META-HEURISTIC FUSION FOR 5G VANETS: A GWO–PSO–ACO FRAMEWORK BALANCING LATENCY, ENERGY, AND SPECTRUM

Abstract

Next-generation vehicular applications such as augmented-reality navigation and cooperative collision avoidance demand sub-second response times, low on-board energy use, and judicious utilisation of the scarce 5G/DSRC uplink spectrum. We address these conflicting requirements by formulating task-offloading in 5G-enabled vehicular ad-hoc networks (VANETs) as a multi-objective optimisation that minimises end-to-end latency and vehicular energy consumption while maximising deadline reliability and spectral efficiency. A detailed system model captures variable-size tasks generated by mobile vehicles, bandwidth-constrained LTE/5G and Wi-Fi channels, finite-capacity edge servers at roadside units (RSUs), and a remote cloud. Soft-deadline penalties are imposed on tasks whose latency exceeds 1s, and channel-congestion costs discourage excessive simultaneous off-loads. To solve the resulting NP-hard problem we propose an integrated GWO–PSO–ACO swarm optimiser: Grey-Wolf encircling provides global exploration, Particle-Swarm velocity updates accelerate exploitation, and Ant-Colony pheromone learning refines discrete task channel assignments. All three sub-swarms share the best candidate each iteration, yielding rapid yet robust convergence. Extensive simulations with random vehicle velocities (20–100 km/h) and varying numbers of vehicles (1–100, each generating one task) demonstrate that the proposed hybrid GWO–PSO–ACO algorithm consistently outperforms standalone PSO, ACO, and GWO baselines. Averaged over realistic workloads (1-40 MB tasks), the Hybrid achieves total latency reductions of approximately 21.9% and 29.3% compared to PSO and GWO, respectively, while lowering vehicular energy consumption by 12-18% and maintaining high reliability levels of ≈ 80-85% up to critical load points. Spectral efficiency is improved by up to 0.5% at low-to-moderate loads, and composite objective values are reduced by as much as 30–40% under heavy-load conditions. Convergence analysis confirms that the Hybrid reaches near-optimal solutions in fewer iterations than the baselines, making it suitable for real-time vehicular scenarios. Parameter variation tests further validate its scalability under heavier loads and stricter spectrum budgets. These results indicate that the proposed Hybrid optimiser is a robust and effective edge–cloud orchestration mechanism for future QoS- and spectrum-aware V2X services.