PaperlayerView exact revision on arXiv

Source-linked AI summary

Intelligent OLSR Routing Protocol Optimization for VANETs

Jamal Toutouh, José García-Nieto, Enrique Alba

arXiv:2501.09716v1cs.NEcs.NI

TL;DR

VANET routing must cope with mobility, wireless limitations, fragmentation, and the lack of central coordination. This paper automatically tunes OLSR with four metaheuristics and simulator-based evaluation, obtaining configurations that improve QoS over RFC 3626 and human-expert settings.

  • Problem

    VANET routing faces packet loss, topology changes, fragmentation, and congestion, while OLSR performance depends on difficult-to-select parameter combinations.

  • Method

    The paper couples PSO, DE, GA, and SA with ns −2 simulations to optimize OLSR configurations across realistic Málaga-based VANET scenarios.

  • Results

    PDR = 100% on the optimization scenario, increasing PDR by 8.34% over the standard configuration and by 6.66%–28.57% over expert configurations.

  • Takeaways & Limitations

    Optimized OLSR configurations reduce routing workload and generally shorten packet delivery times while preserving delivered-data degradation below 5%.

Abstract

from arXiv · show

Recent advances in wireless technologies have given rise to the emergence of vehicular ad hoc networks (VANETs). In such networks, the limited coverage of WiFi and the high mobility of the nodes generate frequent topology changes and network fragmentations. For these reasons, and taking into account that there is no central manager entity, routing packets through the network is a challenging task. Therefore, offering an efficient routing strategy is crucial to the deployment of VANETs. This paper deals with the optimal parameter setting of the optimized link state routing (OLSR), which is a well-known mobile ad hoc network routing protocol, by defining an optimization problem. This way, a series of representative metaheuristic algorithms (particle swarm optimization, differential evolution, genetic algorithm, and simulated annealing) are studied in this paper to find automatically optimal configurations of this routing protocol. In addition, a set of realistic VANET scenarios (based in the city of Málaga) have been defined to accurately evaluate the performance of the network under our automatic OLSR. In the experiments, our tuned OLSR configurations result in better quality of service (QoS) than the standard request for comments (RFC 3626), as well as several human experts, making it amenable for utilization in VANET configurations.

I. Introduction

VANET routing is difficult because mobility, limited wireless coverage, obstacles, and the absence of central coordination cause instability and delivery problems. The paper addresses this challenge by automatically tuning OLSR with metaheuristics and realistic Málaga scenarios.

  • High mobility, limited WiFi coverage and capacity, and obstacles generate packet loss, topology changes, and network fragmentation.
  • Routing is challenging because VANETs lack a central entity responsible for finding paths among nodes.
  • OLSR is selected for automatic tuning because it offers competitive delays, adapts to topology changes, and is simple to integrate.
  • OLSR quality of service depends strongly on parameter selection, which can affect data rates and network load.
  • The paper couples PSO, DE, GA, or SA with ns −2 to search OLSR configurations using realistic VANET instances based on Málaga.
  • Automatically tuned OLSR configurations outperform the standard configuration and human-expert configurations in the reported experiments.

II. Problem Overview

OLSR is presented as a suitable proactive routing protocol for dynamic VANETs, but its control overhead limits scalability in dense networks. Parameter tuning is therefore formulated as a way to improve QoS before deployment.

  • OLSR is a proactive routing protocol suited to ad hoc networks with frequent and rapid topological changes.
  • OLSR maintains routing tables for possible routes, creating control-message overhead that can consume bandwidth and provoke congestion in large dense networks.
  • The protocol’s performance depends on parameters such as the interval for broadcasting HELLO messages.
  • The methodology is presented as applicable beyond OLSR to other VANET routing protocols.

A. OLSR Protocol

OLSR is a proactive link-state protocol that periodically disseminates topology information through HELLO, TC, and MID messages. Multipoint relays reduce the transmissions required for this flooding process.

  • OLSR maintains network topology through periodic flooding of control information using multipoint relays.
  • HELLO messages support link sensing, neighborhood detection, and multipoint-relay selection among one-hop neighbors.
  • Topology-control messages are generated by multipoint relays and forwarded across the network for routing-table calculations.
  • MID messages report the network interfaces used by nodes participating in communications.
  • OLSR parameters include message-hold times, node willingness to act as a multipoint relay, and the duplicate-message hold time.

B. OLSR Parameter Tuning

The paper tunes OLSR parameters by representing configurations as solution vectors and evaluating them through a simulator-based communication cost function. The function uses packet delivery ratio, routing load, and end-to-end delay.

