Link Duration, Path Stability and Comparesion of MANET. Routing Protcols. Sanjay Kumar, Haresh Kumar and Zahid Yousif

Transcription

1 Link Duration, Path Stability and Comparesion of MANET Routing Protcols Sanjay Kumar, Haresh Kumar and Zahid Yousif A Bachelor thesis submitted to the Department of Electrical Engineering COMSATS Institute of Information Technology Islamabad In partial fulfillment of the requirements for the degree of Bachelor of Science in Electrical (TELECOM) Engineering Dr. Nadeem Javaid, Supervisor Department of Electrical Engineering COMSATS Institute of Information Technology ISLAMABAD June 2012 Copyright 2012 Sanjay Kumar, Haresh Kumar and Zahid Yousif All Rights Reserved

2

3 ABSTRACT We propose a framework for routing protocols in wireless multi-hop Networks (WMhNs). In this framework we model, simulate and compare impact of link availability probability, path stability and route stability. For this analysis, we select most widely used routing three reactive protocols; AODV, DSR, DYMO and three proactive protocols; DSDV, FSR and OLSR. Number of connections, number of nodes (node density) and relative velocity of nodes are the performance evaluation parameters. A novel contribution of this work is to improve the efficiency of protocols as MOD-AODV, MOD-DSR, MOD-DYMO, MOD-FSR and MOD-OLSR. The comparison-based on simulation of both default and modified routing protocols is carried out under the performance parameters; Packet Delivery Ratio (PDR), Average End-to-End Delay (AE2ED) and Normalized routing Overhead (NRO). Further we analyze the impact of both factors; routing protocols (default and modified) and node density in performance parameters using 2 k Factorial method that how they are affecting them with different variations. From extensive simulations, we observe that MOD- DSR, DSR, MOD-DYMO and DYMO outperform all the remaining routing protocols in terms of link availability probability, path duration and route stability at the cost of higher values of NRO. While analyzing with 2 k Factorial method, it is observed that MOD-DSR and DSR performing good enough by showing higher values in PDR, whereas low values in AE2ED at the cost of high value of NRO. For the simulation work we use Network Simulator NS-2 and Matlab to verify the results observed from NS-2. Keywords: VANETs, DSDV, DYMO and MOD-DYMO, OLSR and MOD-OLSR, Routing, PDR, Routing Load, Delay.

4

5 ACKNOWLEDGMENTS The project that we have done, it was not possible without the grace and kind of divine authority "ALLAH". First of all we would like to thank Dr. Nadeem Javaid, who encourages and supports us at every stage of project and without Sir, it was difficult to complete this project with such a great experience. We are very grateful to our Madam. Ayesha, who provides guidance in every task at which we stuck and she always treated us like younger brothers and helped in every kind of problem during this project. Last but not least, we are very very thankful to our families. They support morally and financially at every stage of life and we always get positive feelings from them for our studies.

6

7 Contents Table of Contents List of Figures vii ix 1 Introduction 1 2 DSDV, DYMO, OLSR: Link Duration and Path Stability Introduction Related Work and Motivation Related Work Motivation Link Duration and Path Stability Modeling Case-I Case-II Case-III Case-IV Experiments and Discussions PDR AE2ED NRO On Link Availability Probability Of Routing Protocols in VANETs Introduction Related Work and Motivation Related Work Motivation Link Availability Probability Experiments and Discussions PDR AE2ED NRO Performance Trade-offs made by Routing Protocols vii

8 viii CONTENTS 4 Evaluating Routing Protocols under Scalability Constraints with 2 k Factorial Analysis Introduction Related Work and Motivation Related Work Motivation k Factorial Analysis Calculating Effects of Factors through Variation Analyzing AODV Calculation of total Variation in AODV Analyzing OLSR Calculation of total Variation in OLSR Experiments and Discussions PDR AE2ED NRO Performance Trade-offs made by Routing Protocols Conclusion 53

9 List of Figures 2.1 Vehicles are moving with obtuse angles Moment of vehicles at different angles Vehicles are moving with obtuse and acute angles PDR vs Scalability Bar charts of PDR AE2ED vs Scalability Bar charts of AE2ED NRO vs Scalability Bar charts of NRO Flow Chart of four cases of velocities Probability of link availability at random variable Vr PDR vs Scalability AE2ED vs Scalability NRO vs Scalability Packet Delivery Ratio Packet Delivery Ratio Average E2E Delay ix

10 x LIST OF FIGURES 4.4 Average E2E Delay Normalized Routing Overhead Normalized Routing Overhead

11 Chapter 1 Introduction As vehicles on roads are increasing sharply in the recent years, so that driving has not stopped from being more challenging and dangerous. To make realize the vehicles what is happening around them, for reducing the number of road accidents and thereby increasing road safety this requires that vehicles should communicate with each other. It is of great importance that information transmitted should be sufficient for safety and without latency. The new advancement in wireless technologies and dedicated short range communication has made possible Inter Vehicular Communication (IVC) and Road-Vehicle Communication (RVC) in mobile ad hoc networks (MANETs). This has given birth a new type of subset of networks known as Vehicular ad hoc networks (VANETs). Vehicular Ad Hoc Networks (VANETs) bring a new concept of wireless ad hoc network environment in which the exchange of information between communicating vehicles without any fixed infrastructure like access points or base stations is an intensive field of research. As we considering the computer nodes as vehicles on the road that are in operation with wireless communication technologies and this results modernize traveling concept [1]. This new concept can be deployed in different applications such as active safety, traffic management and entertainment. VANETs are the special case of MANETs, in which mobile nodes are vehicles with radio com- 1

