Analysis of Handoff Schemes in Wireless Mobile …...Alagu S#1 and Meyyappan T*2 #1Faculty,...

IJCES International Journal of Computer Engineering Science , Volume1 Issue2, November 2011

ISSN : 2250:3439

https://sites.google.com/site/ijcesjournal

http://www.ijces.com/

1

Analysis of Handoff Schemes in Wireless Mobile Network

Alagu S#1

and Meyyappan T*2

#1Faculty, Department of Information Technology

Rai Business School, Chennai, TamilNadu, India

[email protected]

*2Associate Professor, Department of Computer Science and Engineering

Alagappa University, Karaikudi, TamilNadu, India

[email protected]

Abstract. This paper analyses the different traffic schemes for handoff handling and call blocking

attempts. As traffic in mobile cellular networks increases, Handoffs will become an increasingly

important issue and as cell sizes shrink to accommodate an increasingly large demand of services,

newer more efficient handoff schemes need to be used. In this paper the author analyses the various

Handoff schemes for multiple traffic system and simulates an ATM based wireless Personal

Communication Network to implement the non-preemptive Measurement Based Prioritization Scheme

(MBPS).

Keywords: Handoff, Threshold, Deterioration, Tethered, Channel, Hysteresis.

INTRODUCTION

Mobility is the most important feature of a wireless cellular communication system. Usually, continuous

service is achieved by supporting handoff (or handover) from one cell to another. Handoff is the process of

changing the channel (frequency, timeslot, spreading code or combination of them) associated with the

current connection while a call is in progress [1]. It is often initiated either by crossing a cell boundary or

by deterioration in the quality of the signal in the current channel. Poorly designed handoff schemes tend to

generate very heavy signaling traffic and thereby a dramatic decrease in quality of service (QOS). The

reason why handoffs are critical in cellular communication systems is that neighboring cells are always

using a disjoint subset of frequency bands, so negotiations must take place between the mobile station

(MS), the current serving base station (BS) and the next potential BS. Other issues like Decision making

and priority strategies during overloading may also influence the overall performance [2].

Existing Work

The existing works in the related area are discussed in Section 2 & 3. In these sections the earlier research

work on different Handoff schemes are discussed, which covers the different traffic model to handle

handoff and new call attempt.

Proposed approach

The objectives of this research work is to analyze the different schemes in traffic model and to simulate an

ATM based wireless personnel communication network and generate Constant Bit Rate traffic for it. Also

the author implements MBPS – Measurement Based Prioritization Scheme for handling handoff failure in

the above said ATM based wireless personnel communication network. The proposed work is discussed

from section 4 to section 6. The results for the same are depicted in section 7.

mailto:[email protected]



ISSN : 2250:3439



2

HANDOFF INITIATION

Handoff is the mechanism that transfers an ongoing call from one cell to another as a user moves through

the coverage area of a cellular system. The handover process is initiated by the issuing of handover request.

The power received by the MS from BS of neighboring cell exceeds the power received from the BS of the

current cell by a certain amount [3]. This is a fixed value called the handover threshold. For successful

handover, a channel must be granted to handover request before the power received by the MS reaches the

receiver‘s threshold. The handover area is the area where the ratio of received power levels from the

current and the target BS‘s is between the handover and the receiver threshold. Each handoff requires

network resources to reroute the call to the new base station. Minimizing the expected number of handoffs

minimizes the switching load. Another concern is delay. If the handoff does not occur quickly, the quality

of service [QoS] may degrade below an acceptable level. Minimizing delay also minimizes co-channel

interference. During handoff there is brief service interruption. As the frequency of these interruptions

increases the perceived QoS is reduced. The chance of dropping a call due to factors such as the availability

of channels increases with the number of handoffs attempts. As the rate of handoff increases, handoff

algorithms need to be enhanced so that the perceived QoS does not degrade and the cost to cellular

infrastructure does not increase.

Numerous studies have been done to determine the shape as well as the length of the averaging window

and the older measurements may be unreliable. Fig.1 shows a MS moving from one BS (BS1) to another

BS (BS2). The mean signal strength of BS1 decreases as the MS moves away from it. Similarly, the mean

signal strength of BS2 increases as the MS approaches it [4][17].

Fig. 1. Signal strength and hysteresis between two adjacent BSs for potential handoff [17].

Relative Signal Strength

This method selects the strongest received BS at all times. The decision is based on a mean measurement of

the received signal [5]. In Fig.1, the handoff would occur at position A. This method is observed to provoke

too many unnecessary handoffs, even when the signal of the current BS is still at an acceptable level.

Relative Signal Strength with Threshold

This method allows a MS to handle handoff only if the current signal is sufficiently weak (less than

threshold) and the other is the stronger of the two. The effect of the threshold depends on its relative value

as compared to the signal strengths of the two BSs at the point at which they are equal. If the threshold is

higher than this value, say T1 in Fig.1, this scheme performs exactly like the relative signal strength

scheme, so the handoff occurs at position A. If the threshold is lower than this value, say, T2 in Fig.1, the


ISSN : 2250:3439



3

MS would delay handoff until the current signal level crosses the threshold at position B. In the case of T3,

the delay is so long that the MS drifts too far into the new cell. This reduces the quality of the

communication link from BS1 and may result in a dropped call [5].

Relative Signal Strength with Hysteresis

This scheme allows the MS to handle handoff only if the new BS is sufficiently stronger (by hysteresis

margin h in figure 1) than the current one. In this case the handoff would occur at point C. This technique

prevents the Ping-Pong effect, the repeated handoff between two BSs caused by rapid fluctuations in the

received signal strengths from both BSs [6].

Relative Signal Strength with Hysteresis and Threshold

This scheme hands a MS over to a new BS only if the current signal level drops below a threshold and the

target BS is stronger than the current one by a given hysteresis margin. In Figure 1, the handoff would

occur at point D if the threshold is T3.

HANDOFF SCHEMES

In urban mobile cellular systems, especially when the cell size becomes relatively small, the handoff

procedure has a significant impact on system performance. Blocking probability of originating calls and the

forced termination probability of ongoing calls are the primary criteria for indicating performance. In this

section several existing traffic models and handoff schemes are discussed.

Traffic Model

In a mobile cellular radio system it is important to establish a traffic model before analyzing the

performance of the system. Several traffic models have been established based on different assumptions

about user mobility.

Hong and Rappaport‟s Traffic Model

This scheme proposes a traffic model for a hexagonal cell (approximated by a circle) [7]. They assume that

the vehicles are spread evenly over the service area; thus the location of a vehicle when a call is initiated by

the user is uniformly distributed in the cell. They also assume that the vehicle initiating a call moves from

the current location in any direction with equal probability and that this direction does not change while the

vehicle remains in the cell.

From these assumptions they showed that the arrival rate of handoff calls is

λH = Ph (1-Bo) λo

1-Phh (1-P'f)

Where

Ph = the probability that the new call that is not blocked would require at least one handoff

Phh = the probability that a call that has already been handed off successfully would require another

handoff

Bo = the blocking probability of originating calls

P'f = the probability of handoff failure

λo = the arrival rate of originating calls in a cell


ISSN : 2250:3439



4

The probability density function (pdf) of channel holding time T in a cell is derived as

FT(t) = µceµct

+ ((eµct) / (1+γc) ) [fTn(t) + γcfTh(t)]-((µce

µct)/(1+γc))[FTn(t)+γcFTh(t)]

where

fTn(t) = the pdf of the random variable Tn as the dwell time in the cell for an originated call

fTh(t) = the pdf of the random variable Th as the dwell time in the cell for a handed-off call

FTn(t) = the cumulative distribution function (cdf) of the time Tn

FTh(t) = the cdf of the time Th

1/µc = the average call duration

γc = Ph (1-Bo) / [1-Phh(1-P'f)]

El-Dolil et al.‟s Traffic Model

An extension of Hong and Rappaport‘s Traffic model to the case of highway microcellular radio

network has been done by El-Dolol et al. [8]. The highway is segmented into microcells with small BSs

radiating cigar-shaped mobile radio signals along the highway. With these assumptions they showed the

arrival rate of handoff calls is

λH = (Rcj – Rsh)Phi + RshPhh

where

Phi = the probability that a MS needs a handoff in cell i

Rcj = the average rate of total calls carried in cell j

Rsh = the rate of successful handoffs

The pdf of channel holding time T in a cell is derived as

fT(t) = ((µc + µni)/(1+G))e(µc+µni)t

+ ((µc+µh)/(1+G))e(µc+µh)t

where

1/µni = the average channel holding time in cell i for a originating call

1/µh = the average channel holding time for a handoff call

G = the radio of the offered rate of handoff requests to that of originating calls

Steele and Nofal‟s Traffic model

Steele and Nofal [9] studied a traffic model based on city street microcells, catering to pedestrians making

calls while walking along a street. From their assumptions, they showed that the arrival rate of handoff call

is

λH = Σ6

m=1[λo(1-Bo)Phβ+λh(1-P'f)Phhβ]

where

β = the fraction of handoff calls to the current cell from the adjacent cells

λh = 3 λo(1-Bo) Phβ

Ph = the probability that the new call that is not blocked would require at least one handoff

The average channel holding time T in a cell is

_ (1+α1)(1-γ) γ (1+α2) α1(1-γ) + γα2


ISSN : 2250:3439



5

T = -------------- + ------------ + --------------------

µw+µc µo+µc µd+µc

where

1/ µw = the average walking time of a pedestrian from the onset of the call until he reaches the boundary

of the cell

1/ µd = the average delay time the pedestrian spends waiting at the intersection to cross the road

1/ µo = the average walking time of a pedestrian in the new cell

α1 = µwPdelay/( µd - µw)

α2 = µoPdelay/( µd - µo)

Pdelay = PcrossPd, the proportion of pedestrians leaving the cell by crossing the road

Pd = the probability that a pedestrian would be delayed when he crosses the road

γ = λH(1-P'f) / [λH(1-P'f) + (λo(1-Bo)]

ASYNCHRONOUS TRANSFER MODE (ATM)

The bandwidth requirements for data traffic within commercial organizations have been increasing steadily

for some time, both in the local area networks and in wide area networks. Workstations have been used to

introduce multimedia applications to the desktop, including components of voice, video and image, besides

growing amount of data. This development requires networks of greater bandwidth than commonly present

today with the capability of handling multiservice traffic on the same network.

The Asynchronous Transfer Mode (ATM) is being developed as a high speed networking technique for

public networks capable of supporting many classes of traffic. Asynchronous Transfer Mode (ATM) has

been accepted universally as the transfer mode of choice for Broadband Integrated Services Digital

Networks (B-ISDN) [10].

ATM is a high-speed, packet switching technique that uses short fixed length packets called cells. Fixed

length cells simplify the design of an ATM switch at the high switching speeds involved. The selection of a

short fixed length cell reduces the delay and most significantly the jitter (variance of delay) for delay-

sensitive services such as voice and video. ATM is capable of supporting a wide range of traffic types such

as voice, video, image and various data traffic.

Quality of Service (QoS)

ATM networks are thought to transmit data with varying characteristics. Different applications need

various qualities of service. Some applications like telephony may be very sensitive to delay, but rather

insensitive to loss, whereas others like compressed video are quite sensitive to loss.

The ATM Forum specified several QoS categories:

CBR (Constant Bit Rate)

VBR (Variable Bit Rate)

ABR (Available Bit Rate)

UBR (Unspecified Bit Rate)

Constant Bit Rate (CBR)

During a connection setup CBR reserves a constant amount of bandwidth. This service is conceived to

support applications such as voice, video and circuit emulation, which requires small delay variations


ISSN : 2250:3439



6

(jitter). The source is allowed to send as the negotiated rate any time and for any duration. It may

temporarily send at a lower rate as well.

Variable Bit Rate (VBR)

VBR negotiates the Peak Cell Rate (PCR), the Sustainable Cell Rate (SCR) and the Maximum Burst Size

(MBS). VBR sources are burst. Typical VBR sources are compressed voice and video. These applications

require small delay variations (jitter).

Available Bit Rate (ABR) and Unspecified Bit Rate (UBR)

ABR and UBR services should efficiently use the remaining bandwidth, which is dynamically changing in

time because of VBR service. Both are supposed to transfer data without tight constraints on end-to-end

delay and delay variation. Typical applications are computer communications, such as file transfers and e-

mail.

UBR service provides no feedback mechanism. If the network is congested, UBR cells may be lost.

An ABR source gets feedback from the network. The network provides information about the available

bandwidth and the state of congestion. The source‘s transmission rate is adjusted in function of this

feedback information. This more efficient use of bandwidth alleviates congestion and cell loss. For ABR

service, a guaranteed minimum bandwidth (MCR) is negotiated during the connection setup negotiations.

Wireless ATM

In recent years there has been an increasing trend towards personal computers and workstations becoming

―portable‖ and ―mobile‖. These ever-increasing groups of mobile users have been demanding access to

network services similar to their ―tethered‖ counterparts. The desire to provide universal connectivity for

these portable and mobile computers and communication devices is fueling a growing interest in wireless

packet networks. At the same time, wire line communication networks have been undergoing a

revolutionary change themselves with the introduction of Asynchronous Transfer Mode (ATM) based

Broadband Integrated Services Digital Network (B-ISDN) which can provide QoS guarantee [18]. Given

these rapid advancements, the communication networks of today are employing wireless media in the local

area and utilizing wire line physical media in the metropolitan and wide area environment.

To support multimedia applications in wireless systems, it is necessary to construct a wireless networking

infrastructure that can support QoS guarantees essential to provide broadband services. Since ATM is the

standard for wire line broadband networks, it has generally been agreed that broadband services are best

provided to wireless users by exploiting ATM in wireless systems. However, since the characteristics of the

wireless communication channels (e.g., high bit error rate and user mobility) are significantly different

from those of wire line channels, solutions that are designed for wire line networks cannot be expected to

work for wireless environments.

PERSONAL COMMUNICATION NETWORK (PCN)

Personal Communication Network (PCN) is an emerging wireless network that promises many new

services. With the availability of the interface cards, mobile users are no longer required to be confined

within a static network premise to get network access. Mobile users may move from one place to another

and yet maintain transparent network access through wireless links. Information exchanged between users,

may be bi-directional, which includes but not limited to voice, data and image, irrespective of location and

time while permitting users to be mobile.

In a PCN, the covered geographical area is typically partitioned into a set of microcells. Each microcell has

a base station to exchange radio signals with wireless mobile terminals. Due to the limited range of wireless

transceivers, mobile users can communicate only with the base stations that reside within the same

microcell at any instance. The number of handoffs during a call will increase as the cell radii decrease, thus


ISSN : 2250:3439



7

affecting the quality of service [11]. As a result of the increase in processing load due to demand for service

and fast handoffs to mitigate the propagation effect, a high speed backbone network for the PCN to connect

base stations is required. The ATM technology, which is been accepted as a predominant switching

technology, is suited to be an infrastructure to interconnect the base stations of the PCN.

Thus the given geographical area is partitioned into a set of disjoint clusters, each of which consists of a set

of microcells. Each microcell has a base station to serve the mobile terminals within the cell. An ATM

switch is allocated within each cluster and each of the base stations in this cluster is connected to one of the

ports of this ATM switch. The ATM switch offers the services of establishing / releasing channels for the

mobile terminals in the cluster. Two neighboring clusters can be interconnected via the associated ATM

switches. The links between the ATM switches are called Backbone Links, and the links between an ATM

switch and base stations are called Local Links. Each base station has a given number of radio channels for

calls generated within its cell [19].

Mobile hosts engaging in a call or data transfer within the same cluster will consume two local channels,

one for each local link, between the base stations and the associated switch and one radio channel [19]. For

intercluster communication, backbone links will be allocated in addition to the local links and radio

channel, and the channels occupied will depend on the communicating path being assigned.

SIMULATION OF ALGORITHM IN ATM BASED WIRELESS PCN NETWORK

The author analyses the ATM based wireless PCN network based on the schemes discussed in section III

and simulates the same using a non-preemptive Channel allocation algorithm called Measurement Based

Prioritization Scheme (MBPS) [12].

Performance metrics for Handovers [20]:

Call blocking probability – The probability that a new call attempt is blocked.

Handoff blocking probability – The probability that a handoff attempt is blocked.

Handoff probability – The probability that while communicating with a particular cell, an ongoing call

requires a handoff before the call terminates. This metric translates into the average number of handoffs

per cell.

Call dropping probability – The probability that a call terminates due to handoff failure. This metric can

be derived directly from the handoff blocking probability and the handoff probability.

Rate of handoff – The number of handoff per unit time.

Duration of interruption – The length of time during handoff for which the mobile terminal is in

communication with neither base station.

Measurement Based Prioritization scheme without Guard channels

In MBPS, if all channels of a cell are occupied, calls originating within that cell are simply blocked and the

handover requests to that cell are queued as per their priority [13][14][15][16][21]. MBPS is a non-

preemptive dynamic priority discipline. The handover area can be viewed as regions marked by different

ranges of values of power ratio, corresponding to the priority levels such that the highest priority belongs to

the MS whose power level is closest to the receiver threshold. On the other end, MS that has just issued a

handover request has the least priority. Obviously the last comes joins the end of the queue. A queued MS

gains higher priority as its power ratio decreases from the handover threshold to the receiver threshold. The

MS‘s waiting for channels in the handover queue are sorted continuously according to their priorities.

When a channel is released, it is granted to the MS with the highest priority. MBPS is designed as a

handover protection method, which a cellular communication network can utilize along with any channel

allocation strategy. The queuing can be performed at the BS or the Mobile Switching Center (MSC)

depending on the intelligence distribution between these cellular network components [13]. The basic idea

is that originating calls are not assigned channels until the queued handover requests are served. (Refer

Fig.2) In MBPS, new calls are served only when a channel is available and no handover request exists in


ISSN : 2250:3439



8

the queue. Probability of forced termination, Pf, is the probability that an originating call is eventually not

completed because of an unsuccessful handover attempt. Pf, therefore gives the percentage of handover

requests that are not served because the power received by the MS from the current BS approaches the

receiver threshold before a channel is granted. The objective is to minimize the time spent by MS in higher

priority, corresponding to the poorer signal reception, by favoring those MS‘s that received the lowest

power level from their current BS in channel assignment.

Measurement Based Prioritization scheme with Guard Channels

The only difference here should be noted that the employment of guard channels has the effect of reducing

the number of handover requests to be queued. However if started with congested cell whose channels are

already occupied, the number of guard channels, if any, will not have an impact on the waiting time

distributions of the arriving handover requests[13] [14][15][16][21].

Fig. 2. Call Flow Diagram for the Measurement Based Prioritization Scheme [21].

Simulation Parameters

busy_channels : Number of channels occupied by calls.

next_event_type : Type of next event New call, New handover, Channel release.

total_calls : Number of calls generated in or handed to the cell.

new_success : New calls which have been assigned channels by the BS.


ISSN : 2250:3439



9

ho_success : Handovers which have been assigned channels by the BS

ho_fail : Handovers which have not been assigned channels.

blocked : New calls which have not been allocated channels.

incell_success : New incell call or handover which have been assigned channels.

incell_blocked : New incell call or handover which have not been assigned channels.

incluster_success : New incluster call or handover which have been assigned channels.

incluster_blocked : New incluster call or handover which have not been assigned channels.

outcluster_success: New outcluster call or handover which have been assigned channels

outcluster_blocked: New outcluster call or handover which have not been assigned channels.

call_type : Type of call; Incell, Incluster, Outcluster

BTS_index : Index of Base Station whose event will occur.

Capacity : Load which a backbone link can handle.

next_call : Time at which next new call will be generated.

next_event_time : Time at which next event will occur.

next_handover : Time at which next new handover will be generated.

ho_delay : Time for which a handover is stored in the handover queue.

niat : Mean inter arrival time. Time difference between successive calls

hmiat :Handoff Mean inter arrival time. Time difference between successive handover.

RESULTS

The results for the three switch ATM network and the six switch ATM network have been computed. The

result comprises of comparison between the Measurement Based Prioritization Scheme (MBPS) and

MBPS using Guard Channels (GMBPS) for call blocking, handoff failures and throughput. Initial starting

parameters are listed below.

Number of Radio Channels = 30 per cell.

Number of Local Link Channels = 30 per cell.

Average time for a new call = 60 sec.

Average time for a handover call = 30 sec.

Maximum handover queue time = 10 sec.

Capacity of Backbone Links = 50 calls.

Number of Reserved Radio channels for GMBPS = 5

Number of Reserved Backbone Links for GMBPS = 5

Formulas Involved

Call Blocking = Total number of calls blocked / Total number of calls processed

Handoff failures = Total number of handovers not assigned channels / Total number of calls processed.

Throughput = (TSC + TSH) / Total number of calls processed


ISSN : 2250:3439



10

Where,

TSC = Total number of calls that have been assigned channels and backbone links.

TSH=Total number of handovers that have been assigned channels and backbone links.