  • The standard OLSR configuration provides moderate QoS in VANETs, motivating optimization of its parameter values before deployment.
  • OLSR parameters are represented as real-valued components of a solution vector that an optimization technique can fine-tune.
  • The optimization framework couples metaheuristic search with ns −2 simulation to evaluate tentative OLSR configurations.
  • The communication cost function evaluates packet delivery ratio, network routing load, and end-to-end delay.

III. Optimization Framework

The framework couples metaheuristic optimization with ns−2 simulation to automatically tune OLSR parameters for VANET scenarios. Candidate configurations are evaluated through a communication-cost fitness function that prioritizes routing effectiveness.

  • Optimization procedure: Four metaheuristics—PSO, DE, GA, and SA—search the continuous OLSR parameter space for optimal or quasi-optimal configurations.The algorithms represent varied search and optimization strategies.
  • Simulation stage: Each tentative OLSR configuration is simulated in ns−2 over a defined VANET scenario to obtain a quantitative fitness value.The simulator returns PDR, NRL, and E2ED information from 10 independent data transfers.
  • Fitness function: The communication-cost function minimizes NRL and E2ED while maximizing PDR through a weighted aggregation.PDR appears with a negative sign because the objective is minimizing communication cost.
  • Fitness function: The weighting scheme gives PDR priority over NRL and E2ED, emphasizing routing effectiveness before communication efficiency.The three weights control each metric’s influence on the resulting fitness value.

IV. Experiments

The experiments use realistic urban VANET scenarios derived from Málaga road data to evaluate automatically tuned OLSR configurations. The setup addresses ns−2’s lack of direct realistic vehicle-behavior modeling.

  • Scenario design: The simulation design aims to approximate real-world VANET environments through realistic urban scenarios.The setup combines scenario construction with ns−2 simulation parameters.
  • Scenario design: ns−2 cannot directly define realistic vehicle behavior, including road movement, traffic lights, and traffic signs.A traffic simulator is therefore used to generate realistic VANET environments from real digital-map areas.
  • Málaga scenario: The Málaga instance contains 30 cars moving through a 1200×1200 m2 downtown area for 3 min.Vehicles exchange data while following roads selected for the urban scenario.
  • Experimental assumptions: The experiments use a fixed data rate of 4 packets/s to assess route discovery and maintenance rather than maximum throughput.The remaining simulation parameters are summarized in Table II.

B. Experimental Setting

The experimental setting compares four metaheuristics and random search under a common evaluation budget using ns−2-based OLSR simulations. The study examines algorithm performance, configuration comparisons, scenario generalization, and parameter influence.

  • Algorithms: The study evaluates PSO, DE, GA, SA, and RAND for OLSR parameter tuning.The four metaheuristics are implemented with MALLBA, while RAND is separately implemented in C++.
  • Evaluation budget: All five methods use the same stopping condition: 1000 fitness-function evaluations.SA and RAND perform 1000 iterations, whereas population-based algorithms run 100 generations with populations of 10.
  • Simulation implementation: The simulation phase uses ns−2.34 with UM-OLSR version 0.8.8, and every technique is evaluated across 30 independent runs.The implementation was executed on specified Pentium IV hardware under Linux Fedora Core 6.
  • Analysis scope: The experiments report algorithm performance, comparisons with literature configurations, robustness across scenarios, and OLSR-parameter influence on QoS.These four perspectives structure the reported experimental analysis.

A. Performance Analysis

SA achieves the strongest overall solution quality, while PSO offers the best quality–time tradeoff. The algorithms differ in convergence behavior and execution cost, with statistical tests confirming differences among their results.

  • Solution quality: SA outperforms the other algorithms in mean, median, and worst communication cost: −0.450297, −0.457451, and −0.406932, respectively.The ordering is SA, DE, PSO, GA, then RAND.
  • Solution quality: PSO returns the lowest communication cost among the best configurations, followed by DE, SA, and GA.RAND produces the least competitive best configuration.
  • Statistical analysis: The Friedman test ranks SA first, followed by DE, PSO, and GA, while RAND ranks last; Kruskal–Wallis comparisons yield p-values ≪0.05.The distributions are reported as statistically different at the stated 95% confidence level.
  • Convergence behavior: Algorithm convergence differs: DE improves mainly early, PSO mainly late, and SA improves gradually throughout the execution.DE, PSO, and SA converge within the same range of solutions.
  • Execution time: PSO has the shortest mean runtime at 5.38E + 04 s among the metaheuristics.DE, SA, and GA follow PSO in mean running time.
  • Execution time: Mean execution times range from 4.36E + 04 to 1.18E + 05 seconds, or 12.11 to 32.66 h.The authors describe the offline design effort as justified by subsequent QoS benefits after deployment.
  • Quality–time tradeoff: SA ranks best globally but requires more time, whereas PSO provides the best tradeoff between solution quality and time.The ranking considers mean fitness, time to best solution, and mean runtime.

