Research on Edge to End Collaborative Computing Unloading Based on Deep reinforcement learning

To address the limited computing resources and real-time data processing requirements of mobile devices in the mobile edge network environment, this paper proposes a deep reinforcement learning-based edge-cloud collaborative computing offloading method. Specifically, considering computational resources, bandwidth, and task offloading strategies comprehensively, we formulate the problem of minimizing task execution delay, energy consumption, and cost for all mobile devices. Based on this, we design an online coordinated offloading algorithm using deep reinforcement learning, which transforms the cost minimization problem into an optimal offloading strategy problem under Markov decision process framework. The best task offloading strategy is obtained by utilizing experience replay for solution. Experimental results demonstrate that the proposed algorithm exhibits good convergence and outperforms other methods in terms of system performance.