Optimal Configuration of OLSR Routing Protocol for VANETs by ...

Optimal Configuration of OLSR Routing Protocol for
VANETs by Means of Differential Evolution
J. Toutouh, J. García-Nieto, and E. Alba
LCC, ETSI Informática, Louis Pasteur 35, University of Málaga, 29071
{jamal,jnieto,eat}@lcc.uma.es
1 Introduction
Vehicular Ad Hoc Networks (VANETs) are self-configuring networks composed of a collection of
vehicles and elements of roadside infrastructure connected with each other without requiring an
underlying infrastructure. Currently, WiFi (IEEE 802.11 based) technologies are used to deploy
such networks. The coverage limitations of WiFi technologies and the high mobility of VANET
nodes generate frequent topology changes and network fragmentation. For these reasons, and without
any central manager entity, the routing task is a challenging work. Thus, offering an efficient
routing strategy is crucial to deploy VANETs in order to offer as high as possible QoS.
A way to obtain a suitable routing protocol for VANETs is to find an optimal configuration
for an existent one. The huge number of possible configurations practically prevents obtaining an
efficient protocol configuration without using automatic intelligent design tools. This motivates the
use of metaheuristic techniques as well-suited tools to solve such kind of problems [3]. Unfortunately,
only a few related approaches can be found in the specialized literature.
OLSR (Optimized Link State Routing Protocol) [2] is a well-known routing protocol defined
specifically for MANETs/VANETs with low bandwidth and high mobility. In the present work, we
propose the use of a metaheuristic technique, the Differential Evolution (DE), for the optimal configuration
of the OLSR protocol. Our goal is to automatically improve its performance overcoming
if possible the decision of both, experts and standard (RFC 3626) [2] configurations.
2 Optimizing OLSR by using DE
The optimization strategy used to obtain a set of efficient OLSR
parameters consists of coupling two different functional blocks (see
Fig. 1): an optimization algorithm and a simulation procedure. The
optimization task is carried out by a metaheuristic method (DE).
The simulation procedure is used to assign a quality value (fitness)
to the OLSR performance of the computed configurations during
the evolutionary process. This procedure is carried out using the Fig. 1. Optimization framework.
popular network simulator ns-2 [1].
In spite of being OLSR specifically designed for MANETs, the
application of its parameter configuration in VANET scenarios usually shoes a limited level of
QoS [5]. Taking into account the impact of the configuration parameters in the network behavior,
we can offer an optimized OLSR configuration to deploy VANETs.
Table 1 shows these OLSR parameters with their standard values as specified in RFC 3626.
The range of values each parameter can take has been defined here by following OLSR restrictions,
with the aim of avoiding pointless configurations. According to this table, we can use these OLSR
parameters to define a real vector solution and a search space. This way, the vector solution can
be fine-tuned automatically by means of the DE algorithm.
Table 1. Main OLSR Parameters. Default values following the RFC 3626 specification.
Parameter Standard Configuration Range
HELLO INTERVAL 2.0 s 1 · · · 30
REFRESH INTERVAL 2.0 s 1 · · · 30
TC INTERVAL 5.0 s 1 · · · 30
WILLINGNESS 3 {0, 1, 3, 6, 7}
NEIGHB HOLD TIME 3 × HELLO INT ERV AL 3 · · · 100
TOP HOLD TIME 3 × T C INT ERV AL 3 · · · 100
MID HOLD TIME 3 × T C INT ERV AL 3 · · · 100
DUP HOLD TIME 30.0 s 3 · · · 100

2 J. Toutouh, J. García-Nieto and E. Alba
In order to evaluate the performance of the different OLSR configurations (DE solutions),
we have measured the resulted QoS indicators of the network by means of three commonly used
metrics in this area: Packet Delivery Ratio (PDR), Normalized Routing Load (NRL), and Average
End-to-End Delay (E2ED) of a data packet. These metrics are returned by ns-2 after simulating a
VANET scenario consisting in a urban area of 4Km 2 from downtown of the Spanish city of Málaga
(taking into account road directions, signal lights, and traffic rules). In this VANET scenario, 50
vehicles exchange information with each other during 300 seconds. After each simulation, ns-2
returns the three QoS indicators and this information is used to compute the fitness function:
fitness = w 1 · (−P DR) + w 2 · NRL + w 3 · E2ED · C (1)
The objective here consists in maximizing PDR, and minimizing both, NRL and E2ED. As
expressed in Equation 1, we used an aggregative minimizing function, and for this reason PDR
was formulated with a negative sign. Factors w 1 , w 2 , and w 3 (0.8, 0.1, and 0.1, respectively)
were empirically assessed for weighing the influence of each metric on the fitness value. Constant
C = 0.01 set the E2ED with the same range to PDR and NRL factors.