B. Optimized Versus Human Expert Configurations

Metaheuristic-tuned OLSR configurations outperform human expert, standard RFC 3626, and RAND configurations across key QoS indicators in the Málaga scenario. The algorithms differ by metric: contrast achieves perfect delivery, DE and PSO minimize routing load, and SA minimizes delay.

  • 100% PDR was achieved by all four metaheuristic configurations, compared with 94.12% for RAND and 71.43%-93.34% for the remaining configurations.
  • 2.71% NRL for DE and 2.90% for PSO were approximately one order of magnitude lower than most competing configurations.GA recorded 13.12%, while RFC 3626 and RAND recorded 9.52% and 6.93%, respectively.
  • 4.73 ms E2ED was obtained by SA, the best result among the compared configurations.PSO, DE, and GA showed moderate E2ED performance because their low routing load limited routing management operations.
  • All optimized OLSR settings delivered packets in less than 20 ms, the highest allowed latency for cooperative vehicular applications.

C. Optimized Configurations on Multiple Scenarios

The tuned configurations were validated across 54 urban Málaga VANET scenarios varying geographical area and traffic density. Collectively, metaheuristics offered the best tradeoff across packet delivery, routing load, delay, and path length.

  • 54 urban VANET scenarios varied traffic density, network use, and geographical area to validate optimized OLSR parameters under different conditions.Three geographical area sizes and three traffic densities—low, medium, and high—were used.
  • 89.56% was the highest global median PDR, obtained by RFC, with configuration differences limited to 1%-5%.GA and SA reached 99.95% in U1, PSO reached 86% in U2, and RFC reached 86.71% in U3.
  • 11.98% overall NRL was obtained by DE, half the RFC value of 23.15%, while SA and DE generally showed the best routing-load values.RFC had the worst overall NRL across the three scenarios.
  • 4.04 ms was the shortest global median E2ED, achieved by SA; GA was shortest in U1 and U2, while DE led U3 at 19.19 ms.
  • Metaheuristic solutions achieved packet delivery above 84%, routing load below 16.5%, E2ED below 10.3 ms, and paths shorter than 2 hops.RFC also reached accurate median PDR values but had routing load above 23%.
  • The standard OLSR parameters generally produced the shortest median paths by exchanging routing information more frequently, but at higher routing load.

D. Global QoS Analysis

OLSR parameter choices create a QoS tradeoff between route reactivity and protocol overhead. Shorter signaling intervals reduce delay but increase routing load, congestion, and packet loss, whereas optimized timers maintain delivery within the latency target.

  • Shorter HELLO_INTERVAL and TC_INTERVAL values reduce E2ED by improving reactivity to route changes and link failures.
  • The same shorter intervals increase NRL and may cause network congestion and lower PDR.
  • More than 84% of packets were delivered within 20 ms using PSO, DE, GA, and SA settings, whose NEIGHB_HOLD_TIME and TOP_HOLD_TIME timers were larger.

VI. Conclusion

The paper automatically tunes OLSR for VANETs by coupling metaheuristic algorithms with network simulation, then evaluates the resulting configurations across realistic Málaga scenarios. The optimized configurations improve delivery and workload-related QoS measures relative to standard and expert settings, while larger scenarios and additional protocols remain future work.

  • 54 urban VANET scenarios based on real downtown Málaga data were used to validate optimized OLSR configurations against RFC 3626 and human expert configurations.
  • Simulated annealing ranked best by the Friedman test, while particle swarm optimization offered the best tradeoff between optimization performance and execution time.Parallel PSO can further reduce computational time for large VANET simulations.
  • PDR reached 100% on the optimization scenario, increasing 8.34% over RFC 3626 and 6.66%-28.57% over expert-defined configurations.The optimized configurations also dramatically reduced OLSR routing load.
  • Validation produced about half the routing load of RFC 3626, routing paths 27% longer, shorter packet delivery times, and less than 5% degradation in delivered data.
  • Metaheuristic-tuned OLSRs were judged more scalable than the standard version, with PSO providing the best QoS-to-routing-workload tradeoff.The methodology couples metaheuristics with a simulator to customize protocol configurations.
  • Future work extends evaluation to larger urban and highway VANET instances and develops parallel algorithms to address simulation time costs.The authors also plan optimized schemes for WAVE, UDP, and other communication protocols, plus real vehicle tests on different roads.
Loading 2501.09716v1…