Results of the Simulation

Fig. 3. (a) Graph showing the comparison of Guarded Scheme and Non Guarded Scheme for Call Blocking

in the three switch ATM network.


ISSN : 2250:3439



11

Fig. 3.(b) Graph showing the comparison of Guarded and non-Guarded schemes for Handoff Failures in the three

switch ATM network

Fig. 3.(c ) Graph

showing the comparison of Guarded and Non Guarded schemes for Throughput in the three switch ATM network

Discussion

1. From fig 3. (a) – (c) it can be seen that with the increase in load offered to the network the call

blocking and handoff failure increase exponentially and hence the throughput decreases for any

scheme used.

2. Since reducing handoff failure is the main concern for any cellular network, many schemes have

been proposed which reduce handoffs but at the cost of increasing call blocking.

3. As can be seen from fig 3. (a), call blocking is high in Guarded Measurement Based Prioritization

Scheme (GMBPS).

4. From 3. (b), it can clearly been seen that the probability of handoff failure is greatly reduced for

the GMBPS scheme when compared to the MBPS. The improvement without using guard

channels is between 30 – 40% and with guard channel is about 5%.

5. The use of guard channels greatly improves handoff failures by almost 75% in the case of MBPS

scheme, which is at the cost of 30% increase in call blocking. Hence overall throughput of the

network improves greatly when guard channels are used.

6. As can be seen from fig 3. (c) The throughput of GMBPS is better than MBPS.

CONCLUSION

This research focuses on the problem of handoffs in a mobile cellular environment. After studying the

currently used schemes, it is clear that there is some room for improvement. Using simulation of two types

of networks it is shown that an allocation of separate channel for handover requests (Guard Channel) shows

considerable improvement. The three commonly used performance metrics for a cellular network showed

improvement when the MBPS scheme is used. The use of guard channels does improve handoff failures but

also causes call blocking to increase. Hence there is a trade-off.


ISSN : 2250:3439



12

Hence the author conclude that, as network topology becomes more and more complicated and the offered

load to the network increases it is very necessary to use a priority based handoff scheme such as MBPS

which would reduce handoff failures without increasing call blocking.

REFERENCES

(1) Vijay K. Garg, Wireless Network Evolution 2G to 3G.

(2) J. E. Flood. Telecommunication Networks. Institution of Electrical Engineers, London, UK.

(3) G.Liodakis and P.Stavroulakis, ―A novell approach in handover initiation for microcellular system‖ Proc. 44th

IEEE VTC 1994.

(4) W.C.Y. Lee, Mobile Communication Design Fundamentals, Wiley, 2nd edition.,1993.

(5) G.Senarath and David Everitt, ―Performance of handover priority and queuing systems under different

handover request strategies for microcellular mobile communication sytems‖, Proc. 45th IEEE VTC, 1995.

(6) S.Shirvani Moghaddam, V.Tabataba vakili and A.Falahati, ―New Handoff Initiation Algorithm (optimum

combination of Hysteresis and Threshold based methods)‖, Vehicular Technology Conference (VTC 2000) ,

Vol.4, 2000, pp.1567-1574.

(7) Q-A. Zeng, K.Mukumoto and A.Fukuda, ―Performance analysis of a mobile cellular radio systems with

priority reservation handoff procedures‖, Proc. IEEE VTC‘94 vol.3, pp. 1829-1833, June 1994.

(8) S.Choi and K.Sohraby, ―Analysis of a mobile cellular systems with Hand-off priority and Hysteris control‖

IEEE INFOCOM 2000, Vol. 1, March 2000, pp 217-224.

(9) R.Steel and M.Nofal, Teletraffic performance of microcellular personal communication networks, IEEE

Proceedings-1, Vol.139, No.4 Aug 1992.

(10) ATM Forum, ―ATM User Network Interface (UNI) Signalling specification version 4.0‖, Doc. Af-sig-

0061.000, July, 1996.

(11) V.Kapoor, G.Edwards and R.Sankar, Handoff criteria for personal communication networks, Proc. IEEE

ICC‘94

(12) M.Gudmundson, Analysis of handover algorithms, Proc. IEEE VTC‘91. May 1991

(13) S.Tekinay and B.Jabbari, A measurement based prioritization scheme for handovers in mobile cellular

network, IEEE J. Select. Areas in commun, Vol.10, No.8, Oct 1992

(14) Sidi M, Starobinski D. New call blocking versus handoff blocking in cellular networks. Wireless Network

1997

(15) Howard G.Ebersman and Ozan K.Tonguz, ―Handoff Ordering Using Signal Prediction Priority queuing in

Personal Communication Systems‖

(16) Chor Ping Low, ―An Efficient Algorithm for Link Allocation Problem on ATM-Based Personal

Communication Network‖

(17) Handbook of wireless Networks and Mobile Computing, Edited by Ivan Stojmenovic. ISBN 0-471-41902-8

© 2002 John Wiley & Sons, Inc.

(18) S.K. Das et al., ―A call admission and control scheme for Quality of Service positioning in next generation

wireless networks‖, Wireless Networks 6., pp17-30, 2000.

(19) O.T.W.Yu and V.C.M Leung, ―Adaptive resource allocation for prioritized call admission over and ATM-

based wireless PCN‖, IEEE J.Select Areas commun., vol.15, pp1208-1225, sep.1997

(20) T.Salih and K.Fidenboylu, ―A comparison of the performance of two – tier cellular networks based on

queuing handoff calls‖, International Journal of Signal processing, vol.1, 2004, pp. 343 – 347

(21) Alagu.S and Meyyappan.T, ―Analysis of Algorithms for handling Handoffs in wireless mobile networks‖,

International Journal of P2P Networks Trends and Technology, Volume1 Issue2 - 2011


ISSN : 2250:3439



13

Preserving Privacy in Data Mining using Data Distortion Approach

Mrs. Prachi Karandikar #,

Prof. Sachin Deshpande *

# M.E. Comp,VIT , Wadala, University of Mumbai

* VIT Wadala ,University of Mumbai

1. [email protected]

[email protected]

Abstract. Data mining, the extraction of hidden predictive information from large databases, is nothing but

discovering hidden value in the data warehouse. Because of the increasing ability to trace and collect large

amount of personal information, privacy preserving in data mining applications has become an important

concern. Data distortion is one of the well known techniques for privacy preserving data mining. The objective

of these data perturbation techniques is to distort the individual data values while preserving the underlying

statistical distribution properties. These techniques are usually assessed in terms of both their privacy

parameters as well as its associated utility measure. In this paper, we are studying the use of non-negative

matrix factorization (NMF) with sparseness constraints for data distortion.

Keywords: Data Mining, Privacy, Data distortion, NMF, Sparseness

INTRODUCTION

Data Mining [1] ,is the extraction of hidden predictive information from large databases, is a powerful

technology with great potential to help companies focus on the most important information in their data

warehouses. Several data mining applications deal with privacy-sensitive data such as financial

transactions, and health care records. Because of the increasing ability to trace and collect large amount of

personal data, privacy preserving in data mining applications has become an important concern. There is a

growing concern among citizens in protecting their privacy . Data is stored either in a centralized database

or in a distributed database . According to its storage there are various privacy preserving techniques used .

Among the techniques that are used for privacy preserving data mining are: Generalization , Data

Sanitation , Data distortion , Blocking & Cryptography techniques . We are focusing our study on the latter

approach i.e. Data distortion via data perturbation . The objective of data perturbation is to distort the

individual data values while preserving the underlying statistical distribution properties. Theses data

perturbation techniques are usually assessed in terms of both their privacy parameters as well as its

associated utility measure. While the privacy parameters present the ability of these techniques to hide the

original data values, the data utility measures assess whether the dataset keeps the performance of data

mining techniques after the data distortion. Our objective is to study the use of truncated non-negative

matrix factorization (NMF) with sparseness constraints for data perturbation. The rest of the paper is

organized as follows. In section 2, we review the non-negative matrix factorization technique. The data

distortion and the utility measures which can be used are reviewed in section 3 and section 4 respectively.

Experimental results are given in section 5. Conclusion and future scope are given in section 6.

Nonnegative matrix factorization

Non negative matrix factorization (NMF) [8] refers to a class of algorithms that can be formulated as

follows: Given a nonnegative n ×r data matrix, V . NMF finds an approximate factorization V ≈WH where

W and H are both non negative matrices of size n×m and m×r respectively. The reduced rank m of the

factorization is generally chosen so that n r<m<nr and hence the product WH can be regarded as a

compressed form of the data matrix V . The optimal choices of matrices W and H are defined to be those

non-negative matrices that minimize the reconstruction error between V and WH . Various error functions


ISSN : 2250:3439



14

have been proposed. The most widely used is the squared error (Euclidean distance) function (i) and K-L

Divergence function (ii) etc [2] .

(1)

(2)

Non negative matrix factorization requires all entries of both matrices to be non negative, i.e., the data is

described by using additive components only.

NMF with Sparseness Constraint

Several measures for sparseness have been proposed. The sparseness of a vector X of dimension n

is given by [8]:

(1)

Usually, most of NMF algorithms produce a sparse representation of the data. Such a representation

encodes much of the data using few active components. However, the sparseness given by these techniques

can be considered as a side-effect rather than a controlled parameter, i.e., one cannot in any way control the

degree to which the representation is sparse. Our aim is to constrain NMF to find solutions with desired

degrees of sparseness. The sparseness constraint can be imposed on either W or H or on both of them. For

example, a doctor analyzing a dataset that describes disease patterns, might assume that most diseases are

rare (hence sparse) but that each disease can cause a large number of symptoms. Assuming that symptoms

make up the rows of her matrix and the columns denote different individuals, in this case it is the

coefficients which should be sparse and the basis vectors unconstrained. We have studied the projected

gradient descent algorithm for NMF with sparseness constraints proposed in [7] .

Truncation on NMF with Sparseness Constraint

In order to control the degree of achievable data distortion, the elements in the sparsified H matrix with

values less than a specified truncation threshold are truncated to zero.

Thus the overall data distortion can be summarized as follows:

(i) Perform sparsified NMF with sparse constraint h s on H to obtain HSh (ii) Truncate the elements in HSh

that are less than to obtain HSh ,. The perturbed dataset is given by W HSh ,Є. Thus the new dataset is

basically distorted twice by our proposed algorithm that has three parameters: the reduced rank m , the

sparseness parameter h s and the truncation threshold Є.

W, H∑ Vij WH

ij

2 .

i,j

Sx = √n –( ∑|x1|) /√∑x1 2

√n -1


ISSN : 2250:3439



15

DATA DISTORTION MEASURES

Throughout this work, we adopt the same set of privacy parameters proposed in [5]. The value difference

(VD) parameter is used as a measure for value difference after the data distortion algorithm is applied to the

original data matrix. Let V and V denote the original and distorted data matrices respectively. Then, VD is

given by :

VD || V V || / ||V || , (1) where || || denotes the Frobenius norm of the enclosed argument.

After a data distortion, the order of the value of the data elements also changes. Several metrics are used to

measure the position difference of the data elements. For a dataset Vwith n data object and m attributes, let i

, Rank j denote the rank (in ascending order) of the jth

element in attribute i.

Similarly, let Ranki j denote the rank of the corresponding distorted element. The RP parameter is used to

measure the position difference. It indicates the average change of rank for all attributes after distortion and

is given by

(2) (2)

RK represents the percentage of elements that keeps their rank in each column after distortion and is given

by

(3)

where Rk

ij 1. If an element keeps its position in the order of values, otherwise Rk

ij 0 .

Similarly, the CP parameter is used to measure how the rank of the average value of each attributes varies

after the data distortion. In particular, CP defines the change of rank of the average value of the attributes

and is given by

(4)

m n

RP = 1 ∑ ∑ Ranki j - Rank

i j

i=1 j=1

nm

1 m n

RK= ∑ ∑ Rki j

nm i=1 j=1

m

CP = 1 ∑ RankVVi - RankVVi

m i=1


ISSN : 2250:3439



16

where RankVVi and RankVVi denote the rank of the average value of the i

th attribute before and after the

data distortion, respectively. Similar to RK, CK is used to measure the percentage of the attributes that keep

their ranks of average value after distortion. From the data privacy perspective, a good data distortion

algorithm should result in a high values for the RP and CP parameters and low values for the RK and CK

parameters.

UTILITY MEASURE

The data utility measures assess whether the dataset keeps the performance of data mining techniques

after the data distortion. The accuracy of a simple K-nearest neighborhood (KNN) [9] can be used as data

utility measure.

EXPERIMENTAL RESULTS

In order to test the performance of our proposed method, we conducted a series of experiments on some

real world datasets. In this section, we present a sample of the results obtained when applying our technique

to the original Wisconsin breast cancer downloaded from UCI machine Learning Repository [10].

For the breast cancer database, we used 569 observations and 30 attributes (with positive values) to perform

our experiment. For the accuracy calculation , TANAGRA data mining tool has been used .Throughout the

experiment KNN classifier has been used with K=19.

Using Squared error function –

From the Figure 1, it is clear that m = 2 provides the best choice with respect to the privacy parameters.

-.- RK

- o RP X CP

+ CK

VD


ISSN : 2250:3439



17

Fig. 4. Effect of the reduced rank m on the privacy parameters.

Table 1 shows the how the privacy parameters and accuracy vary with the sparseness constraint Sh .

Sh

RP

RK

CP

CK

VD

Acc

0

180.28

0.0371

0.266

0.866

0.0431

93.14 %

0.45

185.90

0.0153

0.266

0.866

0.3751

92.23 %

0.65

187.41

0.0115

0.6

0.6

0.98

91.21 %

0.75

186.43

0.015

0.26

0.86

0.99

92.61 %

TABLE I EFFECT OF THE SPARSENESS CONSTRAINT ON THE PRIVACY PARAMETERS AND

ACCURACY

Table 2 shows the effect of threshold on the privacy parameters. From the table, it is clear that there is a

trade-off between the privacy parameters .

RP

RK

CP

CK

VD

0.035

184.90

0.0179

0.2667

0.8667

0.1398

0.040

195.54

0.0173

0.2667

0.8667

0.1044

0.045

196.11

0.0166

0.2667

0.8667

0.0952

0.050

197.71

0.0144

0.2667

0.8667

0.0830


ISSN : 2250:3439



18

TABLE II EFFECT OF THRESHOLD PRIVACY PARAMETERS AND ACCURACY

Using K-L Divergence error function –

From the Figure 2, it is clear that m 3 provides the best choice with respect to the privacy parameters.

So, we fixed m 2 throughout the rest of our experiments with this dataset.

Fig. 5. Effect of the reduced rank m on the privacy parameters

Table 3 shows the how the privacy parameters and accuracy vary with the sparseness constraint Sh .

0 5 10 15 20 250

1

2

3

4

5

6

7

8

9

10

Reduce rank m

Privacy p

ara

mete

rs &

Accura

cy

RP/100

RK*100

CP

CK*10

VD

ACC/10

Sh

RP

RK

CP

CK

VD

Acc

0

187.51

0.0113

0

1

0.0536

92.14%

0.3

187.28

0.0103

3.2

0.53

0.999

90.56 %

0.65

185.93

0.0108

6.06

0.366

0.999

90.86 %


ISSN : 2250:3439



19

TABLE III EFFECT OF THE SPARSENESS CONSTRAINT ON THE PRIVACY PARAMETERS AND

ACCURACY

From the results in Table 3, it is clear that Sh = 0.65 not only improves the values of the privacy parameters,

but also improves the classification accuracy.

Table 4 shows the effect of threshold on the privacy parameters and accuracy. From the table, it is clear

that there is a trade-off between the privacy parameters and the accuracy.

RP

RK

CP

CK

VD

0.005

186.09

0.0108

6.066

0.3667

0.999

0.01

187.064

0.0108

6.066

0.3667

0.999

0.05

194.84

0.0083

6.066

0.3667

0.999

0.1

198.51

0.0064

6.066

0.3667

0.999

TABLE IV EFFECT OF THRESHOLD PRIVACY PARAMETERS

CONCLUSION AND FUTURE SCOPE

Non-negative matrix factorization with sparseness constraints can provide an effective data perturbation

tool for privacy preserving data mining. In order to test the performance of the proposed method we would

like to conduct a set of experiments on some standard real world data sets . While using the above

mentioned privacy parameters , we would like to test the ability of these techniques to hide the original data

values .

REFERENCES

[1] M. Chen, J. Han, and P. Yu, "Data Mining: AnOverview from a Database Prospective", IEEE Trans. Knowledge and

Data Engineering, 8, 1996.Z.

[2] Yang, S. Zhong, R. N. Wright, ―Privacy preserving classification of customer data without loss of accuracy,‖ In proceedings

of the 5th SIAM International Conference on Data Mining, Newport Beach, CA, April 21-23, 2005.

[3] Saif M. A. Kabir1, Amr M. Youssef2 and AhmedK.Elhakeem1 Concordia University, Montreal, Quebec, Canada , ― On data

distortion for privacy preserving data mining ―

[4] Rakesh Agrawal and Ramakrishnan Srikant, ―Privacy-preserving data mining,‖ In Proceeding of the ACM SIGMOD Conference

on Management of Data, pages 439–450, Dallas, Texas, May 2000. ACM Press.

[5] Shuting Xu, Jun Zhang, Dianwei Han, and Jie Wang, Data distortion for privacy protection in a terrorist Analysis system. P.

Kantor et al (Eds.):ISI 2005, LNCS 3495, pp.459-464, 2005

[6] V. P. Pauca, F. Shahnaz, M. Berry and R. Plemmons. Text Mining using non-negative Matrix Factorizations, Proc. SIAM Inter.

Conf. on Data Mining, Orlando, April, 2004.

[7] Patrik O. Hoyer. Non-negative Matrix Factorization with Sparseness Constraints. Journal of Machine Learning Research 5

(2004) 1457–1469


ISSN : 2250:3439



20

[8] D. D. Lee and H. S. Seung. Algorithms for nonnegative matrix factorization. In Advances in Neural Information Processing 13

(Proc. NIPS 2000). MIT Press, 2001.

[9] R.Duda,P.Hart,and D. Stork, ―Pattern Classification,‖ John Wiley and Sons, 2001.


ISSN : 2250:3439



21

SECURED TRANSFER OF MESSAGES AGAINST

MALICIOUS ATTACKS USING EFFICIENT ALGORITHM

F. EMILY MANOZ PRIYA Department of Computer Science and Engineering

SASTRA University

Thanjavur, India-613402

[email protected]

P.S. RAMESH Asstt. Prof., Department of Information Technology

SASTRA University


[email protected]

B. SANTHI Prof., Department of Information and Communication Technology

SASTRA University


[email protected]

Abstract- Wireless Sensor Network (WSN) is a new class of networking technology .When we use sensor

network in brutal environment, security is most important concern. The technology may face against various

attacks. These attacks produce vulnerability against authentication, confidentiality and trustworthiness. This

paper introduces an adaptive method for securing the transformation of messages in wireless sensor networks in

the harsh environment. The light weight protocols are highly suitable for achieving authentication. The efficient

matching algorithm will be used for performing packet matching and also it detects the malicious attack

efficiently within the transformation of data. Finally, the encryption/decryption algorithm secures our original

data.

Keywords: wireless sensor network, security, light weight protocol, attack.

I. INTRODUCTION

Wireless Sensor Network (WSN) are a fascinating and challenging area of research and a key enabler

for new applications involving smart objects interacting with the physical environment. A wireless sensor

network is a self-configuring network of small sensor nodes communicating among themselves using radio

signals and deployed in large quantity to sends, monitor and understand the physical world. It contains

nodes, where each node is connected to one sensor. The wireless sensor nodes are often called as Motes.

Each motes has several fields: a radio transceiver (a transmitter and a receiver) with an antenna, a micro

controller and a battery [1]. The main characteristics of WSN include low power consumption, ability to

withstand hostile environmental conditions, mobility and scalability of nodes etc. Sensor nodes are

specified in terms of their interface and components. A sensor network is composed of a gateway and a

base station [2].

Sensor messages flow from the wireless sensors to the gateway where it is processed on the base station.

Sensor controls the message flow from base station to gateway. Sensor nodes are normally low power and

low cost devices. Therefore, we have to allow the nodes to use the available power carefully. The general

architecture of the wireless sensor network [12] is shown in figure 1

.


ISSN : 2250:3439



22

FIGURE 1: General structure of Wireless sensor network

In this diagram they show how the information is transferred from the sensor network to the gateway and

then transferred to visualizer. For the message transaction within the wireless sensor network cryptography

plays a vital role. It provides authenticity, confidentiality, and replay protection of exchanged messages.