12 2 Chapter 1 Introduction munication range of 250 to 300 meters [2]. In VANETs, nodes have high mobility that causes fast change in the topology; therefore its link stability is low. Speeds of vehicles moving in same direction are similar most of the time; therefore they remain in radio contact for longer time than vehicles moving in opposite directions. So, path stability in VANETs depends on vehicle density and number of connection between these vehicles. When vehicle density and numbers of connections are less, then link breakage will be more and therefore, link stability decreases and routing overhead increases. As far as safety is concerned, VANETs are more appropriate and reliable for this purpose because they exhibit road accidents and traffic jams. In VANETs routing is done on the basis of routing protocols. The basic purpose of routing protocol is to transfer packets from one node to another node. These protocols find a route for correct delivery of packets from source to destination. Hence it is know that vehicles communicate with each other in VANETs by wireless routing protocols. There are two types of routing protocols depends upon the algorithms; Distance Vector and Link state Routing used. Commonly routing protocols are distributed in two types; reactive and proactive routing protocols. Three routing protocols; Ah-hoc On-demand Distance Vector (AODV) [3],[4], Dynamic Source Routing DSR [4],[5] and DYnamic Manet On-demand DYMO [4],[6] are multi hop on demand routing protocols. These protocols are bandwidth efficient and On-demand routing protocol for Vehicular Ad-Hoc Networks. The protocols are based on two main functions Route Discovery and Route Maintenance. Route Discovery function is responsible for the discovery of new route, when one is needed while Route Maintenance is responsible for the detection of link breaks and repair of an existing route. In this type of routing for route discovery the broadcast based method is used. In this method if the source wants to send the data to destination and it has no direct route to the destination then it broadcast the route request packets to its neighbors and these neighbors broadcast route request packets to their neighbors. This process will go on until the route to the destination is located. There are three main types of on-demand routing protocol AODV, DSR and DYMO. In

13 3 VANETs these protocols have great impact on scalability due to reactive nature because these protocols make route on demand. DYMO and DSR use source routing while AODV uses distance vector algorithm along with sequence number for reliable transmission. For path calculation, all three protocols use flooding based RD (Route Discovery). RD and RM (Route Maintenance) are the main features of these routing protocols. The main feature which distinguishes AODV from other reactive routing protocols is LLR (Local Link Repair). The Gratuitous Route Reply (grat. RREP) strategy is used for optimizations during RD process and is used by both AODV and DSR. Among reactive protocols, AODV performs better in wireless network of tens to thousands of nodes. DSR uses Route Cache (RC) and maintains multiple routes to destination entry. If primary route fails, source has alternate routes to a particular destination. Though DSR packets contain the complete source information, but Packet Salvaging (PS) technique makes it more scalable as compared to DYMO. In medium as well as in more data loads, DSR degrades the performance, because it does not have any explicit mechanism to delete the expired stale routes in RC, except those which are deleted by RERR messages. DYMO does not implement any mechanism except the basic Exponential Back o f f (EB) algorithm (used by AODV and DSR) to handle data traffic loads. DYMO uses HELLO messages to check connectivity and Source Routing SR dissemination, which results more bandwidth consumption in high scalabilities and high data traffic rate. So, DYMO degrades at medium and high traffic rates (as compared to AODV) due to absence of grat.rrers. Destination-Sequenced Distance Vector DSDV [4],[7], Fish-eye State Routing FSR [4],[8] and Optimized Link State Routing OLSR [4],[9]; all three routing protocols are table driven. In proactive routing protocols, the route is created on the basis of one or more routing tables, which contains routing information from source to destination. For packet forwarding, these protocols use hop-by-hop routing. Both DSDV and FSR uses Distributed Bellman Ford (DBF) algorithm for path calculations, while OLSR uses Di jkstras algorithm. DSDV generates periodic updates, P updates and trigger updates, T updates to maintain consistency in routing tables, when information

14 4 Chapter 1 Introduction about new links is available. OLSR uses Multi Point Relays (MPRs) redundancy mechanism to achieve route accuracy while FSR uses graded f requency (GF) mechanism to achieve route accuracy. DSDV uses two type of packets f ull dump:(carries all available routing information) and incremental:(carries only information changed since the last full dump) to reduce routing information carried by routing packets, whereas FSR uses (FS) Fish Eye State technique to reduce overhead, generated by exchanging whole routing table for convergence and GF mechanism is used to keep routing overhead low as network size grows. Whereas, MPR mechanism is used to reduce number of retransmission while forwarding broadcast packets.

