Computation offloading Optimization in Edge Computing based on Deep Reinforcement Learning

By considering an MEC system consisting of multiple mobile devices with stochastic task arrivals, a computational offloading and resource allocation strategy based on Deep Reinforcement Learning (DRL) is proposed. Specifically, a continuous action space based DRL approach named deep deterministic policy gradient (DDPG) is adopted to learn efficient computation offloading policies independently at each mobile user. Thus, powers of both local execution and task offloading can be adaptively allocated by the learned policies from each user's local observation of the MEC system. Through simulation, it can be verified that efficient policies can be learned at each mobile device, and the performance of the DDPG-based strategy is better than the traditional deep Q network (DQN) -based discrete power control strategy, which reduces the computation cost.