Authenticity is very important in wireless sensor network to get reliable information. Confidentiality is to

prevent the revelation of exchanged data for unauthorized parties. Replay protection prevents the attackers

to record messages for replay attack. This paper introduces an adaptive method for securing our original

data in the harsh environment. They achieve authentication, confidentiality and security. The paper is

organized into four sections as follows: Section two discusses about security analysis. In that we identified

various attacks and the requirements to overcome. In section three we describe the proposed adaptive

method. In section four, we described the implementation of this adaptive method and section five

concludes the paper with future direction.

II. SECURITY ANALYSIS

Generally, in real-time environment the network faces lot of attacks every day. The attacks are based on

authentication, confidentiality and security. Here we see the important requirements for security in wireless

sensor network [3, 4, 5]

Confidentiality

This can be achieved by encrypting the data. It means to assure that information contained in the data is

only disclosed to users for which the data was intended.

Replay protection

To assure that an attacker is not able to record the message and use it successfully. Replay protection is

achieved by adding unique information to each message. For example, add number of counters to the

message and increment it.

1 Authenticity

Authentication is an important for many applications in sensor network [4].These are the first step to

achieve the secured transformation of data. Adversary can easily inject the message, so a receiver of data

able to conform that the data originates from the claimed sender.

1.1 Light weight protocol

The light weight protocol is a communication protocol that is designed with less complexity in order to

reduce overhead in terms of more computations.

1.1.1 HB protocol

Over the past few years several protocols have been analyzed.

In 2001, Hopper and Blum are the two authors proposed a protocol called HB protocol [6].

The working of HB protocol as follows: The reader computes and sends the query to tag. Tag performs dot

operation with query and secret key then XOR the result with noise value and computes result r and sends it

to reader. Reader compares the values as r‘= r and if so accept the process otherwise reject it.

This works well against passive attack but an active attack breaks its secureness. Hence an efficient

protocol is needed to address this problem.

1.1.2 HB+ protocol

Juels and Weis (2005) modified HB against active attack from adversaries. The working of HB+

protocol is different from HB in two regards:

1. In HB+ the tag uses two secret keys instead of one.

2. The tag can produce a blinding vector and the remaining process is similar to HB.

This protocol is susceptible to active attack but it‘s not well secure against man- in-the-middle attack.

1.1.3 HB++ protocol

In 2006, Bringer modified HB+ protocol to secure against man-in-the-middle attack from adversaries in

Gilbert (2005). It is called as HB++protocol. The working of HB++ is as follows:


ISSN : 2250:3439



23

The tag can initiate the process .It has four secret keys, and computes the value of z and z‘ and send it to

reader. The reader recalculate the value of z and z‘ and check against the value which it was received. This

protocol is also not highly suitable for man-in-the-middle attack in the harsh environment. The above

algorithms much of the time fails in authentication. By considering that we are forced to go for a new

authentication scheme.

1.2 Matching algorithm

Packet matching can be performed using Matching algorithm. In previous papers the authors revealed

several methods like first match, Recursive Flow Classification (RFC), decision tree classifier[11] etc.

The main drawbacks in the previous approaches are time complexity, poor search time and space

complexity. Apart from these basic drawbacks the malicious attack is also to be vulnerable to our original

data.

The adaptive method deals with an efficient matching algorithm which will be used for detecting the

malicious attack and transfer the message securely.

1.3 Encryption/Decryption It‘s the process of transforming information (plain text) by using an algorithm and makes it unreadable

(cipher text) to unauthorized users. In decryption side, it is the reverse process of encryption to make the

encrypted information readable again. The conventional algorithm will be used for securing our original

message.

III. PROPOSED METHOD

The data flow of the proposed method is shown below.

FIGURE 2: Proposed work flow

3.1 Authentication protocol

Light

weight

protocol

Sweep Line

Algorithm

Binary

Search

Packet Matching

Encryption/

Decryption

Secured Data Tans

mission

Authentication

Wireless Node


ISSN : 2250:3439



24

Wireless sensor networks are performed in low power energy sources, and having limited buffers for

memory [8]. So we are using a Light weight protocol which provides more flexibility and high performance

rate. In this paper the light weight protocol provides an authentication for efficient transaction. A modified

HB++ protocol is an authentication protocol used it here which overcomes the man in middle attack and

eliminates the vulnerability of the existing protocols. To achieve this, the protocol uses mutual

authentication in both the sides.

3.2 Matching algorithm The matching algorithm will be developed for performing packet matching efficiently. By providing an

efficient rule sets the algorithm will detect the malicious attack easily and prevent the packet transaction.

By using this algorithm we will compute all intersecting pairs. They perform several operations[11]

,[7]like calculating the intersection points between two curves and list out the curves participating at each

intersection point, also find whether an intersection point between any two curves exist or not.

Basically every packet has fields like source address, destination address, source port, destination port,

PAN id etc.

Packets can be examined as per the prescribed rules by using Binary Search algorithm to detect the

malicious attacks and to avoid unauthorized users access. The data structure of the packets having the fields

to perform matching algorithm is shown below

FIGURE 3: Packets data structure

3.3 Encryption/decryption algorithm

The algorithm for encryption/decryption process utilizes the conventional encryption algorithm. Here we

considered Play fair cipher, because the algorithm has limited computing capability. But it produces cipher

text that is more secure.

IV. PROPOSED ALGORITHM

The Architecture of the proposed method is shown below.

FIGURE 4: Architecture of Adaptive method

Authenti

cation N

Packet

Matchin

g

Encryption

/Decryptio

n

N


ISSN : 2250:3439



25

Step 1: The data will be transferred from the wireless node.

Step 2: By using the Light weight protocol the authentication for the transferred data will be checked out.

Step 3: Packet matching can be performed using matching algorithm and detect the malicious attack.

Step 4: If the data is efficient against malicious attack it will then encrypted using conventional algorithm

Finally secured data will be reached to another node.

V.CONCLUSION

In this paper, we presented a new approach for secured transformation of messages against various attacks.

Our technique provides a modified version of authentication protocol has been developed for achieving

authentication. Packet matching using efficient matching algorithm is to detect the malicious attacks for

every transaction. The information is then subjected to encryption algorithm for physical transmission.

In future by implementing the proposed algorithm we have to perform the comparison of this algorithm

with the existing algorithms in terms of high performance, power consumption, time complexity and space

complexity.

REFERENCES

[1] Pawan Kumar Goel, Vinit Kumar Sharma ―Wireless sensor network: a security model‖, IJSTM Vol. 2,

Issue 2, April 2011

[2] Brian Carter and Ram Mohan Ragade ―Message transformation services for wireless sensor network‖

(MTS-WSN)computer engineering and computer science, University of Louisville.

[3] Mohammed Mana, Mohammed Feham and Boucif Amar Ben saber ―A Light weight protocol to

provide location privacy in wireless body area networks‖ IJNSA,vol.3,No.2,March 2011.

[4] Xiuli Ren and Haibin Yu‖ Security mechanisms for wireless sensor networks‖, IJCSNS, Vol.6, No.3,

March 2006.

[5] Mayank Saraogi ―Security for wireless sensor networks ―Department of Computer science, University

of Tennessee, Knoxville.

[6] Julien Bringer, Herv´e Chabanne and Emmanuelle Dottax‖ HB++: a Lightweight Authentication

Protocol Secure against Some Attacks‖ Sagem Defense Securite Avenue du Gros Chene 95610

Eragny sur Oise, France.

[7] www.cgal.org : 2Dsweep line of planar curves.

[8] Prashant Agarwal, Tan sun teck andAnandaA.L‖A light weight protocol for wireless sensor

networks‖, Center for internet research, school of computing. National university of

Singapore,Singapore.

[9] Xuefei Leng, Mayes and Markantonkis ―An improvement on HB-MP protocol‖, Dept of math‘s,

university of London, May 2008.

[10] Roi Saltzman Adi Sharabani‖ Active man in middle attacks‖ A security advisory, A white

paper from IBM Rtional application security group,febuary 2009.

[11] Dmitry Rovniagin and Avishai Wool,‖The Geometric Efficient Packet Matching for Firewalls‖,IEEE

transaction on dependable and secure computing,Vol.8,No.1,Jan-Feb 2011.

[12] Carsten Buschmann, Dennis Pfisterer and

Stefan Fischer, ―Spyglass: A Wireless sensor

Networkvisualizer‖,Institute for Telematics,

University of Lubeck.

[13]http://openwsn.berkeley.edu

http://www.cgal.org/

http://openwsn.berkeley.edu/


ISSN : 2250:3439



26

Human head effects on the planar inverted-F antenna

performances

Saida Ibnyaich1, Raefat Jalila El Bakouchi2, Samira Chabaa

1, Abdelilah

Ghammaz2, Moha M‟rabet Hassani1,

1 Department of Physics,Electronics and Instrumentation Laboratory,

Faculty of Sciences Semlalia, Cadi Ayyad University,Marrakesh Morocco

2 Department of Physics, Laboratory of Electrical Systems and Telecommunications

Faculty of Sciences and Technology, Cadi Ayyad university,Marrakesh Morocco

{s.ibnyaich, s.chabaa,hassani}@ucam.ac.ma, [email protected], [email protected]

Abstract. With the current expansion and the anticipated further increase in the

use of cellular telephones and other wireless communication devices,

considerable research effort is devoted to investigations of interactions between

antennas on handsets and the human body. This interaction significantly

changes the antenna characteristics from that in free space or even on the device

(handset, laptop). In this paper and in order to study this problem, firstly a

planar inverted-F antenna (PIFA) was designed and simulated to operate over

the frequency 2,45 GHz , then the influence of the human head on the return

loss and on the radiation efficiency of the antenna has been studied.

Keywords: Planar inverted-F antenna; PIFA; Human head-Antenna

Interaction; SAR.

1 Introduction

It is well known that big efforts have been undertaken by researchers all over the

world to address the problem of optimizing performance of mobile communication

devices and increase radiation efficiency of antennas [1-3]. At the same time, it is

clear that radiation properties of the antenna in free space are different from those in

practical situations when it is located in locality of the user body or head due to

electromagnetic coupling.

These activities are motivated by two factors [4], the first factor is the need to

evaluate deterioration of the antenna performance and to develop better antennas, and

the second is a need to evaluate the rates of RF energy deposition, called specific



ISSN : 2250:3439



27

absorption rates (SAR), in order to evaluate potential health effects and compliance

with standards [5-7].

In this paper we are interested on the first factor; we study the influence of the head

on the return loss and radiation proprieties of a planar inverted-F antenna. Its

important evaluates deterioration of the antenna performance and develop better

antennas, which improve the radiation.

2 PIFA antenna configuration

2.1 Planar inverted-F antenna (PIFA)

PIFA is the abbreviation of Planar-Inverted-F-Antenna. The PIFA antenna has

advantages of having small and multiband resonant properties, simple design,

lightweight, low cost, conformal nature, attractive radiation pattern, and reliable

performance [8]. These characteristics make the PIFA a suitable antenna candidate to

mobile phones.

The inverted-F antenna is evolved from a quarter-wavelength monopole antenna. It is

basically a modification of the inverted F antenna IFA which is consisting of a short

vertical monopole wire.

To increase the bandwidth of the IFA a modification is made by replacing the wires

with a horizontal plate and a vertical short circuit plate to obtain a PIFA antenna.

The conventional PIFA is constituted by a top patch, a shorting plate and a feeding

plate. The top patch is mounted above the ground plane, which is connected also to

the shorting pin and feeding pin at proper positions. They have the same length as the

distance between the top patch and the ground plane. The standard design formula for

a PIFA is [9]:

f =

(1)

Where f is the resonant frequency of the main mode, C is the speed of light in the free

space; W and L are width and length of the radiation patch, respectively.

The Figure.1 shows the illustration of our developed PIFA antenna.

2.2 Antenna Configuration

The configuration of the studied PIFA antenna consists of a radiating top plate with

the dimensions WxL, and the ground plane dimensions are WgxLg . The dielectric

material used above the rectangular ground plane is FR-4 having a thickness t and a

relative permittivity εr, this is meant for the application when the antenna is integrated

with the printed circuit board (PCB). The antenna height is h, and the space between


ISSN : 2250:3439



28

the top plate and the substrate is filled with air (free space). The shorting plate has

dimensions of Ws x (h+t), and the feed plate has dimensions of Wf x h. The distance

between the shorting plate and the feeding plate is Fs(Figure 1).

Fig. 1. The Geometry of the studied PIFA antenna

Table 1. The overall dimensions of the studied PIFA antenna.

Parameter Designation Value

W Width of the radiating plate 40 mm

L Length of the radiating plate 21.5 mm

Wg Width of the ground plane 40 mm

Lg Length of the ground plane 55 mm

T Thickness of the ground plane 1 mm

H Antenna height 10.2 mm

Wf Width of the feeding plate 18 mm

Ws Width of the shorting plate 1 mm

Fs Distance between the shorting

plate and the feeding plate

15 mm

Feeding Plate

Ground Plane Radiating

Plate

Shorting

Wg

W

Wf

Lg

L Ws

h

F

s


ISSN : 2250:3439



29

3 Simulation Results and Discussion

3.1 The simulated results without human head model

The simulated return losses of the proposed PIFA antenna without the human head is

presented in Figure 2, we note that the maximum return loss is -39.43 dB at 2.35

GHz. The upper and lower band frequencies are 2.12 GHz and 2.62 GHz respectively.

Fig. 2. The simulated return loss for the proposed antenna without human head

Fig. 3. The radiation pattern for the proposed antenna without head


ISSN : 2250:3439



30

We have also simulated the far field radiation patterns of the studied PIFA antenna

before adding the human head model (Figure 3).

3.2 The simulated results with human head model

In this section, a model is built as shown in Figure 4, of the antenna next to the left

ear on SAM model; the shape of the head model is similar with real human head

shape. The head model consists of homogenous dielectric representing the human

tissue with relative permittivity εr = 41.5 and electric conductivity 0.97 S/m [10-11].

Fig. 4. The PIFA antenna next to the model of the human head

(a) Front view

(b) Left view

(a) (b)


ISSN : 2250:3439



31

Fig. 5. The simulated return loss for the proposed antenna with the human head model

The simulated return loss of the PIFA antenna with the human head model is shown

in Figure 5. We note that the maximum return loss now is -30.99 dB at 2.25 GHz. The

upper and lower band frequencies are 1.77 GHz and 2.75 GHz respectively. As

expected, the resonant frequency of the PIFA antenna is decreased by adding the

human head model.

Fig. 6.. The radiation pattern for the proposed antenna with the human head model


ISSN : 2250:3439



32

The 3D radiation pattern for the PIFA antenna with the human head model is shown

in Figure 6. As can be seen, the interaction with the human head results in noticeable

changes to the shape, polarization, and directivity of the pattern.

Table II provides a comparison between a PIFA antenna with and without a human

head model in terms of resonance frequency, directivity and the specific absorption

rate (SAR).

Table 2. Effect of human head on the PIFA antenna performances.

PIFA without human head model PIFA with human head model

Frequency 2.36 GHz 2.25 GHz

Directivity 4.84 dBi 4.13 dBi

SAR10g -- 2.78 W/Kg

SAR1g -- 4.32 W/Kg

Frequency 2.36 GHz 2.25 GHz

4 Conclusion

In this paper a planar inverted-F antenna has been designed and its performance and

interaction with the human head evaluated in terms of the resonance frequency,

radiation pattern and SAR in the head. It‟s possible to conclude that the radiation

patterns have a strong change with the human head. Moreover, the return loss is

affected to.

References

[1] T. Zervos, A. Alexandridis, V. Petrović, K.Dangakis, B. Kolundžija, D. Olcan, A. Đorđević,C. Soras, ―Accurate measurements and modelling of interaction between the human head and the mobile handset‖, Proc. 7th WSEAS Int. Multiconf. CSCC, 2003.

[2] A. Alexandridis, V. Petrović, K. Dangakis, B. Kolundžija, P. Kostarakis, M. Nikolić, T. Zervos, A. Đorđević, ―Accurate modelling and measurements of a mobile handset EM radiation‖,in Proc. 2nd Intern. Workshop on Biol. Effects of Electromagnetic Fields, pp.251-259,2002.

[3] M.A. Jensen and Y. Rahmat-Samii, ―EM interaction of handset antennas and a human in personal communications‖, Proc. IEEE, vol. 83, pp. 7-17, 1995.

[4] M. Okoniewski, M. A. Stuchly, ―A study of the handset antenna and human body interaction‖, IEEE Trans. Microwave. Theo. and Tech., vol. 44, no.10, pp. 1855 - 1864, Oct. 1996.

[5] COST244 WG3, ‖Proposal for numerical canonical models in mobile communications‖, Proc. of COST244, pp. 1 - 7, Rome, Nov. 1994.


ISSN : 2250:3439



33

[6] H. C. Taylor, J. A. Hnad, ―Solution of canonical problems using the finite-difference timedomain method‖, Proc. of COST244, pp. 87 - 89, Rome, Nov. 1994.

[7] A. And jar, J. Anguera, C. Puente, ―Ground Plane Boosters as a Compact Antenna Technology for Wireless Handheld Devices‖, Antennas and Propagation, IEEE Transactions ,Volume59,Issue:5,pp.1668-1677,ISSN: 0018-926X,May.2011.

[8] M.Jung, K.Yunghee, B. Lee, ―Dual frequency meandered PIFA for Bluetooth and WLAN applications‖ Antennas and Propagation Society International Symposium, IEEE Volume 2,pp.958 - 961, 22-27 June 2003 .

[9] G.R. Kadambi, T.S. Hebron, T.B. Meza, S. Yarasi, ―Applications of annular and L-shaped slot in PIFA design‖ Antennas and Propagation Society International Symposium, 2003. IEEE Volume 3, pp. 22-27 June 2003.

[10] B.B. Beard, et all,‖Comparisons of computed mobile phone induced SAR in the SAM phantom to that in anatomically correct models of the human head‖ Electromagnetic Compatibility, IEEE Transactions,Volume:48,Issue:2 , pp. 397 – 407, on2006.

[11] Ae.Kyoung Lee; Hyung-Do Choi; Jae-Ick Choi; ETRI, Daejeon ―Study on SARs in Head Models With Different Shapes by Age Using SAM Model for Mobile Phone Exposure at 835 MHz‖ Electromagnetic Compatibility, IEEE Transactions ,Volume:49, Issue:2,pp.302-312, on May2007.

http://ieeexplore.ieee.org/xpl/RecentIssue.jsp?punumber=8

http://ieeexplore.ieee.org/xpl/tocresult.jsp?isnumber=5762404

http://ieeexplore.ieee.org/search/searchresult.jsp?searchWithin=Authors:.QT.Beard,%20B.B..QT.&newsearch=partialPref







ISSN : 2250:3439



34

Study of Route Reconstruction Mechanism in DSDV

Based Routing Protocols

Sharma Shelja, Kumar Suresh and Rathy R. K.

Department of CSE, FET, MRIU, Faridabad, India

Email: [email protected], [email protected], [email protected]

Abstract: Ad hoc networks are infrastructure-less collection of mobile nodes,

characterized by wireless links, dynamic topology and ease of deployment. Proactive

routing protocols maintain the network topology information in the form of routing

Tables, by periodically exchanging the routing information. Mobility of nodes leads to

frequent link breaks, resulting in loss of communication and thus the Information in the

Table may become stale after some time. DSDV routing protocol follows proactive

approach for routing and uses stale routes in case of link break, which is the major cause

of its low performance as mobility increases. We have focused on two variants of DSDV

namely Eff-DSDV and I-DSDV, which deals with the broken link reconstruction and

discussed in these protocols, the process of route reconstruction due to broken links. To

analyze this route reconstruction mechanism, we have used a terrain of size 700m × 800

m with 8 nodes placed randomly. Analysis shows that both Eff-DSDV & I-DSDV,

perform better than DSDV in Packet Delivery Ratio and Packet Loss with slight increase

in Routing Overheads.

Keywords: Route Reconstruction, Packet Loss, Packet Delivery Fraction, DSDV

1 Introduction

Ad-hoc network is a collection of self organizing, autonomous mobile nodes. The

nodes in the network can move randomly and arbitrarily, to form a dynamic topology

in the absence of pre-established infrastructure or central coordinator. The network

utilizes multihop wireless links to forward the packets to the destination with the help

of intermediate nodes.

On the basis of routing information update mechanism there are three categories of

routing protocols for ad-hoc wireless networks [8], namely, proactive (Table driven),

reactive (On-demand) and hybrid routing protocols. Proactive routing protocols

constantly maintain the network topology information through periodic exchange of

routing information, namely DSDV [10], WRP [12], CGSR [2] etc. On the other

hand, reactive protocols create routes on-demand basis. Few of the on-demand

protocols are DSR [2], [4], AODV [11], TORA [9], ABR [13] etc. Hybrid Protocols

combines the features of both proactive & reactive protocols, example includes ZRP

[3], SLURP [1] etc.





