Intelligent Task Offloading Method using Deep Q-Network for Collaborative Edge Computing System

Abstract

Recently, various applications using artificial intelligence (AI) are deployed in edge network. In particular, An intelligence applications demanded with high computation and low end-to-end latency are executed on edge computing environments. Thus, in this paper, for the optimization of the resource of edge servers in multi-edge network environments, we propose the intelligent task offloading method based on Deep Q-network that can optimize computation capability of the multi-edge computing environments. For this, first at all, we formulate the problem of multi-edge computing allocation with a Markov decision process and propose the policy for allocating edge resource adopting a deep reinforcement learning algorithm. In the simulation, the results show the proposed method gets a better performance in terms of the end-to-end latency of the offloaded task than the existing methods.