Improved Deep Deterministic Policy Gradient Algorithm based on PER for Partial Task Offloading

Mobile edge computing (MEC) is the deployment of storage and computing resources at the edge of the network for latency-critical applications. However, it is often challenging for wireless devices (WDs) to make offloading decisions, when WDs offload tasks to a BS in a dynamic and stochastic environment with time-varying wireless channels. In this paper, We propose an improved deep deterministic policy gradient (DDPG) algorithm based on prioritized experience replay (PER) to minimize the average task processing delay. We consider a time division multiple access (TDMA) system with wireless power transfer (WPT), where data is transmitted using packetized communication. The computing task offloading and resource allocation problem are formulated as a constrained Markov decision process with a hybrid action space. Simulation results demonstrate the efficiency of the proposed Improved-DDPG algorithm in enhancing the task processing rate and achieving a 50% reduction in task processing latency,