ISSN : 2250:3439



35

Destination-Sequenced Distance Vector (DSDV) routing protocol [10] for mobile

ad-hoc networks (MANET), is a Table driven routing scheme, based on Bellman Ford

algorithm. The main contribution of the algorithm is to solve routing loop problem by

associating each node with a sequence number. It follows hop-by-hop routing, in

which the route to a destination is distributed in the next-hop of the nodes along the

route. Packets are transmitted between the nodes using route tables stored at each

node. Each node maintains routing table with entry for every other node in the

network. Route table entries are: Dest_addr, Dest_seq_no, next-hop, hop_count, and

install_time. Sequence numbers are even if a link is present; else, an odd number is

used to maintain the consistency of the route tables in a dynamically varying

topology, each node relies on periodic exchange of routing information. It uses stale

routes in case of link break due to mobility, which is the major cause of its low Packet

Delivery Ratio and higher Dropped Packet rate.

2 Recent Works

The limited resources in MANETs have made designing an efficient and reliable

Routing strategy a very challenging problem. An intelligent routing strategy is

required to efficiently use the limited resources while at the same time be adaptable to

the changing network conditions such as network size, traffic density, and network

partitioning. Several variants of DSDV have been proposed [5], [6], [7], [14] in order

to increase the performance of the DSDV protocol.

In Eff-DSDV Khan et. al. [5], presents broken link route reconstruction scheme to

overcome the problem of stale routes, and thereby improving the performance of

DSDV. In this protocol, when an immediate link from any node ‗S‘ to the destination

say ‗T‘ breaks, the node ‗S' suspends sending packets and creates a temporary link

through a neighbor which has a valid route to the destination node ‗T‘. The temporary

link is established by sending one-hop ROUTE-REQUEST and ROUTE-ACK

messages. They implemented the protocol on NCTUns Simulator and performance

comparison was made with DSDV protocol using performance metrics such as Packet

Delivery Ratio, End-to-End delay, Dropped Packets, Routing Overheads, and route

length by varying the number of nodes in the network and the mobility speed of the

ad-hoc nodes. They observed from the simulation results that the performance of Eff-

DSDV is better than DSDV with respect to all the said metrics except the Routing

Overheads. The Routing Overhead is bound to be higher for Eff-DSDV due to the

extra route requests and route reply messages which otherwise is not present in the

DSDV protocol.

Ting Liu et. al. [7], presented I-DSDV, to improve the Packet Delivery Ratio of

DSDV routing protocol in mobile ad-hoc networks with high mobility. They

proposed that, when a route becomes invalid due to link breakage, the node that

detects the link breakage tries to create a new loop-free route through message

exchange with its neighbors. Further when route reconstruction in one-hop area is not

accomplished; the area of message exchange for invalid route reconstruction is

enlarged gradually on-demand. On the basis of simulation results they observed that


ISSN : 2250:3439



36

I-DSDV, reduces the number of Dropped Data Packets with little increased

Overheads at higher rates of node mobility.

In OPR: DSDV based new proactive routing protocol [6], the authors made an

attempt to reduce the control traffic flowing in the network. Each OPR node

periodically determines its 1-hop neighbors and 2-hop neighbors using hello

messages. Any change in the topology only needs to be propagated to the neighbors,

thus reducing the amount of processing and storage required at each node.

S-DSDV [14], postulates that each node creates two-way hash chains in relation to

each node in the network, including itself. One is used for guarding against the

decreasing metric attack and the other for against increasing metric attack.

3 Reconstruction of Broken Link

Link break can be very frequent in MANETs due to mobility of nodes, which is the

major cause of lower Packet Delivery Ratio & higher Dropped Packet Rate of routing

protocols. In DSDV routing protocol, link break affects the entire network due to

periodic/event triggered route update at each change of neighborhood and broken link

is reconstructed through periodic updates, which is a time consuming process and

degrades the performance of the protocol.

In this paper we have made an attempt to analyze the mechanism of reconstruction

of broken links through periodic routing update as in DSDV protocol, through 1-hop

route-request & route-ack messages as in Eff-DSDV and through message exchange

with neighbors as in I-DSDV.

We have considered the Ad-hoc network scenario of 8 nodes which are placed

randomly, in a terrain of size 700m * 800m The initial node positions of the nodes are

taken as: (100,700), B: (250, 600), C: (400, 500), D: (600, 500), E: (300, 300), F:

(450,350), G: (550, 400) and H: (500, 600) as shown in Figure 1.To illustrate the

process, let node ‗A‘ be the source node and node ‗D‘ is the destination. Further

suppose during transmission, link from node ‗C‘ to destination node ‗D‘ breaks.

Fig. 1. An Ad-Hoc Network considered


ISSN : 2250:3439



37

Initially, route tables are developed for the entire network as in DSDV. For example

the routing tables for the nodes ‗B‘, ‗C‘ and ‗E‘ are shown in Table1, Table2, Table3

respectively.

Table 1. Route Table of Node ‗B‘

Destination Next Hop Hop-count Seq. No.

A A 1 100

B B 0 150

C C 1 206

D C 2 310

E C 2 412

F C 2 250

G C 3 500

H C 2 334

Table 2. Route Table of Node ‗C‘

Destination Next

Hop

Hop-

count

Seq.

No.

A B 2 100

B B 1 150

C C 0 206

D D 1 310

E E 1 412

F F 1 250

G F 2 500

H H 1 334

Table 3. Route Table of Node ‗E‘


A C 3 100

B C 2 150

C C 1 206

D C 2 310

E E 0 412

F F 1 250

G F 2 500

H C 2 334

3.1 Reconstruction of Broken Link (Through periodic routing update) Using

DSDV Protocol

a) As node ‗C‘ detects link break to destination ‗D‘ as shown in Figure 2, it updates its route table, increases sequence number of node ‗D‘ by 1 (only case when Seq. No.


ISSN : 2250:3439



38

is incremented by the node other than the destination node) and updates hop-count

for ‗D‘ to infinity as shown in Table 4.

Fig. 2. Link from Node ‗C‘ to ‗D‘ breaks

Table 4. Updated Route Table of Node ‗C‘

Destination Next

Hop

Hop-

count

Seq.

No.

A B 2 100

B B 1 150

C C 0 206

D D ∞ 311

E E 1 412

F F 1 250

G F 2 500

H H 1 334

b) Node ‗C‘ advertises its updated route table to local neighbors ‗B‘,‘E‘, ‗H‘, and ‗F‘, to inform about link break to destination ‗D‘. Meanwhile Node ‗H‘ and ‗F‘ also

receives periodic route update from their neighbors, containing valid route for

destination ‗D‘ as shown in Figure 3.


ISSN : 2250:3439



39

Fig. 3. Node ‗C‘ advertises link break

c) Neighbors ‗B‘ and ‗E‘ had route to ‗D‘ only through node ‗C‘ (which is the broadcasting node), so they update their route tables with broken link information

for destination ‗D‘ as shown in Table 5 and Table 6. On the other hand, node ‗H‘

and ‗F‘ compares the update received from the node ‗C‘ to the route information

available in their route table and found that node ‗C‘ has stale route information and

sequence number of the available route for destination ‗D‘ is higher than the route

update received from node ‗C‘. So, node ‗H‘ and ‗F‘ does not update their route

tables with broken link information, as they have fresh route available to destination

‗D‘.

Table 5. Updated Route Table of Node ‗B‘


A A 1 100

B B 0 150

C C 1 206

D C ∞ 311

E C 2 412

F C 2 250

G C 3 500

H C 2 334

Table 6. Updated Route Table of Node ‗E‘

Destination Next

Hop

Hop-

count

Seq. No.

A C 3 100

B C 2 150

C C 1 206

D C ∞ 311

E E 0 412

F F 1 250

G F 2 500

H C 2 334

d) Nodes ‗H‘ and ‗F‘ advertise their loop free route for destination ‗D‘ to their neighbors as shown in Figure 4.


ISSN : 2250:3439



40

Fig. 4. Nodes ‗H‘ and ‗F‘ advertises their route to Node ‗D‘

e) Node ‗C‘ receives one route from node ‗F‘ and another from node ‗H‘, to

destination ‗D‘. After comparing the two routes, node ‗C‘ found that both routes

have same sequence number but the route through node ‗H‘ has a better metric, so

node ‗C‘ update its route table with new route to destination ‗D‘ via ‗H‘ as shown

in Figure 5.

Fig. 5. Node ‗C‘ reconstructs its route to ‗D‘ via ‗H‘

3.2 Reconstruction of Broken Link (Through 1-hop Route-Request & Route-

Ack ) Using Eff-DSDV Protocol

a) When the Link from node ‗C‘ to node ‗D‘ breaks, node ‗C‘ suspends sending

packets as shown in Figure 2.

b) Node ‗C‘ broadcasts Route-Request message to its 1–hop neighbors ‗B‘, ‗H‘, ‗E‘

and ‗F‘. Route-Request packet includes the node ID and the destination as shown in

Figure 6.


ISSN : 2250:3439



41

Fig. 6. Node ‗C‘ Broadcasts 1-hop Route-Request to its Neighbors

c) Nodes ‗H‘ and ‗F‘ responds with Route-Ack packet along with the hop-count &

route update time (indicates the freshness of a route) as shown in Figure 7. While

nodes ‗B‘ and ‗E‘ do not respond as they have route to node ‗D‘ through ‗C‘, which

is the broadcasting node. Table 7 shows the route update at node ‗C‘.

Fig. 7. Neighbors ‗H‘ and ‗F‘ respond with Route-Ack Packet

Table 7. Route Update at Node ‗C‘

I-Hop-

Neighbor

Next

Hop

Hop-

count

Update

Time

F G 2 940

H D 1 1250

d) The node ‗C‘ chooses the best neighbor, based on the least number of hops to the

destination. If there is more than one node having the same number of hops, it

selects the node with the latest routing update time. The packets are then forwarded

using the latest route, found till, the routing table of node ‗C‘ is updated by the

conventional DSDV protocol. From the Table 7, it can be seen that the node ‗H‘


ISSN : 2250:3439



42

has the minimum hop-count and latest update time so node ‗C‘ resumes sending

packets to destination ‗D‘ through ‗H‘ as a next-hop as shown in Figure 5.

3.3 Reconstruction of Broken Link (Through message exchange with

neighbors) Using I-DSDV Protocol

a) Node ‗C‘, detects that the link to one of its neighboring node ‗D‘ breaks as shown in Figure 2, any route through that link is assigned with an invalid type as shown in

Table 8.

Table 8. Route Update at Node ‗C‘

Destination Type Metric Sequence no.

D Invalid 1 311

A Valid 2 100

B Valid 1 150

C Valid 0 206

E Valid 1 412

F Valid 1 250

G Valid 2 500

H Valid 1 334

b) Node ‗C‘ broadcast an invalid route update (as shown in Figure 8), which contains the former metric & sequence number of the invalid route as shown in Table 9.

Type of path and former metric are additionally used to identify the broken links

other than infinite metric & an odd sequence no.

Fig. 8. Node ‗C‘ broadcasts its invalid route update

Table 9. Invalid Route Update

Destination Type Former

Metric

Sequence

no.


ISSN : 2250:3439



43

D Invalid 1 311

c) Upon receiving an invalid route update from node ‗C‘, nodes ‗B‘, ‗H‘, ‗E‘ & ‗F‘

compares it to the existing route in their route table, if node has a valid route

through node ‗C‘ as the next hop to node ‗D‘, then it replaces its own route with

the invalid one. Otherwise if it has a valid route that has fresh sequence number

or the same sequence number but a hop-count is less than or equal to J + 1, then

it immediately broadcasts a route reconstruction message that contains its valid

route information. Here, J is the metric that the invalid route update contains.

Nodes ‗B‘ and ‗E‘ replaces their route to node ‗D‘ with the invalid one, as they

have a valid route through node ‗C‘ as its next-hop. So nodes ‗B‘ and ‗E‘ do not

respond as they have no other valid route to destination ‗D‘ and in turn broadcast

the invalid route information about ‗D‘. On the other hand neighbors ‗H‘ and ‗F‘

broadcast their new loop free route to destination ‗D‘, as shown in Figure 9.

Fig. 9. Node ‗H‘ & ‗F‘ broadcast their new route to node ‗D‘

d) Node ‗C‘ receives two new loop free routes for destination ‗D‘. One through ‗H‘ with metric 2 and another through ‗F‘, with metric 3 but with the same

destination sequence number. On the basis of better hop-count node ‗C‘,

reconstructs its route to destination ‗D‘ through node ‗H‘ as shown in Figure

10.


ISSN : 2250:3439



44

Fig. 10. Node ‗C‘ reconstructs its route to Node ‗D‘

e) Node ‗C‘, broadcast new route information for destination ‗D‘, to its local

neighbors and neighbors in turn broadcast received information to their

neighbors as shown in Figure 11.

Fig. 11. Node ‗C‘ broadcasts its new route to Node ‗D‘

f) Node ‗B‘ reconstructs its route to destination ‗D‘ through ‗C‘. Node ‗E‘ compares

the new route with the old one (through ‗F‘) and would find that metric is same for

the two routes but the sequence number of the route received from node ‗C‘ is

higher, so node ‗E‘ also reconstructs its route through ‗C‘ to destination ‗D‘ as

shown in Figure 12. Meanwhile, the node ‗F‘ receives periodic route update to

destination ‗D‘ from node ‗G‘, with same sequence number thus finds that metric

of old route is better than the new route update. Therefore, it doesn‘t update its

route to destination ‗D‘, through node ‗C‘.


ISSN : 2250:3439



45

Fig. 12. Neighbors reconstruct their route to node ‗D‘

4 Conclusion

DSDV is a single path routing protocol and no immediate route is available with the

node to transmit the remaining data packets as link breaks. New route is reconstructed

through periodic exchange of routing information, which is very time consuming

process and also consumes other Network resources. We have discussed the broken

link route reconstruction process in DSDV and its variants Eff-DSDV & I-DSDV. On

comparative study of Eff-DSDV and I-DSDV, it has been observed that in both these

variants, Packet Delivery Ratio is improved, Dropped Packed Rate is reduced,

Routing Overhead is slightly increased but End-to-End delay of I-DSDV is higher

than DSDV as shown in Table 10.

Table 10. Qualitative Analysis of Eff-DSDV and I-DSDV

Performance

Parameters

Eff-DSDV I-DSDV

Packet Delivery Ratio Higher than DSDV Higher than DSDV

Routing Overhead Slightly Higher than DSDV Slightly Higher than

DSDV

End-to-End Delay Less than DSDV Higher than DSDV

Packets Dropped Relatively Lower than

DSDV

Lower than DSDV

In future work we intend to implement multipath scheme in DSDV protocol and

carry out simulation based comparisons among variants of DSDV and other routing

protocols.

References


ISSN : 2250:3439



46

1. Abolhasan, M., Wysocki, A., Dutkiewicz E.: A review of routing protocols for mobile ad

hoc networks. www.elsevier.com/locate/adhoc,Ad Hoc Networks, pp.1--22 (2004)

2. Broch, J., Johnson, D. B., Maltz D. A.: The Dynamic Source Routing Protocol for Mobile

Ad Hoc Networks. IETF Internet draft, draft-ietfmanet-dsr-01.txt, (work in progress)

(1998)

3. Hass, Z. J., Pearlman, R.: Zone routing protocol for ad-hoc networks. Internet Draft, draft-

ietf-manet-zrp-02.txt, work in progress, (1999)

4. Johnson, D. B., Maltz, D. A.: Dynamic source routing in ad hoc wireless networks. In:

Mobile Computing, edited by Imielinski, T., Korth H., chapter 5, pp. 153--181 (1996)

5. Khan, K. Ur R., Reddy, K. A., Zaman, R. U., Reddy, A. V., Harsha T. S.: An efficient

DSDV routing protocol for MANET and its usefulness for providing Internet access to Ad

Hoc Nodes. In: Proceedings of IEEE TENCON, pp. 1--6 (2008)

6. Kumar, S., Rathy, R. K., Pandey, D.: OPR: DSDV Based New Proactive Routing Protocol

for Ad-Hoc Networks. In: Proceedings of IEEE IACSITSC, pp. 204--207 (2009)

7. Liu, T., Liu, K.: Improvements on DSDV in Mobile Ad Hoc Networks. In: International

Conference on Wireless Communications, Networking and Mobile Computing, pp. 1637--

1640 (2007)

8. Murthy, C. S. R., Manoj, B. S.: Ad Hoc Wireless Networks. Pearson Education, pp. 213--

226 (2005)

9. Park, V. D., Corson, M. S.: Temporally-Ordered Routing Algorithm (TORA).version 1:

Functional specification. Internet-Draft, draft-ietf-manet-tora-spec-00.txt (1997)

10. Perkins, C. E., Bhagwat, P.: Highly dynamic Destination- Sequenced Distance-Vector

routing (DSDV) for mobile computers. In: Proceedings of SIGCOMM‘ 94 Conference on

Communications Architectures, Protocols and Applications, pp.234--244 (1994)

11. Perkins, C. E., Royer, E. M.: Ad-hoc On-Demand Distance Vector Routing. In:

Proceedings of 2nd IEEE Workshop. Mobile Comp. Sys. and Apps., pp. 90--100 (1999)

12. Royer, E. M., Toh C. K.: A review of current routing protocols for ad hoc mobile wireless

networks. In: Proceedings of IEEE, vol. 6 , pp. 46--55 (1999)

13. Toh, C.K.: Long-lived ad hoc routing based on the concept of associativity. Internet draft,

IETF (1999)

14. Wang. J.W., Chen, H.C., Lin Y.P.: A Secure DSDV Routing Protocol for Ad Hoc Mobile

Networks. In: Fifth International Joint Conference on INC, IMS and IDC, pp. 2079--2084

(2009)

http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=4766808&queryText%3DAn+efficient+DSDV+ROUTING+PROTOCOL+FOR+manet%26openedRefinements%3D*%26searchField%3DSearch+All



http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=5169340&queryText%3DOPR%26openedRefinements%3D*%26searchField%3DSearch+All

http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=5169340&queryText%3DOPR%26openedRefinements%3D*%26searchField%3DSearch+All

http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5169290

http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=4340187&queryText%3DDSDV+routing+protocol+in+MANET%2CPublication+2005-2010%26openedRefinements%3D*%26pageNumber%3D2%26searchField%3DSearch+All


http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=760423&queryText%3DA+review+of+current+routing+protocols+for+ad+hoc+mobile+wireless+networks%26openedRefinements%3D*%26searchField%3DSearch+All

http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=760423&queryText%3DA+review+of+current+routing+protocols+for+ad+hoc+mobile+wireless+networks%26openedRefinements%3D*%26searchField%3DSearch+All


http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=5331842&queryText%3DDSDV+routing+protocol+in+MANET%2CPublication+2005-2010%26openedRefinements%3D*%26searchField%3DSearch+All

http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=5331842&queryText%3DDSDV+routing+protocol+in+MANET%2CPublication+2005-2010%26openedRefinements%3D*%26searchField%3DSearch+All



ISSN : 2250:3439



47

Synthesizable Implementation of Safety Fuzzy Logic Controller of 1oo2

architecture in FPGA

Mohammed Bsiss, Amami Benaissa,

Laboratory of computer science systems and telecommunications (LIST)

Faculty of Science and Technology of Tangier,

University Abdelmalek Essaadi Tangier Morocco

90 000 BP 416 Tangier Morocco [email protected], [email protected]

Abstract. Currently, the achievements of security systems is becoming more and more ground in different areas

not only through the development of new technologies of programmable circuits, with the ability to achieve very

complex systems in a single chip but thanks also a common and coherent organization of the different safety

standard.This paper describes the implementation for a safety fuzzy logic controller (SFLC) on the basis of Safety

Norm 61508. The SFLC is programmed with the hardware description language VHDL and implemented in

FPGA.

Keywords: Safety norm 61508, Safety fuzzy logic controller, one-out-of-two architecture, field programmable

gate array (FPGA), very high speed hardware description language (VHDL).

1 Introduction

Today the implementation of a safety fuzzy logic controller is necessary. This paper shows how an SFLC can be

realized in FPGA. The internal structure of FPGAs is composed of an array of configurable logic blocks (CLBs)

along with interconnection channels, blocks of SRAM and input/output blocks (IOBs). Certain FPGAs also contain

PLLs (Phase Locked Loop), DLLs (Delay Locked Loop), (DCM: DIGITAL Clock Manager) and also simple ALUs

(Arithmetic Logic Units).

However, this complex structure can contain a large number of failures such as stuck-at fault, bridging fault,

interconnect defect, CLB defect.

One strategy discussed in [1] is based on creating several application circuits and testing them with test vectors

developed specifically for each circuit.

The second strategy is based on testing the internal structure and reconfigurability of an FPGA and is called the

Multi-Configuration Strategy (MCS) [2].

The third strategy [3] is based on the concept of Built-In Self-Test (BIST).

