Multi-Stage PSO-Based Cost Minimization for Computation Offloading in Vehicular Edge Networks

With the fast development of autonomous driving, the demand of computing resources becomes a big challenge for resource-constrained vehicles. To alleviate this issue, vehicular edge computing (VEC) has been proposed to offload real-time computation tasks from vehicles. However, complex physical constraints in real VEC applications make computation task offloading become a fundamental issue in VEC. A high-quality offloading strategy can not only complete computational tasks, but also minimize the cost of computing and resource offloading. The work proposes a multi-stage particle swarm optimization (MPSO)-based offloading method for VEC. It significantly optimizes the energy cost under specified delay limits. Compared with original PSO, it improves the convergence by applying a staged optimization strategy. Experiments show that it saves 91%-97% of cost than a typical random offloading strategy, depending on delay limits and vehicle numbers. Moreover, it has 31% improvement of convergence than a PSO-based method under the same simulation parameter setting.