| Summary: | To address the demands for low latency, high energy efficiency, and security in vehicular edge computing scenarios, a task offloading and resource allocation strategy based on deep reinforcement learning, combined with intelligent reflecting surface technology was proposed. Firstly, unlike approaches that optimized a single objective, a weighted comprehensive index of delay, energy consumption, and task completion rate was constructed as the system optimization objective. Furthermore, a resource-optimized twin-delayed deep deterministic policy gradient algorithm was designed, which was capable of perceiving task completion status. By introducing a task completion rate-based reward function and incorporating a successful trajectory prioritized replay mechanism, the learning stability of the algorithm in sparse environments was enhanced, thereby improving offloading success rate and resource utilization efficiency under poor signal coverage or unstable links. Experimental results demonstrated that, compared with state-of-the-art baseline algorithms, the proposed method reduces average system cost by about 25%, average latency by 23.2%, and average energy consumption by 17.6%, while increasing the task completion rate by 7%.
|
|---|