Generally all strategies mentioned above are normally employed by the chip manufacturer and used by

unprogrammed FPGAs. The challenge is how can we detect this failure using programmed FPGAs? Or, in other

words, how can we be sure that the generated VHDL code for the simple fuzzy logic controller (FLC) architecture

[4], [5] is correctly operated on the device? Under these circumstances, a simple-structure for an FLC without a

safety function (redundancy) does not provide reliability and the safety is only partial.

Particularly with regard to safety-related systems, model structures are necessary in order to allow safe operation in

case of system failure.

This security must meet the requirements defined in Security Norm IEC61508 [6].

2 The security norm IEC61508

Safety Norm IEC61508 is a international norm for the functional safety of electrical/electronic/programmable

electronic safety-related systems and consists of seven parts [8] (general requirement, requirements for E/E/PES

safety-related systems, software requirements, definitions and abbreviations, examples of methods for the

determination of safety-integrity levels, guidelines on the application of Part 2 and Part 3 and an overview of

techniques and measures).



ISSN : 2250:3439



48

Safety Norm 61508 [8] defines the safety integrity levels (SIL) as an order of the magnitude levels of risk reduction.

There are four SILs. SIL4 has the highest level of safety integrity and SIL1 the lowest in terms of risk reduction. The

four levels are defined in Table 1 [6] for various fractions of failures leading to a safe state as follows:

Table 1. Type a safe failure fraction

Safe Failure fraction

Hardware fault tolerance (see note 1)

0

1 2

< 60 % SIL1

SIL2 SIL3

60 % - < 90 % SIL2

SIL3 SIL4

90 % -<99% SIL3

SIL4 SIL4

>99% SIL3

SIL4 SIL4

NOTE 1 A hardware fault tolerance of N means that N+1 faults could cause a loss of the safety function

3 Safety fuzzy logic controllers

The safety fuzzy logic controller (SFLC) consists of the fuzzification process, rule evaluation process and

defuzzification process. The details of this process can be found in the following publications [4], [5] and [7].

The difference to the simple controller FLC is that these processes are created redundantly. The used architecture of

SFLC is one-out-of-two (1oo2), which means that in these particular cases the system must operate at least so that

the security function can react when an error occurs, and thus bring the system to a condition of safety. This

structure allows the system to meet the requirements of Safety Norm IEC61508 [6].

Figure 1 shows a basic model for a safety fuzzy logic controller with redundancy architecture (1oo2).

Fig. 2. Safety fuzzy logic

controller of 1oo2 architecture

As shown in Figure 1 the safety fuzzy

logic controller consists of the

following components:

Two Fuzzy Logic

Controllers.

One Flash PROM to save

Test Data Pattern.

Register Data_Comp

Register Data_Test.

Register Result_Test.

14-bit register to save data

from analog-to-digital

converter LTC6912-1[9]

12-bit register to give data

to the digital-to-analog converter LTC2624 [10]

Circuit for

control the

motor speed

FPGA

Circuit for

temperature

measurement

Fuzzy logic

controller

FLC2

12Bit

DA_ValueSPI

14Bit

ADC_ValueData_Comp

Fuzzy logic

controller

FLC1

D/A

Data_Test

Result_Test

CLK_FLC1

CLK_FLC2

Interrupt

signal

A/D

Flash

PROM

SPI


ISSN : 2250:3439



49

The design of the fuzzy logic controller architecture is shown in figure 2.

Fig.2. Fuzzy logic controller

3.1 Fuzzification

There are two variables for controlling the temperature input:

The error of temperature Te

And the rate of change of temperature dTe/dt.

The fuzzy set of membership functions of error of temperature is partitioned into 4 zones as shown in figure 3

below.

Temperature

T[°c]

μ

0

3FFFcold cool mild hot

0

A1C

+10

BB2h

+20

D49h

+30

EDFh+40

1075h

+120

1D28h

Fig. 3. The membership functions of error of temperature Te.

The analog to digital converter [9] has a 14-bit resolution which means that the membership degree μ = 1 is equal to

3FFFh or 16383 in decimal.

The second input variable for SFLC is the rate of change of temperature dTe/dt.

The membership functions of the derivation dTe/dt consist of three linguistic terms (slow, moderate and fast). The

graphic representation of membership functions is shown in figure 4.

Rule base

12BitDA_Value

FuzzifierTe error

Inference

Mamdani

(Min-Max)

FuzzifierTe/dt

Fuzzification

R(1) : MAX(MIN(uTe(1),uTe/dt(3)),

MIN(uTe(1),uTe/dt(2)))

R(2) : MAX(MAX(MIN(uTe(2),uTe/dt(1)),

MIN(uTe(2),uTe/dt(3))),

MAX(MIN(uTe(3),uTe/dt(1)),

MIN(uTe(3),uTe/dt(2))))

R(3) : MAX(MIN(uTe(4),uTe/dt(1)),

MIN(uTe(4),uTe/dt(3)))

uTe(1)

uTe(2)

uTe(3)

uTe(4)

uTe/dt(1)

uTe/dt(2)

uTe/dt(3)

)(

)(*)(3

1

iR

iRiyi

defuzzification

FLC

14BitAD_Value


ISSN : 2250:3439



50

Temperature

T[°c]/dt

μ

0

3FFFslow moderate fast

-5°c

950

-3°c

9A20°c

A1C

+3°c

A96+5°c

AE7

Fig. 4. The membership functions of the rate of the temperature change.

Figure 5 illustrates how part of this process is implemented in VHDL:

Fig.5. The fuzzification process in VHDL

3.2 Rule inference engine

In this step, each input value is applied to its membership function in order to determine the value of the fuzzy input.

In this application there are two inputs, one with four membership functions and the other with three, which makes

seven degrees of membership functions to be calculated.

After that, the degrees of membership function are determined in a fuzzification step; the next step is to use

linguistic rules to decide what action should be taken in response to a set of data.

The Mandani min-max technique [7] is used to calculate the numerical results of linguistic rules based on the input

values of the system.

Before we commence calculating how many rules might be needed for the system, we must define the output

membership functions.

The output membership functions consist of 3 singletons. The linguistic terms of the singletons output (slow,

moderate and fast) are used.

The graphic representation of membership functions is shown in figure 6.


ISSN : 2250:3439



51

Courant

[mA]

μ

0

FFF

5

$369

11

$733

17

$B03

slow moderate fast

Fig.6. The membership functions of the output

The SFLC has two inputs, one with four linguistic terms and the other with three and an output with three linguistic

terms. This makes a total of 4*3*3 different rules that may be used to describe the strategy of total control.

3.3 Defuzzification

After the rules for each output have been established, the next step is to combine them into a single output value that

can be used to control the output.

The center of gravity [7] will be used to obtain the release of the final system. In this application, the defuzzification

takes the weighted average of all fuzzy outputs. Each fuzzy output is multiplied by the corresponding singleton, and

then the sum of these products is divided by the sum of all fuzzy outputs for the final result.

The following pseudo-code illustrates how this process is implemented in VHDL:

Fig.7. Defuzzification process in VHDL

3.4 System test

The SFLC is tested by using following register „Data_test‟, „Data_Comp‟ and „Result_Test‟.

The register „Data_Test‟ compares the data from the first fuzzy logic controller FLC1 with the second FLC2. The

comparison takes place during each clock cycle. The comparison of the faulty data is avoided by periodically testing

the system with different data patterns. These patterns are stored in external flash PROM see Figure 1 and are based

on measurements made on the system. Note that when a channel is tested, the functionality of the system is ensured

by the second channel.


ISSN : 2250:3439



52

Therefore the SFLC allows not only to detect all stuck-at faults,, but also places the system in a safe state. The safety

integrity levels (SIL) reached by this architecture is SIL1.

4 Functional and timing simulations

After converting the VHDL code into gate-level schematics, a test bench for the safety fuzzy logic controller was

designed using the Isim tools [11] from Xilinx. This allows the timing and the correct functionality of the system to

be verified.

A synthesizable simulation result (timing diagram) is shown in figure 8.

Fig.8. Waveform of functional simulation of the safety fuzzy logic controller

The waveform in figure 8 shows the values of the temperature inputs and the corresponding output current in hex

form at the various instances determined by the stimuli in the test bench.

The parallelism in the FPGAs allows the fuzzification process and the defuzzification process of both fuzzy logic

controllers to be carried out during one time period T = 0.9us. The result of the first FLC1 and the second FLC2 are

stored within T = 1us in the registers flcdataout1 and flcdataout2.

As seen in figure 8 the result of the first FLC is exactly the same as the result of the second. If the interrupt signal

(sig_irg) switches to high, it means there is a discrepancy in the results and the system goes into the safe state. When

the system goes into the safe state, it means that the output is switched off.

5 Synthesis of the safety controller ‘SFLC’

The synthesis processes convert the VHDL code into gate-level schematics. The synthesis process is carried out

using Xilinx tools ISE Design Suite 12.2 [13]. The VHDL code of the safety fuzzy logic controller is synthesized for

converting into RTL view.

Part of synthesized netlist design of the safety fuzzy logic controller is shown in figure 9.


ISSN : 2250:3439



53

Fig. 9. Part of netlist design of SFLC

The generated synthesized netlist of the SFLC has been downloaded into the Sparan-3AN FPGA Starter Board Kit

[12] for verification of the correctness of the algorithm functionality.

6 Conclusions

A novel approach to realize a safety fuzzy logic controller with a field-programmable gate array has been proposed

in this paper. The design has been realized in VHDL using Xilinx12.3 [13.]

The SFLC operates at a frequency of 50 MHz and is not only able to react very quickly to the output whenever an

error occurs, it can also put the systems into a safe state.

The simulation and the implementation of safety fuzzy logic controller with ISim 12.3 from Xilinx demonstrated

complete functionality while meeting all the initial system requirements.

References

1. C.Jordan, W.P Marnane, 1993 . ”Incoming inspection of FPGAs”. In Proc. European Test Conf pp371-377.

2. M.B Tahoori, E J. Mc.Cluskey and M.Renovell, P. Faure, 2004. “A Multi-Configuration Strategy for Application Dependent

Testing of FPGAs” In Proc. VLSI Test.

3. C.Stroud, P Chen, S.Konala, and M.Abramovici, 1995. “using ILA testing for BIST in FPGAs”. In proceedings of IEEE VLSI

Test Symposium, pp.259-265

4. Philip T.Vsuong, Asad M.Madni and Jim B. Vuong 2006. “VHDL Implementation for a Fuzzy Logic Controller”, In Bei

technologies, Inc.

5. Zeyad assi Obaid, Nasri Sulaiman and M. N. Hamidon July 2009. “FPGA-based Implementation of Digital Logic using Altera

DE2 Board” in IJCSNS International Journal of Computer Science and Nezwork Security, VOL.9 No.8.

6. A Summary of the IEC 61508 Standard for Functional Safety of Electrical/Electronic/Programmable Electronic Safety-

Related Systems , version 2.0, january2, 2006 http://www.exida.com/articles/iec61508_overview.pdf

7. Fuzzy sets. Information and Control. 1965;8:338-353

8. David J Smith, Kenneth G L Simpson 2004. “Functional Safety” A Straightforward Guide to applying IEC 61508 and Related

Standards. Elsevier Butterwoth-Heinemann, 2nd edition

9. Datasheet Analog to digital converter with 14-bit from Linear technology LTC6912 “Dual programmable gain Amplifiers

with Serial Digital Interface”. http://cds.linear.com/docs/Datasheet/6912fa.pdf

10. Datasheet Digital to analog converter with 12 bit from lineartechnologyLTC2624.

http://cds.linear.com/docs/Datasheet/2604fd.pdf

11. http://www.xilinx.com/support/documentation/sw_manuals/xilinx11/plugin_ism.pdf

12.Spartan-3A/3AN FPGA Starter Kit Board User Guide UG334 (v1.1) June 19, 2008,

http://www.exida.com/articles/iec61508_overview.pdf

http://cds.linear.com/docs/Datasheet/6912fa.pdf

http://cds.linear.com/docs/Datasheet/2604fd.pdf

http://www.xilinx.com/support/documentation/sw_manuals/xilinx11/plugin_ism.pdf


ISSN : 2250:3439



54

13. www.xilinx.com

http://www.xilinx.com/


ISSN : 2250:3439



55

Vectored Interrupt Controller Implementation of Advanced Bus Architecture on

FPGA

Vasanth H 1, Dr.A.R.Aswath

2

1 VLSI Design & Embedded systems, M.Tech,

DSCE, Bangalore-78,

[email protected] 2 Dept. of E&C, Professor, E&C

DSCE, Bangalore-78.

Abstract- Interrupt controller is designed with the concept of priority based selection

of peripherals which requires immediate attention or service. Here AHB is optimized

to interface with VIC to initiate data transfer on the AHB. Both read and write cycles

are designed with AHB bus.

Keywords: AMBA, AHB, Interrupt Controller, VIC.

1. INTRODUCTION

As no of components or peripherals are increasing on a single chip design of system

on chip is getting complicated. Design of System-on-Chip (SoC) receives a great

deal of attention in recent days. Interrupt controller are important due to the fact that

Processors and peripherals usually communicate with each other with interrupt [1].

With the development of SoC technique, the communication between processors and

peripherals becomes a problem as processors have limited interrupt ports which is far

less than the total interrupt signals of peripherals and other processors.

Fig1: An Interrupt controller in AMBA based system

An AMBA based microcontroller typically consists of a high performance system

backbone bus, able to sustain the external memory bandwidth, on which the CPU on-

chip memory and other vectored interrupt controller devices reside [2]. This bus

provides a high bandwidth interface between the elements that are involved in the

majority of transfers.

The AHB slave main function is an interface unit that allows AHB logic to initiate

a data transfer on the AHB. The AHB specifies the type transaction to be executed

on the slave through a user friendly interface.


ISSN : 2250:3439



56

The VIC is an Advanced Microcontroller Bus Architecture (AMBA) compliant

System-on-Chip (SoC) peripheral that is developed, tested, and licensed by ARM

[3].

II. METHODOLOGY:

In this project it requires the usage of Verilog HDL for designing the RTL code.

Simulation is done using Model Sim and corresponding synthesis is done with

Xilinix 12.2v and finally implementing on SPARTAN 3XC3S 400 FPGA board.

III. AMBA Specifications A. Overview of AMBA Specifications:

The Advanced Microcontroller Bus Architecture (AMBA) specification defines an

on-chip communications standard for designing high-performance embedded

microcontrollers [3]. Three distinct buses are defined within the AMBA

specification:

Advanced High-performance Bus (AHB)

Advanced System Bus (ASB)

Advanced Peripheral Bus (APB)

B. AMBA AHB:

AHB is a new generation of AMBA bus which is intended to address the

requirements of high-performance synthesizable designs [3]. It is a high-performance

system bus that supports multiple bus masters and provides high-bandwidth

operation. AMBA AHB implements the features required for high-performance, high

clock frequency systems including:

burst transfers

split transactions

single-clock edge operation

non-tristate implementation A typical design contains the following components:

AHB master

AHB slave

AHB arbiter

AHB decoder

IV. INTERRUPT CONTROLLER

A. Interrupt Controller:

It is a device used to combine several sources of interrupt onto one or more CPU

lines while allowing priority levels to be assigned to its interrupt outputs. Interrupt

with the highest priority level is asserted to processor for interrupt processing [4].

Many structures of interrupt controller have been proposed by former papers, as in

[5], but they are neither priority configurable nor interrupt-combinable.

The VIC is an Advanced Microcontroller Bus Architecture (AMBA) compliant

System-on-Chip (SoC) peripheral that is developed, tested, and licensed by ARM.

The VIC provides a software interface to the interrupt system. In a system with an

interrupt controller, software must determine the source that is requesting service and


ISSN : 2250:3439



57

where its service routine is loaded. A VIC does both of these in hardware [6]. It

supplies the starting address, or vector address, of the service routine corresponding

to the highest priority requesting interrupt source.

B. Interrupts in ARM:

In an ARM system, two levels of interrupt are available:

◦ Fast Interrupt Request (FIQ) for fast, low latency interrupt

handling

◦ Interrupt Request (IRQ) for more general interrupts.

C. Features of Interrupt Controller:

The features supported are:

Uses the AMBA AHB protocol.

Up to 32 interrupt source.

High level sensitive, interrupt source type.

Support for 32 vectored interrupts.

Fixed interrupt priority level.

Fixed IRQ and FIQ generation.

Software interrupts generation.

Interrupt enable.

Raw interrupt status.

Interrupt source get acknowledgment.

V. Design OF AMBA AHB INTERRUPT CONTROLLER

The vectored interrupt controller is mainly divided in to three blocks namely:

Peripheral interface

CPU interface

AHB slave interface

The Block diagram of vectored interrupt controller is shown in figure 2

Fig 2: Block diagram of vectored interrupt controller


ISSN : 2250:3439



58

A. Peripheral Interface:

Interrupt Request Logic:

Interrupt request logic receives 32 Intr_src lines from CPU peripherals and

combines with the software interrupt which are written by CPU on Soft_Int

register and enable the user selected interrupts by gated enabling and separate

the 32 request lines into 16 fast interrupt request and 16 general interrupt

request and also encode the filtered output generates two separate request id for

FIQ‘s and IRQ‘s.

Figure 3: Interrupt request logic

B. CPU Interface:

FIQ request handling:

Here we use fixed priority logic for lower 16 bits of Intr_src and generates the nfiq

signals which is active low and selects the vector address of the respective peripheral

from vectored table to the CPU.The FIQ request handling shown in figure 6 asserts

the nfiq signal. i.e. if FIQ_status is nonzero, set the nfiq as low. It selects the

vectored address of the corresponding fast interrupt request. Send it to CPU through

AHB slave interface


ISSN : 2250:3439



59

Figure 4:FIQ request handling

. It will select the vectored address from the vectored address table, vectored

address table is the memory configuration space, which contain the subroutine of the

each interrupt request. The vectored addresses in the vectored table are

programmable. nfiq is active low signal for CPU.

IRQ Request Handling:

Here we use fixed priority logic for lower 16 bits of Intr_src and generates the nfiq

signals which is active low and selects the vector address of the respective peripheral

from vectored table to the CPU.The IRQ request handling shown in figure 7 Asserts

the nirq signal. i.e., if irq_status is nonzero, set the nfiq as low and selects the

vectored address of the corresponding interrupt request. Send it to CPU through

AHB slave interface. It will select the vectored address from the vectored address

table, vectored address table is the memory configuration space, which contain the

subroutine of the each interrupt request.

IRQ_status acts as a select line for vectored address selection. nirq is active low

signal for CPU.

Figure 5:IRQ request handling

D. AHB Slave Interface:

An AHB bus slave responds to transfers initiated by bus masters within the

system.

The AHB slave shown in figure 6 maps the memory configuration space with the

interrupt controller and perform the data transaction as AHB asserts its signal. In this

block asserts Hready_out as high and Hresp as OKAY, because we designed

Interrupt controller as a single slave.


ISSN : 2250:3439



60

Figure 6: AHB slave interface

VI. SIMULATION RESULTS:

A. AHB Write Cycle: Figure 7 shows how AHB signals got asserted to perform Data write through the

AHB to Vectored Interrupt Controller. In order to write the Data Hwrite must be

high. Data write to the address which is specified on the Haddr bus through HWdata

bus

Figure 7: AHB Write Cycle

Red line on Hread signal is due to write operation being performed.

B. AHB Read Cycle

Figure 8 shows how AHB signals got asserted to perform Data read through the

AHB from Vectored Interrupt Controller. In order to read the Data Hwrite must be


ISSN : 2250:3439



61

low. Data read from the address which is specified on the Haddr bus through HRdata

bus.

Figure 8: AHB Read Cycle

C. Integrated Interrupt controller simulation:

Figure 9 shows how interrupt controller got asserted to perform generation of fiq,

irq to ack signals from CPU.

Figure 9: Interrupt Controller

VII IMPLEMENTATION RESULT


ISSN : 2250:3439



62

Synthesis of the Integrated Interrupt Controller implementation is shown below with respect to small functions

like gates, inverters large units like multiplexers, encoders [8].

Figure 10: Implementation results

VIII CONCLUSION

Vectored interrupt can serve two interrupt request at a time, one FIQ request and one IRQ request and it will give

two separate acknowledgements for both FIQ and IRQ requests. Due to the parallel processing of FIQ and IRQ

requests the VIC has a low latency has a faster execution speed.

Finally interrupt controller is implemented and its results are discussed

VII. References

1. Wei Chipin,―Design of a Configurable Multichannel Interrupt Controller ― 2010 Second

Pacific-Asia Conference on Circuits, Communications and System (PACCS)

2. ARM Corporation, Prime cell Vectored interrupt controller PL192, reference manual, 2002.

3. ARM Corporation, AMBA specification 3.0, reference manual, 2006

4. Horelick, Dale; ―Simple, Versatile CAMAC Crate Controller and Interrupt Priority ENC Oding Module

―Nuclear Science, IEEE Transactions on Volume: 22 ,Issue: 1 Digital Object Identifier:

10.1109/TNS.1975.4327692 Publication Year: 1975,

5. De Gloria, A. Faraboschi, P.Olivieri, M.;‖A self timed interrupt controller: a case study in asynchronous

micro-architecture design‖, ASIC Conference and Exhibit, 1994. Proceedings, Seventh Annual IEEE

International Digital Object Identifier: 10.1109/ASIC.1994.404555 Publication Year: 1994,

Page(s): 296 – 299

6. Vectored Interrupt Controller Usage and Applications-www.altera.com/literature/an/AN595.pdf

7. Digital Design principles and practices, John F. Wakerly, third edition, Prentice Hall

Publication, 2000.