3 Results
The main results obtained after experimentations are presented in Table 2. They consist on: results
obtained by three configurations (#1, #2 and #3) reported by experts Gómez [4], results of
using the default parameters from RFC 3626 configuration, and results of the best configuration
optimized by our DE (average best out of 30 independent runs with the Málaga city scenario).
Table 2. OLSR configurations obtained by DE, RFC 3626 definition, and experts (Gómez et al. [4])
Configuration Fitness PDR NRL E2ED
#1 46.19 90.00% 1170.02 kbps 1197.25 ms
Gómez [4] #2 -15.31 90.00% 554.75 kbps 1208.91 ms
#3 -29.47 66.00% 208.84 kbps 2435.22 ms
RFC 3626 61.22 80.00% 328.42 kbps 1347.22 ms
DE Best -68.35 94.00% 68.34 kbps 8.36 ms
As we can observe, the OLSR configuration optimized by DE shows the best performance in
terms of both, the fitness value and the QoS indicators. We can notice that DE configuration
obtained better PDR than standard RFC and Gómez et al. [4] configurations (PDR of 90% in #1
and #2, 80% in RFC, and 94% in DE).
However, it is in the network load (NRL) and in the average delay (E2ED) where optimized
configurations by DE reached the highest improvement, being a reduction of 67.28% (from 328.42
kbps to 68.34 kbps) concerning NRL and 99.66% (from 1347.22 ms to 8.36 ms) in E2ED over #3,
the best configuration in terms of fitness of Gómez et al. [4].
In the light of these preliminary results, we can conclude that our approach is able to outperform
the standard OLSR configuration and even parameters set by human experts in this matter. In
this sense, we will explore the use of our optimization model in the configuration of the OLSR
protocol for other VANET scenarios in future experiments.
Acknowledgements Authors acknowledge funds from the CICE, J. Andalucía, contract P07-TIC-03044
(DIRICOM http://diricom.lcc.uma.es) and Spanish Ministry MICINN and FEDER contract TIN2008-
06491-C04-01 (M* http://mstar.lcc.uma.es). José García-Nieto is supported by grant BES-2009-018767
from the MICINN
References
1. The Network Simulator Project - Ns-2. [online] http://www.isi.edu/nsnam/ns/.
2. T. Clausen and P. Jacquet. Optimized Link State Routing Protocol (OLSR). IETF RFC 3626, 2003
[online] in URL http://www.ietf.org/rfc/rfc3626.txt.
3. J. Garca-Nieto, J. Toutouh, and E. Alba. Automatic tuning of communication protocols for vehicular
ad hoc networks using metaheuristics. Engineering Applications of AI, 23(5):795 – 805, 2010.
4. C. Gómez, D. García, and J. Paradells. Improving performance of a real ad hoc network by tuning olsr
parameters. In ISCC ’05, pages 16–21, Washington, DC, USA, 2005. IEEE Computer Society.
5. J. Santa, M. Tsukada, T. Ernst, O. Mehani, S. Gómez, and F. Antonio. Assessment of VANET multi-hop
routing over an experimental platform. Int. J. Internet Protoc. Technol., 4(3):158–172, 2009.