Efficient Deep Reinfocement Learning Framework in Edge Computing Environments

In this paper, we propose a new framework for deep reinforcement learning in edge computing environments. First, we describe the application of distributed reinforcement learning techniques from previous research to edge computing environments. Based on network delay and the computational capabilities, we propose a framework that enhances the learning efficiency of deep reinforcement learning algorithms. In the proposed framework, the roles of the actors and the learner, running on the edge and cloud servers respectively, are dynamically adjusted based on available resources and network latency. The performance of the proposed framework is validated through experiments on a testbed.