8. J. Basker, a Verilog HDL Synthesis 1st Edition, 1998.

http://ieeexplore.ieee.org/search/srchabstract.jsp?tp=&arnumber=4327692&queryText%3Dinterrupt+controller%26openedRefinements%3D*%26searchField%3DSearch+All




http://dx.doi.org/10.1109/TNS.1975.4327692





http://dx.doi.org/10.1109/ASIC.1994.404555

http://www.altera.com/literature/an/AN595.pdf


ISSN : 2250:3439



63

Abstract— Temperature is a very important parameter in industrial production. Recently, lots of researches have been

investigated for the temperature control system based on various control strategies. This paper presents the comparison of

GA-PID, fuzzy and PID for temperature control of water bath system. Different control schemes namely PID, PID tuning

using Genetic Algorithms(GA-PID), and Fuzzy Logic Control, have been compared through experimental studies with respect

to set-points regulation, influence of impulse noise and sum of absolute error. The new algorithm based on GA-PID improve the

performance of the system. Also, it's fit for the complicated variable temperature control system. The simulation results show that

the validity of the proposed strategy is more effective to control temperature.

Keywords-Water Bath System, GA-PID controller, Fuzzy, PID

I. INTRODUCTION

emperature control is an important factor in many process control system [12,13]. If the temperature is too high

or too low, the final product is seriously affected. Therefore, it is necessary to reach some desired temperature

points quickly and avoid large overshoot. Since the process-control system are often nonlinear and tend to

change in an unpredictable way, they are not easy to control accurately. In general, most of the temperature control

systems use the conventional PID as it is non-linear, time varying and big lag. However, the conventional PID for

this non-linear system is difficult to achieve the desired effect of control. In addition, the parameters of PID need

make the corresponding adjustment when the characteristic of controlled object changes.

Fuzzy logic has been mainly applied to control problems with fuzzy if–then rules [4]. In most fuzzy control systems

fuzzy if–then rules were derived from human experts. Recently, several approaches have been proposed for

generating fuzzy if-then rules from numerical data [4].

A genetic algorithm (GA) is a parallel, global search technique that emulates operators. A GA applies operators

inspired by the mechanics of natural selection to a population of binary string encoding the parameter space at each

generation; it explores different areas of the parameter space, and then directs the search to regions where there is a

high probability of finding improved performance. In this paper, GA is used to tune gain of PID controller

In the following section, basic concepts &, modeling of controllers has been presented. Experimental setup is given

in section 3. Simulation results and comparison of various models is shown in section 4. Conclusions follow in

section 5.

II. DESIGN OF CONTROLLER

A. PID Controller

PID stands for Proportional-Integral-Derivative. This is as type of feedback controller whose output, a control

variable(CV), is generally based on the error (e) between some user-defined set-point (SP) and some measured

process variable (PV). Each element of the PID controller refers to a particular action taken on the error. For the PID

control, a velocity-form discrete PID controller [11] is used and is described by

[ ]

[ ]

=KP [e(k)-e(k-1)] + KI e(k)+ KD [e(k)-2e(k-1)+e(k-2)] (1)

Where

SARITA RANI1. SANJU SAINI

2, SANJEETA RANI

3

1,2Deenbandhu Chhotu Ram Univ. of Science & Technology,Murthal

3University Institute of Instrumentation Engineering ,Kurkshetra

2Astt. Prof. at Department of Electrical Engineering

A Comparative Analysis of GA-PID, Fuzzy and PID for Water Bath System

T


ISSN : 2250:3439



64

KP=-

KI, KI=K

, KD=K

(2)

Proportional: Error multiplied by a gain, Kp. This is an adjustable amplifier. In many systems Kp is responsible for

process stability: too low and the PV can drift away; too high and the PV can oscillate.

Integral: The integral error is multiplied by a gain Ki. In many systems Ki is responsible for driving error to zero,

but set Ki too high is to invite oscillation or instability or integrator windup or actuator saturation.

Derivative: The rate of change of error multiplied by a gain, Kd. In many systems Kd is responsible for system

response: too high and the PV will oscillate; too low the PV will respond sluggishly. The designer should also note

that derivative action amplifies any noise in the error signal.

“Tuning of a PID involves the adjustment of Kp, Ki and Kd to achieve some user-defined „optimal‟ character of a system

response.”

Although much architecture exists for control systems, the PID controller is mature and well-understood by

practitioners. For these reasons, it is often the first choice for new controller design.

B. Fuzzy Logic Controller

The Fuzzy controller developed here is a two-input single output controller. The two inputs are the deviation from

set point error, E, and error change rate, EC. This is usually used for temperature control system [8]. The error

means the difference between temperature measured and setting temperature. The error change rate means the

derivative of error change. The single output is the change of operating value, ΔU. The Fuzzy logic control system

and operational structure of Fuzzy controller are showed in Figure 1 which compromise four principal component:

Fuzzificaion interface

Knowledge Base

Decision Making Logic

Defuzzification Interface

Figure 6: Fuzzy Logic Controller

Fuzzification interface: Fuzzification is related to the vagueness and imprecision in a natural language. The

fuzzification module converts the input data into suitable linguistic values. In this paper, the domain of language

variable of input, system error e(k) and the rate of error ∆e(k), are{ -1, -0.8, -0.6, -0.4, -0.2, 0, 0.2,0.4,0.6,0.8 },the

fuzzy sets are {NB, NM, NS, ZO, PS, PM, PB}, the elements of the subset stand for the negative big, the negative

medium, the negative small, zero, small, middle, big. Gaussian membership function is used. The output variable

u(k) is the PWM signal between 0% and 100%.

Knowledge base: The knowledge base compromise knowledge of application domain and control goals. It consists

of ―data base‖ and ―linguistic control rule base‖. The data base provides necessary definitions, which are used to

define linguistic control rules and fuzzy data manipulation in an FLC.

Decision making logic: The decision-making logic is the kernel of FLC. It has the capability of simulating human

decision-making based on fuzzy concepts and of inferring fuzzy control action employing fuzzy implication and rule

of fuzzy inference in fuzzy logic.

Defuzzfication interface: The defuzzification performs the following functions:

A scale mapping, which converts the range of values of output value into corresponding universe of

discourse

Defuzzfication, which yields a non-fuzzy control action from an inferred fuzzy control action.


ISSN : 2250:3439



65

Figure 7: Membership functions of the three fuzzy variables of the water bath system. (a) Input variable e(k), (b) input variable ∆e(k), and (c) output variable ∆u(k)

C. GA-PID

GAs are powerful search optimization algorithms based on the mechanism of natural selection and genetics. Gain of

PID are tuned using GA.GA is implemented using matlab. The GA is coded using MATLAB. Typical values of

different parameters of GAs are taken. The programs use static values for maximum number of generations

(maxgen=120), probability of crossover (pc=0.08), and probability of mutation(pm=0.05). The initial population is

randomly generated. The population size (psize=60), is selected based on the observation. Roulette-wheel method is

used to select individuals for reproduction process. In the method, two strings from the population are selected at

random with their probability of selection being proportional to their fitness values. The selected strings undergo

crossover and mutation and become members of the new population. The GA uses the absolute error i.e. the

difference between the actual output and the reference input as a fitness function.

III. EXPERIMENTAL SETUP

Problem Statement

To see whether the proposed GA-PID can achieve good performance and overcome the disadvantages of the FLC

and PID, we compare it with the FLC and PID under the same aforementioned circumstances on a simulated water

bath temperature-control system. Consider a discrete time SISO temperature-control system

yP(k+1)=a(Ts)yP(k)+

[ ] (3)

where a(Ts) and b(Ts) are given by (11). The parameters for simulation are 1.00151e-4, 8.67973e-3,

40.0 and Y0=25C, which were obtained from a real water bath plant [8, 9].

Figure 3: Schematic diagram of Water bath temperature control system.

The plant input u(k) is limited between 0 and 5 V, and it is also assumed that the sampling period is Ts=30s . The

schematic diagram of the real water bath process is depicted in figure3. A personal computer reads the temperature

of the water bath through a link consisting of a diode-based temperature sensor module and an 8-bit analogue to

digital converter. The plant input produced by the computer is limited between 0 and 5 V and controls the duty cycle


ISSN : 2250:3439



66

for a heater through a pulse-width-modulation scheme. The task is to control the simulated system to follow three set

points.

{

IV.SIMULATION STUDIES

For the aforementioned controllers (GA-PID controller, PID controller and manually designed fuzzy controller), two

groups of computer simulations are conducted on the water bath temperature control system. In the first set of

simulations, the regulation capability of the three controllers with respect to set-point changes is studied. Three set-

points to be followed are

{

To test their regulation performance performance index, sum of absolute error (SAE), is defined by

SAE= ∑

where yref (k) and y(k) are the reference output and the actual output of the simulated system ,respectively.

,

(a) (b)

(c) Figure4: Regulation performance of three controller for water bath system. (a).GA-PID (b).FLC (c). PID

Results shows that GA-PID controller exhibits good regulation capability.FLC shows small overshoot at 55c..,but

PID controller shows poor regulation performance.

In second set of experiment, two impulse noise values 5.0C

and -5.0C

is added at output at the fortieth and

eightieth sampling instants, respectively. A set-point of 50.0C is performed in this set of experiments. Figure5 shows

behavior of controller to impulse noise

0 20 40 60 80 100 12020

30

40

50

60

70

80

k

Tem

pera

ture

(degre

e)

0 20 40 60 80 100 12020

30

40

50

60

70

80

k

Tem

pera

ture

(deg

ree)

0 20 40 60 80 100 12020

30

40

50

60

70

80

k

Tem

pera

ture

(deg

ree)


ISSN : 2250:3439



67

(b)

(c)

Figure 5: Behavior of the three controllers under the impulse noise for the water bath system. (a).GA-PID (b).FLC (c). PID

A summary of comparisons among the three controllers on the water bath temperature-control system is shown in

Table 1. Experimental study shows that GA-PID shows better result as comparison to conventional control like PID

controller

TABLE 1: SUMMARY OF COMPARISONS AMONG THE THREE CONTROLLERS ON THE EXPERIMENTAL WATERBATH SYSTEM

Design

Criteria

Regulation

performance

Influence of

impulse noise

GA-PID SAE=394.2 SAE=283.2

FUZZY SAE=409.3 SAE=295.3

PID SAE=419.4 SAE=305.7

V. CONCLUSION

In this paper, GA-PID, has resulted in better regulation performance & thus minimizing overall absolute error The

results show that the GA-PID can be easily applied to the in presence of unknown noise .Performance of GA-PID is

also better than the FLC and PID controller. SAE value of GA-PID is also less than the FLC and PID. In addition,

the proposed method would be applied in much industrial process with better performance.

REFERENCES [1] Hongbo Xin ,Tinglei Huang,Xiaoyu Liu, Xiangjiao Tang,―Temperature Control System Based On Fuzzy Self-Adpative PID

Controller‖,Third International Conference On Gentic And Evloutionary Computing ,2009

[2] Zang Huai-quai,LI Quan, ―The Automatic Temperature System With Fuzzy Self-adpative PID Control In Semiconductor Laser‖Proceeding Of IEEE International Conference On Automation And Logistics,2009

[3] Zhang Yongjun, Wang Zhixing, Wang Lili,―The Design Of Fuzzy PID Multi-channel Temperature Control System Based On NeuralNetwork‖, International Technology And Innovation Conference,2007

[4] Jia-Xin Chen, Wei Li, ―Application Of Fuzzy Control PID Algorithm In Temperature Controlling System‖, Proceeding Of Second

International Conference On Machine Learing And Cybernetics ,2003

0 20 40 60 80 100 12025

30

35

40

45

50

55

k

Tem

pera

ture

(deg

ree)

0 20 40 60 80 100 12020

30

40

50

60

k

Tem

pera

ture

(degre

e)

0 20 40 60 80 100 12025

30

35

40

45

50

55

k

Tem

pera

ture

(deg

ree)


ISSN : 2250:3439



68

[5] Aeenmeher, A. Yazdizadeh, M. S.Ghazizadeh,―Neuro-PID Control Of An Industrial Furnace Temperature‖, Symposium On Industrial

Electronic And Application IEEE,2003 [6] T.Thyagarajan, P.G.Rao,J.Shanmungam,M.Ponnavaikko,―Advanced Control Schems For Temperature Regulation Of Air Heat‖,

International Fuzzy System Conference IEEE,1999

[7] Chinn-Tang Lin,Chia-Feng, Chung-Ping Li,―Temperature Control With Neural Fuzzy Inference Network‖,IEEE Transaction On System,

Man And Cybernectics,Vol.29 No.3,2009

[8] Y. J.Chang, Y.M.Chen, C.A.Lee,Y.H.Wang,Y.C.Chen, ―Improving Temperature Control Of Laser Module FuzzyLogicTheroy‖,20th

IEEE Semi-Therm Symposium,2004

[9] Marzuki Khalid,Sigeru Omatu,―A Neural Network Controller for Temperature Control System‖IEEE Control Sytem,1992

[10] Mo Zhi ,Peng Xiao, Xiao Laisheng―Research and Application on Two –stage Fuzzy Temperature Control System for Industrial Heating

Furnace‖Secnode international conference on intelligent computattion technology IEEE,2009

[11] Adriano Cruz,Mestrdo NCE ―ANFIS-Adaptive Neuro-Fuzzy Inference System‖

[12] K. Ogata, Discrete-Time Control Systems. Englewood Cliffs, NJ:Prentice-Hall, 1987

[13] J.Tanomaru, S.Omatu ,―Process C ontrol By On-line Trained Neural Controller,‖IEEE Trans.Ind Electron.39,1992


ISSN : 2250:3439



69

Abstract— Temperature is a very important parameter in industrial production. Recently, lots of researches have been

investigated for the temperature control system based on various control strategies. This paper presents the comparison of

GA-PID, fuzzy and PID for temperature control of water bath system. Different control schemes namely PID, PID tuning

using Genetic Algorithms(GA-PID), and Fuzzy Logic Control, have been compared through experimental studies with respect

to set-points regulation, influence of impulse noise and sum of absolute error. The new algorithm based on GA-PID improve the

performance of the system. Also, it's fit for the complicated variable temperature control system. The simulation results show that

the validity of the proposed strategy is more effective to control temperature.

Keywords-Water Bath System, GA-PID controller, Fuzzy, PID

III. INTRODUCTION

emperature control is an important factor in many process control system [12,13]. If the temperature is too high

or too low, the final product is seriously affected. Therefore, it is necessary to reach some desired temperature

points quickly and avoid large overshoot. Since the process-control system are often nonlinear and tend to

change in an unpredictable way, they are not easy to control accurately. In general, most of the temperature control

systems use the conventional PID as it is non-linear, time varying and big lag. However, the conventional PID for

this non-linear system is difficult to achieve the desired effect of control. In addition, the parameters of PID need

make the corresponding adjustment when the characteristic of controlled object changes.

Fuzzy logic has been mainly applied to control problems with fuzzy if–then rules [4]. In most fuzzy control systems

fuzzy if–then rules were derived from human experts. Recently, several approaches have been proposed for

generating fuzzy if-then rules from numerical data [4].

A genetic algorithm (GA) is a parallel, global search technique that emulates operators. A GA applies operators

inspired by the mechanics of natural selection to a population of binary string encoding the parameter space at each

generation; it explores different areas of the parameter space, and then directs the search to regions where there is a

high probability of finding improved performance. In this paper, GA is used to tune gain of PID controller

In the following section, basic concepts &, modeling of controllers has been presented. Experimental setup is given

in section 3. Simulation results and comparison of various models is shown in section 4. Conclusions follow in

section 5.

IV. DESIGN OF CONTROLLER

D. PID Controller

PID stands for Proportional-Integral-Derivative. This is as type of feedback controller whose output, a control

variable(CV), is generally based on the error (e) between some user-defined set-point (SP) and some measured

process variable (PV). Each element of the PID controller refers to a particular action taken on the error. For the PID

control, a velocity-form discrete PID controller [11] is used and is described by

[ ]

[ ]

=KP [e(k)-e(k-1)] + KI e(k)+ KD [e(k)-2e(k-1)+e(k-2)] (1)

Where

KP=-

KI, KI=K

, KD=K

(2)

A Comparative Analysis of GA-PID, Fuzzy and PID for Water Bath System

SARITA RANI1. SANJU SAINI

2, SANJEETA RANI

3

1,2Deenbandhu Chhotu Ram Univ. of Science & Technology,Murthal

3University Institute of Instrumentation Engineering ,Kurkshetra

2Astt. Prof. at Department of Electrical Engineering

T


ISSN : 2250:3439



70

Proportional: Error multiplied by a gain, Kp. This is an adjustable amplifier. In many systems Kp is responsible for

process stability: too low and the PV can drift away; too high and the PV can oscillate.

Integral: The integral error is multiplied by a gain Ki. In many systems Ki is responsible for driving error to zero,

but set Ki too high is to invite oscillation or instability or integrator windup or actuator saturation.

Derivative: The rate of change of error multiplied by a gain, Kd. In many systems Kd is responsible for system

response: too high and the PV will oscillate; too low the PV will respond sluggishly. The designer should also note

that derivative action amplifies any noise in the error signal.

“Tuning of a PID involves the adjustment of Kp, Ki and Kd to achieve some user-defined „optimal‟ character of a system

response.”

Although much architecture exists for control systems, the PID controller is mature and well-understood by

practitioners. For these reasons, it is often the first choice for new controller design.

E. Fuzzy Logic Controller

The Fuzzy controller developed here is a two-input single output controller. The two inputs are the deviation from

set point error, E, and error change rate, EC. This is usually used for temperature control system [8]. The error

means the difference between temperature measured and setting temperature. The error change rate means the

derivative of error change. The single output is the change of operating value, ΔU. The Fuzzy logic control system

and operational structure of Fuzzy controller are showed in Figure 1 which compromise four principal component:

Fuzzificaion interface

Knowledge Base

Decision Making Logic

Defuzzification Interface

Figure 8: Fuzzy Logic Controller

Fuzzification interface: Fuzzification is related to the vagueness and imprecision in a natural language. The

fuzzification module converts the input data into suitable linguistic values. In this paper, the domain of language

variable of input, system error e(k) and the rate of error ∆e(k), are{ -1, -0.8, -0.6, -0.4, -0.2, 0, 0.2,0.4,0.6,0.8 },the

fuzzy sets are {NB, NM, NS, ZO, PS, PM, PB}, the elements of the subset stand for the negative big, the negative

medium, the negative small, zero, small, middle, big. Gaussian membership function is used. The output variable

u(k) is the PWM signal between 0% and 100%.

Knowledge base: The knowledge base compromise knowledge of application domain and control goals. It consists

of ―data base‖ and ―linguistic control rule base‖. The data base provides necessary definitions, which are used to

define linguistic control rules and fuzzy data manipulation in an FLC.

Decision making logic: The decision-making logic is the kernel of FLC. It has the capability of simulating human

decision-making based on fuzzy concepts and of inferring fuzzy control action employing fuzzy implication and rule

of fuzzy inference in fuzzy logic.

Defuzzfication interface: The defuzzification performs the following functions:

A scale mapping, which converts the range of values of output value into corresponding universe of

discourse

Defuzzfication, which yields a non-fuzzy control action from an inferred fuzzy control action.


ISSN : 2250:3439



71

Figure 9: Membership functions of the three fuzzy variables of the water bath system. (a) Input variable e(k), (b) input variable ∆e(k), and (c) output variable ∆u(k)

F. GA-PID

GAs are powerful search optimization algorithms based on the mechanism of natural selection and genetics. Gain of

PID are tuned using GA.GA is implemented using matlab. The GA is coded using MATLAB. Typical values of

different parameters of GAs are taken. The programs use static values for maximum number of generations

(maxgen=120), probability of crossover (pc=0.08), and probability of mutation(pm=0.05). The initial population is

randomly generated. The population size (psize=60), is selected based on the observation. Roulette-wheel method is

used to select individuals for reproduction process. In the method, two strings from the population are selected at

random with their probability of selection being proportional to their fitness values. The selected strings undergo

crossover and mutation and become members of the new population. The GA uses the absolute error i.e. the

difference between the actual output and the reference input as a fitness function.