15 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability 2.1 Introduction In recent years, Wireless Communication becomes so important, specially when there is high mobility. When we discuss about high mobility in ad hoc networks it means we are focussing on VANETs because it is a network with the high movement of vehicles. Where the routing protocols play an vital role in communication between these vehicles and here the performance of proactive and reactive routing protocols in accordance of performance parameters for a urban scenario in VANETs is discussed. Nakagami model is used for simulation work in NS-2 because in [10], authors conclude that Nakagami model experimentally performs well among the available propagation models. We simulate and analyze both default and modified routing protocols; DSDV, DYMO and OLSR under the performance parameters; PDR, AE2ED and NRO. Changes have been made in some parameters of routing protocols to modify their performance. In DYMO, NETWORK_DIAMET ER from 10 to 30 hops and Route Request wait time from 1000 to 600 5

16 6 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability ms. While, we change Topology Control Interval from 5 sec to 3 sec and Hello Message Interval to 1 sec which is originally 2 sec in OLSR. Through these modifications, comparison of routing protocols; DYMO and OLSR is carried out in accordance of performance parameters. 2.2 Related Work and Motivation Related Work Authors in [11], predict the link duration and stability of nodes in MANETs. They find link duration (for how much time link is available between nodes) of nodes and also calculated the mean duration. On the basis of link duration, they found link stability of nodes keeping one node at fixed position while other is moving with relative velocity. In [12], authors evaluate three routing protocols; selected from categories; geographic routing (i.e. GPSR), geographic opportunistic routing (i.e. GOSR), and trajectory based routing (i.e. SIFT) for VANETs in urban environments. In order to model realistic vehicular pattern, Vehicular Mobility Model (VMM) is used. They analyze routing protocols varying vehicle density and speed against the performance parameters; PDR, Packet Loss Ratio (PLR), Throuput, AE2ED, Average number of hops and control overhead. [13] presents the evaluation of IEEE p with IEEE a. Simulations are performed in NS-2 using both MAC protocols; a and p, then are compared for performance parameters; AE2ED, throughput, packet drops during various models. From the observed results, it is concluded that p performs better than a under the performance parameters. In [14], authors evaluate the routing protocols; AODV, OLSR and Modified OLSR in VANETs urban scenario. For Modified OLSR, they change existing routing parameters to get better results in terms of PDR, AE2ED and NRL varying node density and number of connections.

17 2.3 Link Duration and Path Stability Modeling Motivation Motivation is taken from [11],[12],[13],[14] as mentioned in related work and also from simulation results discussed in section IV. In this chapter we have done simulation in urban scenario that are previously select in [11],[12],[13],[14]. Authors in [11], find link duration D oa of link L oa between two nodes A and O from current time t to time at which L ao is broken, keeping one node at fixed position, while other is moving. In [12], routing protocols are evaluated in VANETs for urban scenario. Paper [13] evaluate and compare parameters; IEEE p and IEEE a under performance parameters. In [14], authors evaluate the routing protocols; AODV, OLSR and Modified OLSR in VANETs urban scenario. Inspired by [12] and [13], we take urban scenario for simulation work using IEEE p. The modifications in routing parameters in routing protocols is done as in [14]. We modeled link duration and path stability which is determined between vehicles like in [11]. We evaluate and compare the performance of both default and modified routing protocols; DSDV, DYMO and OLSR with four different cases of angles between vehicles in VANEts urban scenario. We consider urban scenario in VANETs, in which nodes (vehicles) are moving with different velocities. The links between vehicles are not available for longer time and the relative velocity (Difference of velocities between two vehicles and expressed as v r ) of vehicles is changed at different time instants. Path stability depends on available links between the vehicles, so their stability also decreases. Therefore, we find link duration and path stability between vehicles for four cases, discussed below. 2.3 Link Duration and Path Stability Modeling In [11], authors find link duration D oa of link L oa between two nodes A and O from current time t to time at which L ao is broken. They take two nodes as mobile, keeping one node at fixed position

18 8 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability while other is moving. Further they calculate mean link duration D oa based on distance d. We consider urban scenario in VANETs, in which nodes (vehicles) are moving with different velocities. The links between vehicles are not available for longer time and the relative velocity (Difference of velocities between two vehicles and expressed as v r ) of vehicles is changed at different time instants. Path stability depends on available links between the vehicles, so their stability also decreases. Therefore, we find link duration and path stability between vehicles for four cases, discussed below Case-I In case-1 let A and B are two vehicles. It is assumed that at time t 0 the distance between two vehicles A t0 and B t0 is d t0. At time t 1, A t0 and B t0 move with distances d 1 and d 2 making angles of α O and β O respectively, as shown in Fig Further we calculate the distances R 1 and R 2 between A t0 and B t1, and, B t0 and A t1, using cosine law and angles Ψ A and Ψ B using sine law respectively, given in eq. 2.1, 2.2, 2.3 and 2.4. Figure 2.1 Vehicles are moving with obtuse angles

19 2.3 Link Duration and Path Stability Modeling 9 R 1 = d t0 2 + d 2 2 2d t0 d 2 cos(β obtuse ) (2.1) Ψ A = arcsin( d 2 sin(β obtuse ) R 1 ) (2.2) R 2 = d t0 2 + d 1 2 2d t0 d 1 cos(α obtuse ) (2.3) Ψ B = arcsin( d 1 sin(α obtuse ) R 2 ) (2.4) Then we determine the distance d t1 between A t1 and B t1 at time (t 1 ) as d t1 = d R 2 1 2d 1 R 1 cos(α obtuse Ψ A ) (2.5) or d t1 = d R 2 2 2d 2 R 2 cos(β obtuse Ψ B ) (2.6) or d t1 = d t0 + d 1 cos(180 α obtuse ) +d 2 cos(180 β obtuse ) (2.7) Let r be the radio communication range of any node, therefore the distance must satisfy for communication d t1 r. So the distance d at time t 1 will be expressed as d = r d t1 (2.8) It is clear that link availability between vehicles depends on two parameters; distance (d t1 ) and relative velocity v r, therefore in order to find link duration LD AB for node A and B, we derive an expression as: LD AB = d v r (2.9) If the link duration increases, path stability becomes high.

