Distributed Service Migration in Satellite Mobile Edge Computing

Abstract

With the emergence of more and more latency-sensitive applications and mobile devices pumped to the edge of network, the burden on the backhaul network is getting heavier and heavier due to the limited transmission resources. Mobile edge computing (MEC) considered as a promising technology becomes more and more popular, which can provide services at the edge of the network. In this paper, we take into account the mobility of users and focus on the problem of service migration. Most of existing works modeled a Markov Decision Process (MDP) model with a high-dimensional state space, and have to solve it by deep reinforcement learning. To tackle this issue, we propose a distributed two-layer decomposition model and generate a series of new MDP problem with Low-dimensional in order to replace original High-dimensional MDP. In our model, the size of state space is reduced from M^2N to N × M2 by decomposing the original optimization problem. Simulation results show that the performance of the proposed two-layer decomposition model is better than the baseline models.