III. EXPERIMENTAL SETUP

Problem Statement

To see whether the proposed GA-PID can achieve good performance and overcome the disadvantages of the FLC

and PID, we compare it with the FLC and PID under the same aforementioned circumstances on a simulated water

bath temperature-control system. Consider a discrete time SISO temperature-control system

yP(k+1)=a(Ts)yP(k)+

[ ] (3)

where a(Ts) and b(Ts) are given by (11). The parameters for simulation are 1.00151e-4, 8.67973e-3,

40.0 and Y0=25C, which were obtained from a real water bath plant [8, 9].

Figure 3: Schematic diagram of Water bath temperature control system.

The plant input u(k) is limited between 0 and 5 V, and it is also assumed that the sampling period is Ts=30s . The

schematic diagram of the real water bath process is depicted in figure3. A personal computer reads the temperature

of the water bath through a link consisting of a diode-based temperature sensor module and an 8-bit analogue to

digital converter. The plant input produced by the computer is limited between 0 and 5 V and controls the duty cycle


ISSN : 2250:3439



72

for a heater through a pulse-width-modulation scheme. The task is to control the simulated system to follow three set

points.

{

IV.SIMULATION STUDIES

For the aforementioned controllers (GA-PID controller, PID controller and manually designed fuzzy controller), two

groups of computer simulations are conducted on the water bath temperature control system. In the first set of

simulations, the regulation capability of the three controllers with respect to set-point changes is studied. Three set-

points to be followed are

{

To test their regulation performance performance index, sum of absolute error (SAE), is defined by

SAE= ∑

where yref (k) and y(k) are the reference output and the actual output of the simulated system ,respectively.

,

(a) (b)

(c) Figure4: Regulation performance of three controller for water bath system. (a).GA-PID (b).FLC (c). PID

Results shows that GA-PID controller exhibits good regulation capability.FLC shows small overshoot at 55c..,but

PID controller shows poor regulation performance.

In second set of experiment, two impulse noise values 5.0C

and -5.0C

is added at output at the fortieth and

eightieth sampling instants, respectively. A set-point of 50.0C is performed in this set of experiments. Figure5 shows

behavior of controller to impulse noise

0 20 40 60 80 100 12020

30

40

50

60

70

80

k

Tem

pera

ture

(degre

e)

0 20 40 60 80 100 12020

30

40

50

60

70

80

k

Tem

pera

ture

(deg

ree)

0 20 40 60 80 100 12020

30

40

50

60

70

80

k

Tem

pera

ture

(deg

ree)


ISSN : 2250:3439



73

(b)

(c)

Figure 5: Behavior of the three controllers under the impulse noise for the water bath system. (a).GA-PID (b).FLC (c). PID

A summary of comparisons among the three controllers on the water bath temperature-control system is shown in

Table 1. Experimental study shows that GA-PID shows better result as comparison to conventional control like PID

controller

TABLE 1: SUMMARY OF COMPARISONS AMONG THE THREE CONTROLLERS ON THE EXPERIMENTAL WATERBATH SYSTEM

Design

Criteria

Regulation

performance

Influence of

impulse noise

GA-PID SAE=394.2 SAE=283.2

FUZZY SAE=409.3 SAE=295.3

PID SAE=419.4 SAE=305.7

V. CONCLUSION

In this paper, GA-PID, has resulted in better regulation performance & thus minimizing overall absolute error The

results show that the GA-PID can be easily applied to the in presence of unknown noise .Performance of GA-PID is

also better than the FLC and PID controller. SAE value of GA-PID is also less than the FLC and PID. In addition,

the proposed method would be applied in much industrial process with better performance.

REFERENCES [14] Hongbo Xin ,Tinglei Huang,Xiaoyu Liu, Xiangjiao Tang,―Temperature Control System Based On Fuzzy Self-Adpative PID

Controller‖,Third International Conference On Gentic And Evloutionary Computing ,2009

[15] Zang Huai-quai,LI Quan, ―The Automatic Temperature System With Fuzzy Self-adpative PID Control In Semiconductor Laser‖Proceeding Of IEEE International Conference On Automation And Logistics,2009

[16] Zhang Yongjun, Wang Zhixing, Wang Lili,―The Design Of Fuzzy PID Multi-channel Temperature Control System Based On NeuralNetwork‖, International Technology And Innovation Conference,2007

[17] Jia-Xin Chen, Wei Li, ―Application Of Fuzzy Control PID Algorithm In Temperature Controlling System‖, Proceeding Of Second

International Conference On Machine Learing And Cybernetics ,2003

0 20 40 60 80 100 12025

30

35

40

45

50

55

k

Tem

pera

ture

(deg

ree)

0 20 40 60 80 100 12020

30

40

50

60

k

Tem

pera

ture

(degre

e)

0 20 40 60 80 100 12025

30

35

40

45

50

55

k

Tem

pera

ture

(deg

ree)


ISSN : 2250:3439



74

[18] Aeenmeher, A. Yazdizadeh, M. S.Ghazizadeh,―Neuro-PID Control Of An Industrial Furnace Temperature‖, Symposium On Industrial

Electronic And Application IEEE,2003 [19] T.Thyagarajan, P.G.Rao,J.Shanmungam,M.Ponnavaikko,―Advanced Control Schems For Temperature Regulation Of Air Heat‖,

International Fuzzy System Conference IEEE,1999

[20] Chinn-Tang Lin,Chia-Feng, Chung-Ping Li,―Temperature Control With Neural Fuzzy Inference Network‖,IEEE Transaction On System,

Man And Cybernectics,Vol.29 No.3,2009

[21] Y. J.Chang, Y.M.Chen, C.A.Lee,Y.H.Wang,Y.C.Chen, ―Improving Temperature Control Of Laser Module FuzzyLogicTheroy‖,20th

IEEE Semi-Therm Symposium,2004

[22] Marzuki Khalid,Sigeru Omatu,―A Neural Network Controller for Temperature Control System‖IEEE Control Sytem,1992

[23] Mo Zhi ,Peng Xiao, Xiao Laisheng―Research and Application on Two –stage Fuzzy Temperature Control System for Industrial Heating

Furnace‖Secnode international conference on intelligent computattion technology IEEE,2009

[24] Adriano Cruz,Mestrdo NCE ―ANFIS-Adaptive Neuro-Fuzzy Inference System‖

[25] K. Ogata, Discrete-Time Control Systems. Englewood Cliffs, NJ:Prentice-Hall, 1987

[26] J.Tanomaru, S.Omatu ,―Process C ontrol By On-line Trained Neural Controller,‖IEEE Trans.Ind Electron.39,1992

.


ISSN : 2250:3439



75

Envision of student‟s concert using supervised learning techniques S. Anupama Kumar

1 and Dr. Vijayalakshmi M.N

2

1 Research Scholar, PRIST University, 1Assistant Professor, Dept of M.C.A. 2Associate Professor, Dept of MCA,1,2 R.V.College of Engineering, Bangalore, India.

[email protected] , 2 [email protected]

Abstract: Educational data mining is an emerging technology concerned with developing methods for exploring the various

unique data that exists in the educational settings and uses them to understand the students as well as the domain in which they

learn. Educational domain consists of a lot of data related to students, teachers and other learning strategies. Classification

algorithms can be used on various educational data to mine the academic records. It can be used to predict student‘s outcome

based on their previous academic performance. The various predictive algorithms like, C4.5, Random tree are applied on

student‘s previous academic results to predict the outcome of the students in the university examination. The prediction would

help the tutor in understanding the progress and attitude of the student towards the studies. It would also help them to identify the

students who are constantly improving in their studies and help them to achieve a higher percentage. It also helps them to identify

the underperformers so that extra effort can be taken to achieve a better result. The algorithms are analyzed based on their

accuracy of predicting the result, the recall and the precision values. The accuracy of the algorithm is predicted by comparing the

output generated by the algorithm with the original result obtained by the students in the university examination.

KEYWORDS: Educational EDM, Prediction, Decision trees, Recall, Precision.

.

1 Introduction

Data mining, also called Knowledge Discovery in Databases (KDD), is the field of discovering novel and potentially

useful information from large amounts of data. It can be applied on educational data to bring out new information

hidden in the data. Enormous data pertaining to student information is available in educational institutions which can

be mined to improve the quality of the students as well as educational institutions. EDM is the process of

transforming raw data compiled by education systems in useful information that could be used to take informed

decisions and answer research questions [2]. The data can be taken from student‘s learning environments, academic

data pertaining to their scores in the examination, grade obtained by them, previous academic achievement etc.

These data often has multiple levels of meaningful hierarchy depending on time and sequence. The meaningful

context of the data play important role in the study and analysis of the students. The various data mining techniques

like classification, clustering and association rule mining can be efficiently used on educational data to bring out

new knowledge from it [1]. Classification techniques like decision trees, naïve bayes algorithm, support vector

machines etc can be efficiently used on student assessment data to predict their outcome in the university

examination using the intermediate marks obtained in the internal examination [6].

Examination and assessment plays a major role in all the stages of a student. The result of a student depends on the

marks obtained by the student. Therefore predicting student‘s result is an important issue for any institution. If the

result of a student can be predicted at an early stage, it would help the institution to bring out betterment in the

result. From [6] it is clear that decision trees can be used to predict the student‘s performance using the marks

obtained by them in the internal examination. Apart from the internal marks various other parameters can also be

used to predict the student‘s performance in the examination. This paper aims to mine the student‘s academic data

available in the records from their school till fourth semester. The paper aims to predict the result which the student

would obtain in the V semester. Classification algorithms like C4.5 and Random Tree are used to mine the records

and predict the outcome of the students in the V semester. These algorithms are analyzed using the following

parameters.

1. The number of instances predicted correctly by the miner and the algorithm

2. The accuracy of the algorithm is analyzed by comparing it with the original result

3. The recall value of the algorithm

4. The precision value of the algorithm

This paper is organized into Introduction, Application of supervised learning techniques on educational data, Data

collection and visualization, Implementation of the decision tree algorithms, result and conclusion.

2 APPLICATIONS OF SUPERVISED LEARNING TECHNIQUES IN EDUCATIONAL

DATA MINING




ISSN : 2250:3439



76

Machine learning is one of the most popular techniques used in the field of data mining. It can be classified as

Supervised and Unsupervised learning. In supervised learning the variables under investigation can be split into two

groups: explanatory variables and one (or more) dependent variables. The target of the analysis is to specify

a relationship between the explanatory variables and the dependent variable as it is done in regression analysis. In

unsupervised learning all the variables are treated in the same way, there is no distinction between explanatory and

dependent variables. Predictive modelling is the process by which a model is created or chosen to best predict the

probability of an outcome. In many cases the model is chosen on the basis of detection theory and tries to guess the

probability of an outcome given a set amount of input data. Classification is a predictive data mining technique

which makes predication about values of data using known results found from different data [4]. Predictive models

have the specific aim of allowing us to predict the unknown value of a variable of interest given known values of

other variables. Predictive modelling can be thought of as learning a mapping from an input set of vector

measurements x to a scalar output y [5]. Classification maps data into predefined groups called as classes. It is often

referred to as supervised learning because the classes are determined before examining the data. They often describe

these classes by looking at the characteristic of data already known to belong to the classes [4]. The various

classification techniques which come under supervised learning techniques include decision trees, naïve bayes

networks, neural networks etc. They can be used to

Predict Academic success for students

Predict the Course Outcomes

Succeeding the next task

Meta cognitive skills, habits and motivation

The author in [7] has predicted the students who are at risk by building a predictive model and implementing that

model for a group of students in a particular course. This paper concentrates on applying the various decision tree

algorithms like C4.5, Random tree in predicting the academic success.

3 DATA COLLECTION AND VISUALIZATION:

Classification is the most commonly applied data mining technique, which employs a set of pre-classified examples

Education system can be either traditional or conventional. In case of higher education, the teaching – learning

process is carried out by using traditional way of teaching and using the ICT‘s for the same. In [8] ,the authors has

explained how clustering techniques can be used in an online education system to predict the student‘s drop out

ratio. Behrouz Minaei-Bidgoli and et al [10] have explained how prediction techniques can be used in an virtual

environment using LONCAPA system. The Indian education system is more traditional where the student‘s are

expected to be physically present in the class and the examination is conducted in the class only using traditional

method. The papers are assessed by the teachers and the scores are recorded. The percentage of marks and grades

obtained by the students of a particular course is taken as a data set for the implementation of the algorithms. The

data set consists of 16 attributes and 55 instances of a particular batch of students of a course. The details of the data

set are given below. Table 1 : Attributes of the dataset

S.No Attribute Information Category Information

1 UID University id Discrete 55 values

2 X percentage %of marks Continuous -

3 X class Class obtained Discrete 3 values

4 XII percentage %of marks Continuous -

5 XII class Class obtained Discrete 3 values

6 Degree percentage %of marks Continuous -

7 Degree Class Class obtained Discrete 3 values

8 I sem Percentage %of marks Continuous -

9 I sem class Class obtained Discrete 3 values

10 II sem Percentage %of marks Continuous -

11 II sem Class Class obtained Discrete 3 values

12 III sem percentage %of marks Continuous -

13 III sem class Class obtained Discrete 2 values

14 IV sem percentage % of marks Continuous -

15 IV sem class Class obtained Discrete 2 values

16 Predicted class Class expected Discrete 2 values


ISSN : 2250:3439



77

The attributes UID consists the university id no the student, the other attributes constitutes the percentage of marks

and class obtained by the students in X std, XII std, Degree and four semesters of a particular course. The

description of the percentage and the class obtained by the student is described below:

Table 2: Description of Data set

Percentage of

marks obtained

Class obtained Description

40_59 SC Second class

60_70 FC First Class

70_100 FCD First Class with

Distinction

These attributes are used to predict the outcome of the student in the fifth semester. Figure 1 explains how the

percentage of the students are split from X std to their percentage in the forth semester. From the graph it is clear

that there is a growth in the graph from X std percentage to IV semester percentage. The student‘s percentage has

started with 45% in X std where the IV semester percentage begins with 62%. Therefore it is clear that there is

increase in the percentage of marks the students are obtaining in the college as compared with schools

Fig 1: Data visualization of X std percentage v/s. IV semester percentage

4 IMPLEMENTATION OF DECISION TREE ALGORITHM

Supervised learning algorithms can be classified as decision tree algorithms, rule based algorithms, naïve bayes

function algorithms etc. In [3] the author has explained how rule based algorithms can be used to predict the

behavior of a student in an e learning system. Zlatko J. Kovačić [7] has explained how CART trees can be used as a

classification technique to devise a predictive model to identify the at risk students so as to yield a better result. The

author of [9] has explained how decision trees can be effectively used to estimate the motivational level of the

students using various parameters. Since the decision algorithms can work efficiently on a various types of

attributes, they are implemented for the above set of educational data.

4.1 C4.5 Decision Tree algorithm:

C4.5 classification algorithm builds a decision tree from a fixed set of examples. The resulting tree is used to

classify future samples. The example has several attributes and belongs to a discrete class. The leaf nodes of the

decision tree contain the class name whereas a non-leaf node is a decision node. The decision node is an attribute

test with each branch being a possible value of the attribute. Prediction class attribute is defined as a target attribute


ISSN : 2250:3439



78

(discrete value FC/FCD) and all other attributes are constituted as input attributes. The output of the algorithm can

be studied in the form of a confusion matrix. Table3

displays the output produced by the C4.5 algorithm for the given data set in the form of a confusion matrix.

Table 3 : Comparison of instances FCD and FC by C4.5

No. of

students in

FCD

No. of

students in

FC

Total No of students

No. of students

in FCD

35 1 36

No.of students

in FC

3 16 19

Total No of

students

38 17 55

From the confusion matrix it is clear that, Out of the 55 instances, 51 instances are correctly predicted and 4

instances are incorrectly predicted by the miner.

The number of instances who are actually in FCD are 38 and predicted are 36. The number of instances who are

actually in FC are 17 and predicted are 19.

The rule followed by the algorithm for prediction is given below:

Rule 1: Class iv in [FCD] then prediction class = FCD (92.11%of 38 examples).

Rule 2 : Class iv in [FC] then prediction class = FC (94.12% of 17 examples)

4.2 Random Tree:

The same data set is implemented using the random tree classification technique. Table 4 explains the confusion

matrix created by the random tree algorithm. From table 4 it is clear that the output predicted by the miner and the

algorithm are same ie all the instances are correctly predicted by the miner as well as the algorithm

Table 4 : Comparison of instances FCD and FC by RnD tree

No. of

students in

FCD

No. of

students in

FC

Total No of

students

No. of

students in

FCD

36 0 36

No. of

students in

FC

0 19 19

Total No of

students

36 19 55

The decision outcomes created by the algorithm for deriving the tree are as follows:

If class iv in [FCD] then If class ii in [FCD] then

If II < 73.3300 then

If III < 69.7150 then

If xiistd < 72.8800 then prediction = FC (100.00 % of 1 examples)

if xiiSTD > = 72.8800 then prediction = FCD (100.00 % of 1 examples)

if III >= 69.7150 then prediction = FCD (100.00 % of 4 examples)

if II >= 73.3300 then prediction = FCD (100.00 % of 29 examples)if class ii in [FC] then


ISSN : 2250:3439



79

if xii STD < 59.8350 then prediction = FC (100.00 % of 2 examples)

if xii STD >= 59.8350 then prediction = FCD (100.00 % of 1 examples)

if class ii in [SC] then prediction = FCD (0.00 % of 0 examples)

if class iv in [FC] then

if II < 74.2350 then prediction = FC (100.00 % of 16 examples)

if II >= 74.2350 then prediction = FCD (100.00 % of 1 examples)

From the above rules it is clear that the tree works on all the given examples and all the attributes are considered for

deriving a confusion matrix

5 Results The results obtained by the algorithms are analyzed using the following criteria.

1. The number of instances predicted correctly by the miner and the algorithm

2. The accuracy of the algorithm is analyzed by comparing it with the original result

3. The recall and precision value of the algorithm

subject using their roll number easily and give them extra coaching so as to get better results. This will also help the

educational institution to bring out quality results from the students. In future, various other attributes like

attendance, the marks scored in the previous semester can also be incorporated to predict the student‘s outcome.

5.1 : Comparison using Instances:

Table5 explains the comparison of the algorithms in terms of the number of instances Table 5: Comparison of C4.5 and Rnd tree

Algorithm No. of instances

correctly predicted

No. of instances

incorrectly predicted

C 4.5 51 4

Random

Tree

55 0

C4.5 algorithm has predicted 51 instances correctly out of 55 instances where there is a variation 1 instance

predicted as FCD by the miner and predicted as FC by the algorithm. Similarly, 3 instances are predicted FC by the

miner and FCD by the algorithm. Random tree algorithm has not brought out any change in the prediction by the

miner as well as the algorithm. 5.2 : Comparison using Accuracy:

The accuracy of the algorithms is verified using the result obtained from the university. Table5 shows the

comparison of the results obtained by the algorithms and the university results.

Table 6: Comparison of results of C4.5 and Random tree

Algorithm No. of instances in

FCD

No. of instances

in FC

C 4.5 38 17

Random

Tree

36 19

University 40 15

From table6 it is clear that C4.5 algorithm could predict better than the Random tree since the results are 95%

accurate towards the results obtained from the university.


ISSN : 2250:3439



80

5.3 Comparison using Recall and precision:

Any classification algorithm holding a lower precision and recall value are considered better than the other which

has a higher precision and recall values. It can be calculated using the true and false positive as well as negative

values obtained from the confusion matrix. Table6 shows the Recall and -precision obtained by the c4.5 algorithfor

the target attribute. Table 7: Recall and Precision values by C4.5

Table7 gives the recall and precision values of both FCD and FC produced by the C4.5 algorithm. It is clear that the

values are less than 1. The lesser the values indicates that the algorithm is more accurate.

The table8 gives the recall and precision values obtained by random tree algorithm.

Table 8: Recall and Precision values by Random tree

Since the algorithm could not produce any incorrect instances the true negative and false negative values are nil.

Therefore the precision values of the attributes are zero. Even though the precision values are zero, the recall value

is positive. Therefore Random tree algorithm is not more effective than the C4.5 algorithm.

6. Conclusion:

The application of classification algorithms on educational data can effectively help the tutor to understand the

student‘s interest towards academics and can bring improvement in their result. The percentage of marks and the

class obtained by the student from his school till he reaches a higher education is definitely an efficient attribute in

predicting the outcome of a student in the form of a grade. C4.5 can be used to accurately predict the outcome of the

student in the university exam. This is proved true for the given data set. The analysis of the algorithm using the

recall and precision values clearly states that the C4.5 algorithm performs better than the random tree algorithm. The

accuracy of the algorithm is also proved by comparing it with the original result. This also states that C4.5 algorithm

is 5% better than the random tree for the given data set. Therefore for a given data set C4.5 algorithm predicts the

student‘s outcome in the examination better than the random tree algorithm.

