑙𝑖𝑛𝑒](14)

where 𝑎𝑇 and 𝑏𝑇 are the breakpoints of the membershipfunctions and 𝑥 is a particular input.

We considered four input variables to perform the net-work assignment mechanism: data rate requirement, SINRrequirement, amount of instantaneous received power, andLOS Euclidean distance between the access point (AP) andthe receiver. The variables are chosen on the basis of appli-cation scenarios. For example, if a user wants to localize itsposition, it will definitely need an excellent SINR rather thanhigh data rate to minimize the localization resolution. On thecontrary, both data rate and SINRmust be high for a real-timevideo call. Moreover, the instantaneous power significantlycontributes to determining the bit-error performance ofconnectivity. In addition, a low received power degrades theuser’s QoS level by increasing the outage probability to a greatextent. On the other hand, the maximum communicationdistance varies for different optical wireless systems anduser achieves satisfactory QoS when the communicationdistance is short. A long distance between the LED andreceiver increases the interference for LiFi, although OCCis less affected by interference. In particular, the maximumcommunication distance for LiFi is very short compared toOCC for stable communications.

The membership functions are chosen on the basis ofseveral experiments involving the use of training data. Thegrades of the membership functions are assigned accordingto the effect of variations in the value of a particular input.Figure 3(b) shows an illustration of the fuzzification of theSINR requirement of a specific user on the basis of servicetype and quality. The procedure is characterized by four dif-ferent membership grades: low, average, high, and excellent.These grades are distributed from –10 to 60 dB. As shown in


1

0

0.2

0.4

0.6

0.8

Ta Tb

(a)

0.5

0

1SINR (dB)

Area of higher membership degreeArea of lower membership degree

p q r s

AverageHighExcellent

Low

Membership grades

(b)

Figure 3: Fuzzification process: (a) a generalized triangular function; (b) SINR requirement.

Figure 3(b), the four membership grades can be representedin a similar approach, which is

𝐿𝑜𝑤 → 𝜇 (𝑥; 𝑝, 𝑞) , 𝑥 ≥ 𝑝𝐴V𝑒𝑟𝑎𝑔𝑒 → {{{

1 − 𝜇 (𝑥; 𝑝, 𝑞) , 𝑥 > 𝑝𝜇 (𝑥; 𝑞, 𝑟) , 𝑥 ≥ 𝑞

𝐻𝑖𝑔ℎ → {{{1 − 𝜇 (𝑥; 𝑞, 𝑟) , 𝑥 > 𝑞𝜇 (𝑥; 𝑟, 𝑠) , 𝑥 ≥ 𝑟

𝐸𝑥𝑐𝑒𝑙𝑙𝑒𝑛𝑡 → 1 − 𝜇 (𝑥; 𝑟, 𝑠) , 𝑥 < 𝑠

(15)

The chosen SINR values of the breakpoints are –10, 10,30, and 40 dB. For example, if a user requires an SINR ofaround 25 dB, then the user will be categorized as “average” inthe fuzzification process. Other inputs are fuzzified througha similar approach. However, the status of the membershipgrades is varied according to the numerical information ofthe input variables. For example, the data rate requirement isfuzzified through three membership grades: low, average, andhigh.

After fuzzifying the inputs, different rules are used toevaluate the performance of the hybrid system [41]. Theseif/then rules are generated by assigning a membership gradeto each of the input variables, and a decision is made aftermultiplying (also can be referred to as “and” operation) therules. For example, if the data rate requirement is low, theSINR requirement is excellent, and the instantaneous receivepower is medium, then the user will be connected via LiFi forthe shortest distance between the light source and receiver(or OCC for the highest distance). It is worth noting herethat the rules are comprehensive and are generated keepingthe nature and quality of the user requirements in mind. Ingeneral, rules are the guidelines generated according to the

membership functions and serve as a basis for why we choosea particular network in a specific kind of service scenario.

The network assignment procedure is illustrated inFigure 4. The user must remain inside the LED cell in orderto get connected via LiFi or OCC. However, the connectionpossibility significantly depends on the FoV or AoV offeredby the PD or camera, respectively. Because the effects of theNLOS components on the optical signal are disregarded, theLED must appear inside the coverage area of the receiver.After getting a new network access request (NAR) from theuser, the service type will be investigated. The examinationon the input variables will be initiated immediately followingthe investigation. Then, the system will go through thefuzzification process described earlier.

Subsequently, the rules are employed and evaluated. Thelast stage of the network assignment mechanism is the markallocation, a process that is also referred to defuzzification.The mark indicates the possibility of choosing a network inthe network assignment process. Two separate outputs areconsidered for LiFi and OCC. Both outputs are characterizedwith triangular membership functions. We have consideredfive membership grades for each output to obtain a preciseresult in the network selection mechanism. In this paper,the mark is termed as network assignment factor (NAF) anddenoted as 𝜗𝑙𝑖𝑓𝑖 and 𝜗𝑜𝑐𝑐 for LiFi and OCC, respectively. Wehave adopted CoG [33] method for defuzzification becauseit shows better performance results than the bisector-of-area(BoA)method,which is realized through several experimentson training data. The NAF is provided as a crisp value by theCoG method, which is represented as

For LiFi, 𝜗𝑙𝑖𝑓𝑖 = ∫1

0𝑧𝜇𝑙 (𝑧) 𝑑𝑧

∫10𝜇𝑙 (𝑧) 𝑑𝑧

For OCC, 𝜗𝑜𝑐𝑐 = ∫1

0𝑧𝜇𝑐 (𝑧) 𝑑𝑧

∫10𝜇𝑐 (𝑧) 𝑑𝑧

(16)


Start

New NAR?

Detect service type

Analyze crisp input variables

Fuzzy inference systemEvaluate IF-THEN rules

Defuzzification

NAF examination

Accept to OCC Accept to LiFi

Data transmission and processing

User inside LED cell

Yes No

Yes

Yes

No

Allocate time resource to remaining users

Is lifi ≤ occ ?

Figure 4: Admission strategy for a new network access request.

where ∫10𝜇𝑙(𝑧)𝑑𝑧 indicates the total area of the region after

combining all the membership functions. NAF values rangefrom 0 to 1. In fact, a higher NAF increases the possibilityof choosing a network. Thus, when a network access isrequested by a new user, the NAFs of both networks will becompared. The network with the higher NAF will be chosen.For

Convergence of Heterogeneous Wireless Networks for 5G-and-Beyond Communications...

Documents

Transcript of Convergence of Heterogeneous Wireless Networks for 5G-and-Beyond Communications...