Task migration for mobile edge computing using deep reinforcement learning

Abstract

Mobile edge computing (MEC) is a new network architecture that puts computing capabilities and storage resource at the edges of the network in a distributed manner, instead of a kind of centralized cloud computing architecture. The computation tasks of the users can be offloaded to the nearby MEC servers to achieve high quality of computation experience. As many applications’ users have high mobility, such as applications of autonomous driving, the original MEC server with the offloaded tasks may become far from the users. Therefore, the key challenge of the MEC is to make decisions on where and when the tasks had better be migrated according to users’ mobility. Existing works formulated this problem as a sequential decision making model and using Markov decision process (MDP) to solve, with assumption that mobility pattern of the users is known ahead. However, it is difficult to get users’ mobility pattern in advance. In this paper, we propose a deep Q-network (DQN) based technique for task migration in MEC system. It can learn the optimal task migration policy from previous experiences without necessarily acquiring the information about users’ mobility pattern in advance. Our proposed task migration algorithm is validated by conducting extensive simulations in the MEC system.