20 10 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability Case-II In the first case, vehicles are moving with making obtuse angles (greater than 90 ). Here the case is different due to the movement of vehicles with acute angles (less than 90 ). For this case, equations that we drove for case-1 will be same with little change of angles α O as α A and β O as β A and also the angles Ψ A and Ψ B will change due acute angle, as shown in Fig. 2.2.(a) below and also in eq. 2.2 and 2.4. Then, the distances R 1 and R 2 and the angles Ψ A and Ψ B are calculated same like case-1, to find the distance d t1 between A t1 and B t1 at time t 1. After that, link duration and path stability between vehicles and distance d t1 in eq are determined. (a) Vehicles are moving with acute angles (b) Vehicles are moving with both acute and obtuse angles Figure 2.2 Moment of vehicles at different angles d t1 = d t0 + R 1 cos(180 Ψ A ) + R 2 cos(180 Ψ B ) (2.10) Case-III In this scenario, we assume that one vehicle is moving with distance d 1 by making an acute angle α A, while other is moving with distance d 2 making an obtuse angle (β O ). Where the angles Ψ A and Ψ B depend on angles (α A ) and (β O ) respectively, as shown in Fig. 2.2.(b) and also in eq. 2.2

21 2.4 Experiments and Discussions 11 and 2.4. Now, to calculate the distance d t1 between A t1 and B t1 at time t 1, we use same equation of case-1 and also by using eq Eq. 2.9 tell us about the remaining time of link between nodes. d t1 = d t0 d 1 cos(α acute ) + d 2 cos(180 β obtuse ) (2.11) Case-IV This case is same like case-3 but we change the angles α A as α O and β O as β A, so the angles Ψ A and Ψ B will also be changed, as shown in Fig Then R 1 and R 2, and Ψ A and Ψ B, are dtermind using eq. 2.1, 2.2, 2.3 and 2.4, to calculate the distance d 1 between vehicles at time t 1 and also using eq for d 1. Further, we find its link duration and path stability using eq Figure 2.3 Vehicles are moving with obtuse and acute angles d t1 = d t0 + d 1 cos(180 α obtuse ) d 2 cos(β acute ) (2.12) 2.4 Experiments and Discussions In this chapter, Nakagami propagation model in NS is used. The implementation of DSDV used is by default in NS-2. For implementation of DYMO and OLSR, DYMOUM [10] and OLSR [11] patchs are used. The map imported in MOVE and scaled down to 4 km x 4 km in size for

22 12 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability reasonable simulation environment. Using MOVE and SUMO, mobility patterns are generated randomly. Table 2.1 shows the complete simulation parameters used in simulation. Table 2.1 SIMULATION PARAMETERS Parameters Values NS-2 Version 2.34 DSDV Implementation NS-2 default DYMO Implementation DYMOUM-patch [15] OLSR Implementation OLSR-patch [16] MOVE version 2.81 SUMO version Number of nodes 30, 50, 70, 90, 120 Number of CBR sessions 6, 12, 18, 24 Tx Range Simulation Area Speed Data Type Data Packet Size MAC Protocol 300m 4KM x 4KM Uniform, 40kph CBR 1000 bytes IEEE Overhauled PHY Standard IEEE p Radio Propagation Model Nakagami The following performance parameters are used to evaluate the performance of routing protocols; DSDV, DYMO, and OLSR. - PDR: The ratio of data packets at the destination and total data packets generated.

23 2.4 Experiments and Discussions 13 - AE2ED: Overall Delay of packet generation at the source and arrival at destination. - NRO: The number of routing packets transmitted per data packet delivered to the destination PDR Fig. 2.4.(a) shows PDR against number of connections. From Fig. 2.4.(a), it is clear that MOD- DYMO attains more PDR than other routing protocols; DSDV, DYMO, MOD-OLSR and OLSR, due to reactive in nature because reactive protocols do not need route calculation before data transmission. So, as number of connections increases, it attains higher PDR than other protocols. While DYMO is showing second highest value in PDR but as the number of connections increase, its PDR goes down. Because in MOD-DYMO NETWORK_DIAMET ER of 30 hops in large area of 4 KM x 4 KM results better scalability as compared to DYMO, in which NETWORK_DIAMET ER is too small (10 hops). In low number of connections, DSDV attains high PDR than MOD-OLSR and OLSR, due to generation of less trigger updates, as trigger updates are generated in response to breakage among active routes. Less number of connections means less active routes. Whereas, for broadcasting DSDV uses simple flooding therefore, more connections introduce more overhead. In high number of connections, there is occurrence of more trigger updates, will cause more routing overhead and PDR decreases. While the MOD-OLSR and OLSR show increasing graph, The main reason of increasing PDR is that the computation of Multipoint Relay (MPRs) mechanism generates more routing packets so due to this, its PDR goes up as increase in number of connections. In Fig. 2.4.(b), we simulate the PDR against node density. In Fig. 2.4.(b), MOD-DYMO and DYMO sustain higher PDR than DSDV, MOD-OLSR and OLSR. The main reason of high PDR value is due to its reactive nature, because reactive protocols do not require more computation for route discovery. That is why MOD-DYMO performs well than all other routing protocols. Whereas DSDV attains high value in low scalability and in medium scalability it comes down.

