As an emerging computing paradigm, Vehicular Edge Computing (VEC) provides computational support for applications requiring real-time processing of massive data. However, the spatial and temporal distribution of in-vehicle tasks and limited system resources make effective task offloading decisions, edge node load balancing, and efficient resource management challenging in VEC networks. To solve these problems, firstly, we construct an Air-Ground heterogeneous Edge Computing (AGEC) network combining VEC and Unmanned Aerial Vehicle (UAV) and use a paid service mechanism to incentivise the edge nodes to provide offloading services to users. Secondly, we built a non-convex optimization problem by considering network utility weighted by multiple performance metrics under resource-limited constraints and different service strategies. Finally, we propose a Deep Deterministic Policy Gradient (DDPG) based Deep Reinforcement Learning (DRL) method for interactive exploration in dynamic environments to find the optimal solution to a nonconvex optimization problem and achieve an optimal resource allocation. Simulation results show that the proposed algorithm improves the average network utility by 16.8%, 240%, and 12.7%, reduces the average energy consumption by 7%, 19.3%, and 1.5%, and reduces the average latency by 4%, 27%, and 4%, respectively, when compared to the other three DRL-based baseline methods.
Li et al. (Wed,) studied this question.