PSOA-CRL: A Hybrid Multi-Objective Routing Mechanism Using Particle Swarm Optimization and Actor-Critic Reinforcement Learning For VANETs

Abstract

Vehicular ad hoc networks (VANETs) serve vehicles and infrastructure systems to communicate in real time for critical safety functions and traffic control. The highly mobile nature of VANETs with rapid topology changes, high mobility, and frequent disconnections is a very challenging situation for routing protocols. Most of the current approaches are static and tend to focus on a single metric rather than being flexible in practical environments. This paper introduces a hybrid routing method that is capable of maintaining a high packet delivery rate, low delay, and stable connectivity in VANETs with dynamic traffic situations. To address these problems, in this paper, we propose the PSOA-CRL, which is a hybrid multi-objective routing algorithm that integrates particle swarm optimization (PSO) with actor-critic reinforcement learning (A-CRL). The offline PSO component generates a variety of optimal routes. where the adaptive CRL just-in-time chooses the best available path. The two-way protocol maximizes the trade-off between the packet delivery ratio (PDR), end-to-end delay (E2E), link reliability, energy consumption, and routing overhead. A performance evaluation of PSOA-CRL with benchmarks under multi-objective optimization (MOO) through network metrics reveal the dominance of PSOA-CRL in most of the performance evaluation metrics. The obtained result reveals that the PSOA-CRL has a 97.8% packet delivery ratio, 41.3 ms end-to-end delay, and 96.1% link reliability. These results indicate that the PSOA-CRL is efficient in realizing reliable, real-time VANET routing and can be practically utilized in intelligent transportation systems (ITS).