24 14 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability DSDV DYMO MOD DYMO OLSR MOD OLSR DSDV DYMO MOD DYMO OLSR MOD OLSR PDR(%) PDR(%) Number of Connections Number of Nodes (a) PDR vs No. of connections (b) PDR vs Node Density Figure 2.4 PDR vs Scalability MOD-OLSR and OLSR in low scalability show less value due to low optimization but as node density increases its PDR value also increases because high optimization of MPRs % 2.94% Default Modified % 3.54% Default Modified PDR(%) PDR(%) DYMO OLSR 0 DYMO OLSR (a) Average PDR vs No. of connections (b) Average PDR vs Node Density Figure 2.5 Bar charts of PDR In Fig. 2.5, we observe that modified routing protocols perform better than default one. MOD- DYMO outperforms DYMO due to decrease in Route Request wait time from 1000 to 600 seconds and NETWORK_DIAMET ER from 10 to 30 hops. Whereas, MOD-OLSR shows good results than OLSR due to decrement in intervals of updates for link sensing as well as for route updations.

25 2.4 Experiments and Discussions 15 Link duration and path stability in MOD-DYMO and DYMO is greater than DSDV, MOD-OLSR and OLSR because of high value of PDR. The main reason is less drop of packets causes more PDR and the link duration in eq. 9. DSDV, MOD-OLSR and OLSR also have good value of link duration and path stability but not as good as DYMO has AE2ED Fig. 2.6.(a) shows AE2ED against number of connections. OLSR and MOD-OLSR show highest value of AE2ED than other routing protocols; DSDV, DYMO and MOD-DYMO. This includes two main reasons; firstly, proactive routing protocols have more AE2ED because before data transmission, they need to calculate routing tables and secondly, generation of Hello and TC messages for checking the link and computing MPRs that causes more delay, therefore MOD-OLSR has less value than OLSR due to decrease in Hello and TC message intervals. DSDV attains high value than MOD-DYMO and DYMO, but less than MOD-OLSR and OLSR. DSDV has two main reasons for its high value, first proactive nature and second, the selection of best routes involves route settling time, creates delay in advertising routes. DYMO has less AE2ED than DSDV, MOD- OLSR and OLSR, because it uses Expanding Ring Search (ERS) algorithm that reduces AE2ED. While MOD-DYMO performs better than DYMO due to decrease in Route Request wait time. In Fig. 2.6.(b), we simulate the AE2ED against node density. MOD-OLSR and OLSR has less delay because there are generation of Hello and TC messages for the link sensing and computing MPRs that causes reduction in delay. In low scalability, MOD-OLSR shows high value of AE2ED than OLSR because of decrement in Hello and TC message intervals. Overall, AE2ED of DSDV is very high, while in medium scalability, it decreases. MOD-DYMO and DYMO show less and almost decrease in AE2ED. In medium scalability, DYMO and OLSR perform well by showing same delay. In Fig. 2.7, DYMO outperforms MOD-DYMO due to decrease in Route Request wait time

26 16 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability DSDV DYMO MOD DYMO OLSR MOD OLSR DSDV DYMO MOD DYMO OLSR MOD OLSR AE2ED(s) AE2ED(s) Number of Connections Number of Nodes (a) AE2ED vs No. of connections (b) AE2ED vs Node Density Figure 2.6 AE2ED vs Scalability Default Modified Default Modified 25.66% E2ED (s) E2ED (s) % % 67.08% 0 DYMO OLSR 0 DYMO OLSR (a) AE2ED vs No. of connections (b) AE2ED vs Node Density Figure 2.7 Bar charts of AE2ED from 1000 to 600 seconds, that causes to decrease in delay. Whereas, MOD-OLSR shows less value of AE2ED than OLSR due to decrement in intervals of updates; TC and Hello Messages. MOD-DYMO and DYMO sustain less value for AE2ED, due to showing best value of link duration and path stability. While, DSDV, MOD-OLSR and OLSR doe not have good value of link duration and path stability like DYMO, because proactive routing protocols have different link sensing updates.

27 2.4 Experiments and Discussions NRO From Fig. 2.8.(a), it is observed that NRO of DYMO and MOD-DYMO is larger than both proactive routing protocols; DSDV, MOD-OLSR and OLSR. The reason is that DYMO is reactive in nature; ERS algorithm is efficient for less delay and high NRO than rest of routing protocols. MOD-OLSR and OLSR has higher value of NRO than DSDV due to one reason that there is generation of Hello and TC messages for checking the link and computing MPRs at routing layer increase in NRO. Whereas, OLSR has less value of NRO than MOD-OLSR due to high interval of updations intervals. DSDV generates less NRO but in higher scalability DSDV sustains more NRO. The reason is that in high scalability, DSDV generates more trigger updates for updating active routes NRO DSDV DYMO MOD DYMO OLSR MOD OLSR DSDV DYMO MOD DYMO OLSR MOD OLSR NRO 3 NRO Number of Connections Number of Nodes (a) NRO vs No. of connections (b) NRO vs Node Density Figure 2.8 NRO vs Scalability In Fig. 2.8.(b), MOD-OLSR has highest value of NRO but as node density increases its NRO increases with rising slope, due to more trigger updates. While OLSR sustains less value of NRO than MOD-OLSR and greater than DSDV and DYMO, due to short intervals of trigger updates; Hello and TC messages. DYMO attains high value of NRO but not greater than OLSR, due to on-demand route discovery. DSDV has less value of NRO but as node density increases, due to

