Mobile Edge Computing Offloading Strategy Based on Deep Reinforcement Learning

Edge computing is a very good choice to solve the shortage of computing power of mobile devices (MD), the bandwidth resources and computing power of edge servers are also limited, therefore, efficient task offloading strategies and resource scheduling can reduce task computing time, reduce system energy consumption and improve user experience. In this study, a three-layer network architecture is proposed, namely a “end-edge-cloud” three-layer network architecture. Secondly, considering the maximum tolerance delay of user service, a comprehensive optimization model for task offloading and resource allocation aiming at the minimum energy consumption of the system is designed. The model clearly defines the state space, action space and reward function of Markov stochastic process, and adopts the simulation based on double Q-network (DDQN) and dueling deep Q-network (Dueling DQN). Experimental results show that the proposed scheme is better than multiple baseline algorithms and the original DQN algorithm.