References

1. S.Anupama Kumar, Dr.Vijayalakshmi M.N, ―A Novel Approach in Data Mining Techniques for Educational Data‖, Proc 2011

3rd Internal Conference on Machine Learning and Computing (ICMLC 2011) , Singapore, Feb 2011,ISBN 978-1-4244-9252-7,pp

V4-152-154.

2. Cecily Heiner, Ryan Baker y Kalina Yacef, -Proceedings of the Workshop on Educational Data Mining at the 8th International

Conference on Intelligent Tutoring Systems Jhongli, Taiwan.,2006.

3. Félix Castro & Àngela Nebot and el al ,‖Extraction of Logical rules to predict student‘s behaviour‖, Proceedings of IASTED

International conference on web education, pp 164-170

4. Margret H. Dunham, ―Data Mining: Introductory and advance topic‖.

Value Recall Precision

FCD 0.9722 0.0789

FC 0.8421 0.588

Value Recall Precision

FCD 1.00 0.0

FC 1.00 0.0


ISSN : 2250:3439



81

.5. David Hand, Heikki, Mannil Padraic smyth,―Principles of Data Mining‖ PHI

6. S.Anupama Kumar and et al , ―Prediction of the student‘s recital using classification Technique ", IFRSA‘s

International Journal Of Computing,Vol1,issue 3,July. ISSN (Online) : 2230-9039 ,ISSN: 2231-2153, 2011 pp 305-309

7. Zlatko J. Kovačić and et al, ―Predictive working tool for at risk students‖, Creative Commons 3.0 New Zealand Attribution

Non-commercial Share Alike Licence (BY-NC-SA).

8. Tuomas Tanner and Hannu Toivonen ,― Predicting and preventing student failure – using the k-nearest neighbour method to

predict student performance in an online course environment‖,

University of Helsinki.

9. Mihaela Cocea & Stephan Weibelzahl, Can Log Files Analysis Estimate Learners‘ Level of Motivation?, Proceedings of

Lernen,Wissensentdeckung-Adaptivitat pp 32-35, Germany.

10. Behrouz Minaei-Bidgoli and et al ,Predicting Student Performance: An Application Of Data Mining Methods With The

Educational Web-Based System Lon-Capa‖, 33rd ASEE/IEEE Frontiers in Education Conference.

11. Dille, B., & Mezack, M.: Identifying predictors of high risk among community college telecourse students. The American

Journal of Distance Education, (1991) 5(1), 24-35.

Propagation Assessment of MB-OFDM Ultrawide

Band (UWB) MIMO Based Communication in

Fading Channels

Affum Emmanuel1 , Edward Ansong

2

1Electrical and Electronic Department, Koforidua Polytechnic, Koforidua, Ghana

Email:[email protected] 2Computer Science Department, Valley View University, Accra, GhanaEmail:

[email protected]

Abstract

Ultra wideband (UWB) technology is one of the promising solutions for future short-range

communication which has recently received a great attention by many researchers. However,

interest in UWB devices prior to 2001 was primarily limited to radar systems, mainly for

military applications due to bandwidth resources becoming increasingly scarce and also its

interference with other commutation networks. This research work provides performance

analysis of multiband orthogonal frequency division multiplexing (MB-OFDM) UWB MIMO

system in the presence of binary phase-shift keying time-hopping (BPSK-TH) UWB or BPSK-

DS UWB interfering transmissions under Nakagami-m and Lognormal fading channels

employing various modulation schemes using MATLAB simulations. The research work

indicates that it is totally impossible to predict the performance of UWB system in

Lognormal channel.

Keywords ─ Multiband (MB), UWB, multiple interferers, orthogonal frequency-

division multiplexing

(OFDM), 802.15.3a, MIMO.

_____________________________________________________________________

_______________________

1 Introduction

Ultra wideband (UWB) characterizes transmission systems with instantaneous

spectral occupancy in excess of 500 MHz [1] and is a fast emerging technology with

uniquely attractive features inviting major advances in wireless communications,

networking, radar, imaging, and positioning systems. Interestingly scholars and

researchers have predicted that it is the promising solutions for future short-



ISSN : 2250:3439



83

range communication thus UWB transmission has recently received significant

attention in both academia and industry for applications in wireless communications

[2]. After two proposals have been made by IEEE 802.15.3a task group as future ultra

wideband (UWB) standards: multiband orthogonal frequency division multiplexing

(MB-OFDM) UWB and Direct Sequence UWB (DS-UWB) approach [3], the demand

for higher data rates has become necessary and one possible solution is to exploit

both spatial and multipath diversities via the use of multiple-input multiple-output

(MIMO) and proper coding techniques. Lui and the group [4] investigated MIMO-

OFDM multiband UWB system in Nakagami fading. Also, in [5], the authors

analyzed error performance of MIMO-based singleband and multiband UWB

system in lognormal fading channel but they did not consider OFDM technique in

their work. Cao and et al [6] looked at uncoded adaptive modulated MIMO-OFDM

benefit from both multiple transmit/receive antenna diversity order and exploiting

the multipath diversity of the UWB channel. The authors in [7] also investigated

BER variation in an UWB OFDM MIMO communications system based on

measurements made in a picocell Wireless environment. Up to now, there is not

much research to evaluate the performance of MIMO- OFDM multiband UWB

system. Authors in [8] concentrated on performance MIMO-OFDM multiband UWB

system and derived a lower bound of pairwise-error probability (PEP) of the system

using lognormal fading channel. The research work is organized as follows;

Section 2, will focus on UWB signals models with Section 3, describing the

OFDM-MIMO transceiver system as well as MIMO frequency selective channel. In

Section 4, we provide simulation results and Section 5 summarizes and concludes the

research work.

2 UWB Spectrum „Spreading‟ Models

The two main approaches to randomizing the pulse train are time hopping (TH)

and direct sequence (DS) techniques [9]. Consider UWB channel with users

and interferers, , the user is assigned a unique random spreading

(of length data length).Let be the spreading sequence associated to

user .User then will be transmitting a spreaded signal consisting of

frames of chips of length .Each chip has power√ the power of each frame is

then(√ ), being the processing gain or the spreading gain. With DS-BPSK

UWB model the user transmits.

∑ ∑√

A typical UWB TH-BPSK can be molded as follows:

∑ √

(1)

(2)


ISSN : 2250:3439



84

where t is the time index, is the transmitted UWB pulse shape w i t h pulse

duration and as energy per pulse. Refers to frame of length The OFDM

received signal is given by [2]:

where is the received signal for the OFDM symbol is the interfering

signal applying the same approach in [2, 3, 12] and is the receiver noise.

3 MIMO-OFDM Multiband System

3.1 Transmitter Description

Consider multiband UWB system with transmit antenna a n d receive

antennas, as shown in Figure 1. At the transmitter, the coded information sequence

from a channel encoder is divided into blocks of bits. Each block is mapped

onto Space–Time–Frequency (STF) codeword matrix [8]

STF Encoder

Maximum

Likelihood

Detector

D/A IFFT Add

Prefix

D/A

A/D

A/D

FFT Remove

Prefix

FFT Remove

Prefix

IFFT Add

Prefix N

t N

r

exp(j2πfct) exp(j2πf

ct)

exp(j2πfct)

exp(j2πfct) .

.

.

.

.

.

1 1

Figure 1 MIMO–OFDM multiband UWB system

(3)


ISSN : 2250:3439



85

[

]T

where [

] for and

[

] , (for represents the complex symbol to

be transmitted over subcarrrier by transmit antenna during OFDM

symbol period. The baseband O F D M s i g n a l t o b e t r a n s mi t t e d b y the

antenna at the transmit antenna over OFDM symbol periods can be expressed

as [8]:

√ ∑

{ }

where and the factor √ guarantees the average transmitted

symbol is independent on the number of transmit antenna. Thus, the transmitted

multiband UWB at the transmit antenna OFDM symbol periods can expressed

as [8]

∑ {

}

where specifies the subband. The carrier frequency can be changed from one

OFDM block to another which enables f r eq uenc y d i ve r s i t y . is t h e same

for each transmit antenna and the transmission from all transmit antennas are

simultaneous and synchronous. Since bits are transmitted during seconds,

the transmit rate (without channel coding) is .

3.2 Frequency Selective MIMO Channel

The general structure of frequency- selective MIMO channel with signals

[ ] .From the input of our system at each time instant and thus obtain

output. Therefore, the output at time instant can be expressed as [9].

[ ] ∑ ∑

[ ] [ ]

[ ]

where denotes the largest number of taps among all the contributing channels.

In (7), the channel matrix has the form [9].

(4)

(5)

(6)

(7)


ISSN : 2250:3439



86

[ ] *

[ ] [ ]

[ ] [ ]

+

3.3 BER Analysis of MB-OFDM UWB in Fading Channels

Having derived the necessary term for the variance of the UWB interference signal,

we can express the bit error probability for the MB-OFDM system as [11]

(√

)

where is the channel impulse at time due to delta function [17] and

∫

√

is Gaussian Q- function or the complementary probability

distribution function for Gaussian distribution [13]. is the received power of the

OFDM system and where represent the noise spectral density. The

average of path equals and assuming that then

∫

K=

√ , then

∫ ∫

[ (

√ )

]

if

√

√

√ √

thus,

∫

√ ∫ √

[ ]

Using,

(8)

(9)

(11)

(12)

(13)

(14)


ISSN : 2250:3439



87

( )

√

therefore,

∫

√

∫

√ ∫

hence, for a given the total average power can

be expressed as[19]

∑

Therefore, for lognormal distribution the probability of error (BER) is expressed as

* (

)+ ( ∑

)

where

and is the variance of the Random Variable being

considered Similarly, for Nakagami distribution then probability is given as

indicated in [11].

* (

)+

(

)

∫

(

)

3.4 Receiver Processing

If coherent single-user matched filter is used where the receiver is assumed to know

the fading coefficients of the user of interest and the transmitted signal from each

antenna [10], then an antenna will receive

(15)

(16)

(17)

(18)


ISSN : 2250:3439



88

(t)

+ (t)

The Optimum decision rule selects that minimizes [16]

∫ ∑

∑ ∫ ∑

∑

This means that the optimum rule decision also known as maximal-ratio combining

for a single –user case is expressed as

( { ∑

})

where

∫

According to the optimum decision rule [10] the inner product of the and

is the sufficient statistic[14]. Therefore the probability of error of a MB-OFDM

UWB MIMO system could be expressed as [16]

[ (

∑

∑ (

)

)]

3.5 CHANNEL CAPACITY OF MIMO SYSTEMS

There are two major problems in broadband mobile communication i) frequency

selective fading due to multi-path a technology for achieving high frequency

(19)

(20)

(21)

(22)

(23)

(24)


ISSN : 2250:3439



89

utilization efficiency in terms of bits per Hertz (bit/s/Hz) or bits per/Hz/cell. As

mention earlier, Space Division multiplexing (SDM) over MIMO channels using

multiple transmitting and receiving antennas is one of the most promising

technologies for improving bits per Hertz (bit/s/Hz). The MIMO channel

capacity is given by [11].

* (

)+

As the parallel channel capacity, where I is by identity matrix, is a channel

matrix, and number of transmitting and receiving antennas, and denotes

the complex conjugate transpose. This equation indicates that the channel capacity

can be increased in proportion to the number of antennas if .

4 Result & Discussions

In this study, the size of FFT for the MB-OFDM analysis was taken as 128 without

error control coding. The modulations were by the following: MPSK and MQAM.

As for the TH and DS-UWB system, the frame duration and the hop width

were chosen are chosen to be 1ns an 0.0625ns respectively, The number of

hops equal 16. The analysis in AWGN indicated the average BER versus Eb/No

of MB-OFDM UWB in noise. 8PSK performs extremely well followed by

average performance by QPSK with BPSK and 16 QAM slightly below average,

thus BER performance of MPSK is improves when the alphabet size inreases, however, the performance improves with the MIMO technology in

AWGN. Similar performances were obtained in fading channels shown in Figure 2,

3 4 and 5. However, the performance in Nakagami-m fading channel outperforms the

Lognormal channel, interestingly, the performance improves with MIMO technology

in all cases. Looking at the results in multiple TH and DS UWB interferences shown

in Figures 2 and 3, UWB system performs well in multiple TH interferences as

compare to that of DS, moreover, the performance in Nakagami fading is better in

comparison to Lognormal Fading channel.

In multiple interference, the performance of UWB communication systems with the

same noise power in Nakagami-m fading channel outperforms that of Lognormal

fading channel in section 4.3. Figure 4.2a shows the performance of MB-OFDM

UWB system in multiple TH and DS interference and Nakagami-m fading channel,

interestingly the performance improves significantly as the MIMO technology is

introduce. Looking at figure 4.2c the performance of the MB-OFDM UWB (4*4)

MIMO system, the BER performance is approximately twice that of the MB-OFDM

UWB (2*2) MIMO system in figure 4.2b when the number of interferes is 5. It is

(25)


ISSN : 2250:3439



90

important to add that the performance of MB-OFDM UWB MIMO system with the

same transmitted power in 4.2b and 4.2c performs poorly as compare to that MB-

OFDM UWB system shown in figure 4.2a MB-OFDM UWB when the number of

interferes is 10, thus in MIMO technology is not suitable for UWB system especially

when the number of interferes is high. In addition, the BER performance of UWB

communication system is better in TH multiple interference than DS multiple

interference.

The results on the performance of UWB system in multiple TH and DS interference

in Lognormal fading channel. From figure 4.3a the performance in Lognormal fading

channel of MB-OFDM UWB system is such that at even 10 Eb/No BER virtually

remains constant and poor performance even by MB-OFDM UWB (2*2) MIMO

system. However, there is an improvement in performance as the number of antennas

in the MIMO system increases as shown in figure 4.3a. Here also the BER

performance of UWB communication system is better in TH multiple interference

than DS multiple interference from the results shown and it is important to stress here

that the performance of MB-OFDM UWB MIMO system with the same transmitted

power in 4.3b and 4.3c performs poorly as compare to that MB-OFDM UWB system

shown in figure 4.2a MB-OFDM UWB when the number of interferes is 10, also it

can be concluded that MIMO technology is not suitable for UWB system especially

when the number of interferes is high in Lognormal fading channel. However, in

figure 4.3c the performance of the MB-OFDM UWB (4*4) MIMO system, is

approximately twice that of the MB-OFDM UWB (2*2) MIMO system in figure 4.2b

whenthe number of interferes is 5.For the same performance as in Nakagami-m fading

channel the bit energy is increase from 1dB to 5dB in Lognormal channel.

Similar results in figure 4.1a was obtained by the author in [14] and the similar results

in, figure 4.2a is also obtained by Mehbodniya et al in [4]. The poor performance in

Lognormal fading channels buttress to point made by [6] that it is impossible to

predict the performance of UWB system in Lognormal fading channels. Lastly, the

performance of all the modulation schemes also support a point raised by Hu and the

group that DS sequence out performs TH in larger values of SNR.

5 Conclusion & Future Work

In conclusion, the performance of UWB communication system is improved by

using MIMO technology employing MPSK modulations with higher values of M

also the performance analysis indicates that UWB communications in Nakagami-

fading outperform that of Lognormal fading channel. . It is however important to

stress here that not only employing multiple input and multiple output (MIMO)

could improve the performance but also using specific error control coding could also

improve the performance tremendously.

Performance in Multiple Interference and in Nakagami-m Fading Channel


ISSN : 2250:3439



91

Performance in Multiple Interference and in Lognormal Fading Channel

Fig. 2 a MB-OFDM UWB Fig.2b MB-OFDM UWB MIMO (2*2) Fig.2 c MB-OFDM UWB MIMO (4*4)

Fig. 3 a MB-OFDM UWB Fig.3b MB-OFDM UWB MIMO (2*2) Fig.3 c MB-OFDM UWB MIMO (4*4)


ISSN : 2250:3439



92

Performance of Modulation Schemes in DS UWB Multiple Interference and

Nakagami-m Fadindg Channel

Performance of Modulation Schemes in DS UWB Multiple Interference and in

Lognormal Fading Channel

Fig.4 c MB-OFDM UWB MIMO (4*4) Fig.4 b MB-OFDM UWB MIMO (2*2) Fig.4 a MB-OFDM UWB


ISSN : 2250:3439



93

References

1 Yang, L., and Giannakis, G., B.,: ―Ultra–Wideband Communications: An

idea whose time has come,‖ IEEE Signal Processing Magazine, vol. 21, pp.26–54,

Nov. 2004.

2 Oppermann, M., Ha¨ma¨ la¨ inen, Iinatti, J.,: ―UWB Theory and

Applications,‖ John Wiley& Sons, 2004.

3 Ghavami, M., Michael, L., B., Kohno, R.,: ―Ultra Wideband Signals and

Systems in Communication Engineering,‖ Second Edition, John Wiley & Sons,

2007.

4 Mehbodniya Aissa, S.,: ―Ultra wideband Technologies Coexistence in

Nakagami-m FadingChannels,‖ IET Communications Journal, Manuscript ID: COM-

2008-0428, Issue 14, July, 2008.

Fig.5 c MB-OFDM UWB MIMO (4*4) Fig.5 b MB-OFDM UWB MIMO (2*2) Fig. 5a MB-OFDM UWB


ISSN : 2250:3439



94

5 Hoai Nam, V., Rajatheva, R., M., A., P.,: ―On the Performance Analysis of

MIMO-OFDM Multiband UWB System in a Lognormal fading channel,‖ IEEE

Conference Magazine, Issue 5, June 2006.

6 Shen, X., Guizani, M., Caiming, Q., Le-Ngoc,T.,: ―Ultra -Wideband

Wireless Communication and Networks,‖ John Wiley & Sons, 2006.

7 Arslan, H., Chen, Z., Benedetto, M., D.,: ―Ultra Wideband Wireless

Communication,‖ John Wiley & Sons, 2006.

8 Siriwongpairat, W., P., Su, W., Olfat, M., and Liu, K., J., R.,: ―Space-Time-

Frequency Coded Multiband UWB Communication Systems,‖ IEEE Wireless

Communication and Networking Conference, pp. 13–17, Mar. 2005.

9 Foerster, J.,: (editor), ―Channel modeling sub-committee report final,‖ IEEE

802.15-02/490 (see http://ieee802.org/15/).

10 Liu, H., Qiu, R., C., Tian, Z.,: ―Error performance of pulse-based ultra

wideband MIMO systems over indoor wireless channels,‖ IEEE Wireless

Communication conference. vol.4, pp. 2939–2944, Nov. 2005.

11 Cao, Z., Mwangoka, J.,W.,: ―Adaptive modulated MIMO-OFDM for ultra

wideband Channel,‖ Microwave, Antenna, Propagation and EMC Technologies for

Wireless Communications (MAPE), IEEE International Symposium, vol. 2, 8, Aug.

2005.

12 Chambers, P., Dowinig, C., and Hussein Baher,: ―BER Variation in an

UWB OFDM MIMO Communications System Based on Measurement made in a

Picocell Wireless Environment,‖ European Conference Antennas Propagation (

EuCAP), Nice, France, 6-10, Nov. 2006.

13 Ghavami, M., Michael, L., B., Kohno, R.,: ―Ultra Wideband Signals and

Systems in Communication Engineering,‖ John Wiley & Sons, 2004.

14 Mehbodniya Aissa, S., and Affes, A.,: ―Performance analysis of MB-OFDM

in the presence of UWB interferers,‖ 20th Canadian Conference on Electrical and

Computer Engineering (CCECE 2007), Vancouver, April 2007.

15 Miller, L., E.,: ―Validation of 802.11a/UWB Coexistence Simulation,‖

national institute of standards and technology, WCTG white paper, 2003.

16 Sergio V.,: ―Multiuser Detection‖ The Press Syndicate of the University of

Cambridge, Pit Building Trumpington Street, Cambridge, CB2, 1RP, United

Kingdom.

http://ieee802.org/15/


ISSN : 2250:3439



95

17 Proakis, G., : ―Digital Communication,‖ Fourth Edition, Irwin/McGraw-Hill,

an imprint of the Mcgraw-Hill Companies, Inc, New York, USA, 2000.

18 Khiong, S., Chond, C.,: ―An Overview of Multigigabit Wireless through

Millimeter Wave Technology: Potentials and Technical challenges,‖ EURASIP

Journal on Wireless Communication and Networking, article ID 78907, 14, June

2006.

19 Kuhn, V., : ―Wireless Communication over MIMO channels application to

CDMA and Multiple Antenna System‖ John Wiley & Sons, 2006.

20 Hu, B., and Beaulieu, C., : ―Accurate performance evaluation of time-

hopping and direct-sequence UWB systems in multi-user interference,‖ IEEE

Transactions

Analysis of Handoff Schemes in Wireless Mobile …...Alagu S#1 and Meyyappan T*2 #1Faculty,...

Documents

Transcript of Analysis of Handoff Schemes in Wireless Mobile …...Alagu S#1 and Meyyappan T*2 #1Faculty,...