28 18 Chapter 2 DSDV, DYMO, OLSR: Link Duration and Path Stability generation of more periodic and trigger updates. 8 7 Default Modified 8 7 Default Modified % NRO % 48.43% NRO % DYMO OLSR 0 DYMO OLSR (a) Average NRO vs No. of connections (b) Average NRO vs Node Density Figure 2.9 Bar charts of NRO In Fig. 2.9, we observe that default routing protocols sustain less value of NRO than modified. MOD-DYMO outperforms DYMO due to decrease in Route Request wait time from 1000 to 600 seconds and network diameter from 10 to 30 hops. Whereas, OLSR shows good results than MOD-OLSR due to decrement in intervals of TC and Hello Message. DYMO and MOD-DYMO sustain link duration for longer time and good value of path stability, that will cause high value of NRO. OLSR and MOD-OLSR have better link duration than DSDV but not than DYMO and MOD-DYMO due to its MPRs mechanism.

29 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs 3.1 Introduction Evaluation of the performance of three routing protocols; AODV, DSR and FSR for VANETs in urban scenario. The evaluation of both default and modified routing protocols has done with performance parameter; PDR, AE2ED and NRO with varying node densities and different number of connections. In case of urban scenario for communication, vehicles can move in any direction. In other words, vehicles can move in both directions (same and opposite). So, we find out link availability probability between nodes with different cases that is described in link availability probability section. 19

30 20 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs 3.2 Related Work and Motivation Related Work In last few years, efforts are made on work that relate to our work. In [17], authors evaluate radio propagation models. Through the study, it is observed that Nakagami model is able to give accurate simulation results for VANETs in urban scenario. Studying in [18], authors improve the routing performance of routing protocols in high mobility and high density VANETs. They propose two different cases of vehicles velocity and find the link available probability for these two cases. For further improvements in PDR and AE2ED, they present two algorithms. [19] discusses the comparison and the performance of routing protocols; AODV, DSDV and DSR in VANETs, with the extensive simulation studies for highway scenarios. Simulation results show the performance parameters; PDR, AE2ED and NRL, and concluded that the routing protocols are unsuitable for VANETs, but dedicated for MANETs. Another study in [20], evaluates the performance of routing protocols; AODV, OLSR and Modified OLSR, under realistic radio channel characteristic using NS-2 Nakagami fading model. They analyze the performance of both routing protocols with performance parameters; PDR, AE2ED and NRL with varying node densities and different number of connections Motivation In this chapter we have done simulation in urban scenario that was due to motivated by [17],[18],[20]. In [17], they have done evaluation of protocols through radio propagation model Nakagami by the knowledge as the best model among all these models. Authors in [18] have analyzed the problem for two different cases of velocity of the nodes and also proposed two algorithm for improving routing performance in high scalability and high mobility. In [20], authors evaluate the

31 3.3 Link Availability Probability 21 routing protocols; AODV, OLSR and Modified OLSR in VANETs urban scenario. Hence in [17], we study that Nakagami model is best among all the radio propagation models so for our simulation work, we used this model. As inspired by [5] and [20], we evaluate and compare the performance of both default and modified routing protocols; AODV, DSR and FSR in urban scenario with more different cases of velocity between nodes. Further we also determine the link availability probability between nodes. 3.3 Link Availability Probability In [18], authors discuss the mobility scenario in VANETs taking two cases of velocity; (i) when both nodes have same velocity and (ii) when both have different velocities. After taking assumption they find out the relative velocity and expected relative velocity between nodes. They determine the link availability probability for these two cases using simple area of covered region between two nodes and radio covering range of any node. We consider an urban scenario in VANETs in which nodes (vehicles) can move in both the directions (same and opposite). Assumptions are taken that two nodes are moving with velocities v 1 and v 2 respectively, the distance between two nodes is d and the radio communication range of a node is expressed as r. These nodes can communicate only when d r. Now we consider the four different cases in the velocities of these moving nodes and for each case link availability probability between nodes is discussed, these cases holds when d r : Case-1: When both the nodes have same velocity and moving in same direction then link is available for longtime t 1 between them. Case-2: In this case any of node has greater velocity than other one and moving in same direction then link is broken after some time t 2. Case-3: This case deals that nodes have same velocity but moving in opposite direction then link

32 22 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs is broken after some time t 3 but t 3 < t 2. Case-4: Here the case is different because both the nodes are moving in opposite direction with different velocities so link is broken after some time t 4 i.e: t 4 << t 2. We discuss four cases above with velocities v 1,v 2 [V min,v max ] and θ 1,2 [0,π] and now we will see the relative velocity between these nodes as: v r = v 1 v 2 (3.1) and using cosine law we can write it as v r = v v 2 2 2v 1 v 2 cosθ 1,2 (3.2) Hence relative velocity for four cases will be. Case-1: v 1 = v 2 and angle θ 1,2 = 0 then v r = 0. Case-2: v 1 = a v 2, where a (1,3) and angle θ 1,2 = 0 then v r = (a 1)v 2. Case-3: v 1 = v 2 and angle θ 1,2 = π then v r = 2v 2. Case-4: v 1 = a v 2, where a > 1 and angle θ 1,2 = π then v r = (a + 1)v 2. For the above cases a flow chart is given below in Fig This flow chart shows relative velocity v r for each of the case and also the link availability time t n between vehicles, where n=1 to 4. From above results, it is observed that v r has different values so we represent it as random variable and according to probability density function (pdf), we can find it s expected relative velocity function as: E(v r ) = v r f (v r )dv r (3.3) for further simplification f (v r ) = f (v 1 ) f (v 2 ) f (θ 1,2 ), we can write eq. 3 as: Vmax Vmax π E(v r ) = f (v 1 ) f (v 2 ) f (θ 1,2 ) V min V min 0 (3.4) v v 2 2 2v 1 v 2 cosθ 1,2 dθ 1,2 dv 2 dv 1

