Load-Balanced Task Offloading Scheme for Vehicular Edge Computing Networks

Owing to the high mobility of vehicles, which triggers frequent disconnections and service handovers, task migration becomes necessary in Vehicular edge computing (VEC) networks. This leads to challenges such as inefficient resource utilization and load imbalance primarily caused by reactive migration strategies that fail to account for server heterogeneity. Frequent task migration incurs high computational and communication costs, yet existing schemes neglect both the computational overhead of migrations and inherent load disparities among VEC servers. To address this, we propose a load-balanced task offloading scheme featuring a novel dynamic migration method for real-time server coordination, implemented via an Adaptive Load Balancing Deep Deterministic Policy Gradient (ALB-DDPG) algorithm. Unlike traditional DDPG (limited to single-server selection), ALB-DDPG enables collaborative multi-MEC task processing through a many-to-one matching mechanism and integrates a load-balancing metric into its reward function to directly optimize server utilization. Extensive simulations show that, compared with existing schemes, the proposed ALB-DDPG reduces the average system cost by 28% and improves load-balancing efficiency by 40%, especially under heavy workloads.