33 3.3 Link Availability Probability 23 Figure 3.1 Flow Chart of four cases of velocities Then expected relative velocity for each of the case discussed below. Case-1: Nodes have same velocity and moving in same direction. E(v r ) = Vmax Vmax V min V min f (v 1 ) f (v 2 )dv 2 dv 1 (3.5)

34 24 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs Case-2: Nodes have different velocity and moving in same direction. E(v r ) = Vmax Vmax V min (a 1)v 2 f (v 1 ) f (v 2 )dv 2 dv 1 V min a (1,3) (3.6) Case-3: Nodes have same velocity and moving in opposite direction. E(v r ) = Vmax Vmax V min V min 2v 2 f (v 1 ) f (v 2 )dv 2 dv 1 (3.7) Case-4: Nodes have different velocity and moving in opposite direction. E(v r ) = Vmax Vmax V min (a + 1)v 2 f (v 1 ) f (v 2 )dv 2 dv 1 V min a > 1 (3.8) The expected relative velocity is function of random variable v r, so this function shows random behavior for random values of v r. Using this expected relative velocity function as random variable to find the probability of link availability, we develop an exponential probability density function given below f (E(v r )) = 1 dt n exp( E(v r) dt n ) where n = 1 to 4 (3.9) This pdf demonstrates the probability of available link between nodes. The function is exponential decaying function, so the term E(v r) dt n in eq. 3.9 increases, pdf decreases that means probability of link availability becomes small. Now we discuss for each of the case that how the pdf varies with varying expected relative velocity v r and link availability time t n. Case-1: From above explanation, it is observed that when nodes have expected relative velocity at v r = 0 and the link availability time is t 1 that is large as compared to other cases. The term

35 3.3 Link Availability Probability 25 E(v r ) dt 1 will have less value for large value of pdf and it illustrates that link availability probability for case-1 is greater than all the cases. Case-2: In this case expected relative velocity has a value for v r between nodes is (a 1)v 2 a (1,3) and the link availability time is t 2 i.e: t 2 < t 1. Now the term E(v r) dt 2 case so the probability for link availability is less than case-1. has more value than previous Case-3: This case is different due to change in direction therefore, its expected relative velocity at v r = 2v 2 is in eq. 3.7 and link availability time is t 3 i.e: t 3 < t 2. Then link availability probability of nodes is smaller than case-2 because the term E(v r) dt 3 relative velocity and decrease in link available time t 3. has large value due to increase in expected Case-4: In the last case the scenario is totaly different because in this case assumption is taken that nodes have different velocity and moving in opposite direction. Its expected relative velocity at v r = (a+1)v 2 a > 1 is in eq. 3.8 and the link availability time is t 4 i.e: t 4 < t 3. Now this time the term E(v r) dt 4 is very very large because in this case v r has large value and t 4 has less value compare to all the cases. Therefore link availability probability between nodes will become so small from all the cases. 1 Probability density function (pdf) d=5m d=10m d=20m d=30m d=40m d=50m d=100m d=150m d=200m d=250m Expected related velocity Vr (m/s) Figure 3.2 Probability of link availability at random variable Vr In order to analyze the cases that are discussed above, we create graph for eq From Fig.

36 26 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs 3.2, we can say that link availability probability of two nodes decreases as expected related velocity increases. Another interesting thing is the distance (d < r) as it increases between nodes its graph shifts to upward due to increase in link availability probability. 3.4 Experiments and Discussions In this chapter, we use Nakagami propagation model in NS The implementation of AODV and DSR used is the default one that comes with NS-2. For implementation of FSR, FSR patch is used [14]. The map imported in MOVE and scaled down to 4 km x 4 km in size for reasonable simulation environment. Using MOVE and SUMO, mobility patterns are generated randomly. Table 3.1 shows the complete simulation parameters used in simulations PDR Fig. 3.3 shows the average percentage of PDR against different number of connections and node density. In Fig. 3.3.(a), it is clear that AODV and DSR outperform FSR. AODV and DSR give almost same result and perform better than FSR. The PDR of AODV increases as increase in the number of connections. Because it uses shortest path to destination and local link repair (LLR) mechanism, when link breakage occurs and consumes less bandwidth. The PDR of DSR also goes up as increment in the number of connections. Route reversal keeps away from the overhead of a possible second route discovery. FSR has low value of PDR due to proactive in nature because proactive routing protocols require more computation than reactive routing protocols. That is why, FSR shows average value of PDR as increment in number of connection. Fig. 3.3.(b) shows the PDR values of three routing protocols against node density. DSR gives slightly better PDR than AODV for high node density because valid routes are available in route cache, so it does not need more computation. Where as AODV uses the distance vector algorithm,

37 3.4 Experiments and Discussions 27 Table 3.1 SIMULATION PARAMETERS Parameters Values NS-2 Version 2.34 AODV Implementation DSR Implementation NS-2 default NS-2 default FSR Implementation FSR-patch [21] MOVE version 2.81 SUMO version Number of nodes 20, 40, 60, 80, 100 Number of CBR sessions 6, 12, 18, 24, 30, 36, 42 Tx Range Simulation Area Speed Data Type Data Packet Size MAC Protocol 300m 4KM x 4KM Uniform, 40kph CBR 1000 bytes IEEE Overhauled PHY Standard IEEE p Radio Propagation Model Nakagami so for every time using this algorithm, it finds the routes to destination. In node density, FSR performs same as for number of connections because it uses scope mechanism, due to this mechanism it shows average value for all the number of connections and node density. Modified routing protocols perform better than default routing protocols attaining highest PDR. MOD-AODV performs better than AODV due to increment in T T L START and T T L T HRESHOLD,

38 28 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs PDR(%) AODV MOD AODV DSR MOD DSR FSR MOD FSR PDR(%) AODV MOD AODV DSR MOD DSR FSR MOD FSR Number of Connections Number of Nodes (a) PDR vs No. of connections (b) PDR vs Node Density Figure 3.3 PDR vs Scalability while MOD-DSR shows good results than DSR, due to increment in buffer size. MOD-FSR has slightly better results than FSR due to decreasing the interval of both InterScope and IntraScope. From the results it is observed that link availability probability of DSR and AODV is higher than proactive routing protocol FSR because DSR has available routes in route cache, while AODV uses Hello messages for link sensing. Due to that reasons, their link availability time increases then the term E(v r) dt n have less value therefore eq. 9 results greater value of link availability probability. Whereas FSR has average value of PDR due to use of scope mechanism that causes less link availability time between nodes so its link availability probability is less than both reactive routing protocols AE2ED Fig. 3.4 shows the AE2ED against different number of connections and node density. From Fig. 3.4.(a), AE2ED of AODV is steady, but it is always more than both DSR and FSR. Due to increase in CBR sources, there is an increase in the number of packets contending for a common wireless channel, which leads to more collisions and more consumptions of bandwidth. So there is a significant drop in the delivery ratio and a corresponding increase in the AE2ED. The flow

39 3.4 Experiments and Discussions 29 of AODV AE2ED that as number of connections increase, the delay decreases due to routing packets used in this protocol for establishment of path. The AE2ED flow of DSR is less than both routing protocols; AODV and FSR. But its delay initially decreases as connection increase (6 to 12) because it has available paths in route cache, if link breakage occurs, it checks route cache and uses another path. But its delay increases (18 to 42) as connections increase due to not getting available path in route cache, whenever link breakage occurs and broadcasts message RREQ for establishment of path. The AE2ED flow of FSR increase (12 to 24) as connections increase due to link breakage occurs because of dynamic movement in topology and source has to broadcast information to its neighbors, to spread the information to whole topology for establishment of path as a proactive routing protocol. AE2ED(s) AODV MOD AODV DSR MOD DSR FSR MOD FSR AE2ED(s) AODV MOD AODV DSR MOD DSR FSR MOD FSR Number of Connections Number of Nodes (a) AE2ED vs No. of connections (b) AE2ED vs Node Density Figure 3.4 AE2ED vs Scalability As it is clear from Fig. 3.4.(b), the AE2ED flow of AODV routing protocol is increasing as nodes increase due to two reasons. Firstly, the protocol uses LLR mechanism, to repair link breakage, which causes increase in path length. Secondly, the protocol uses message types, Route Requests (RREQs), Route Replies (RREPs), and Route Errors (RERRs) for establishment of Path which causes much delay. The flow decreases when there are less number of connections, because data packets transmit and then again for path establishment RREQ is broadcasted. AE2ED of DSR

40 30 Chapter 3 On Link Availability Probability Of Routing Protocols in VANETs also increases due to message types but less than AODV s delay because it checks all available routes in Route cache whenever link breakage occurs. The AE2ED of FSR is more than DSR because, whenever link breakage occurs the source has to broadcast information to its neighbors to spread the routing information to whole topology. All paths are established, the data packets are transmitted to specified destination. Delay is marginally constant after path is established. Modified routing protocols perform better than default routing protocols showing lowest value of AE2ED. MOD-AODV performs better than default one due to decrement in network Diameter. Whereas MOD-DSR shows good results than DSR due to decrement in cache size size. MOD-FSR has slightly better results than default FSR due to decreasing the interval of both inner and outer scope. Through the results of AE2ED, we can conclude that AE2ED delay of AODV is larger than both the routing protocols; DSR and FSR due to use of more message types and link sensing before transmission. If AE2ED of any of the routing protocols increases, the link between nodes is not available for longer time due to high mobility in VANETs. So we can say that link availability probability of DSR and FSR is better NRO Fig. 3.5 shows the NRO of routing protocols against number of connections and node density. In Fig. 3.5.(a), it is observed that NRO of FSR is less than both reactive routing protocols; AODV and DSR. As the number of connections increase, NRO of FSR decreases due to use of scope mechanism, that is good for more number of connection. FSR has lowest value among these routing protocols due to use of periodic updates to exchange topology map and also reducing the control messages. AODV shows the highest NRO value in the Fig. 5.(a) increasing graph for NRO. The reason for highest NRO of AODV is use of large number of control packets. DSR shows average behavior for NRO, as number of connections increase, due to stale entries in it s