Cloud game computing offload based on Multi-Agent Reinforcement Learning

In this paper, we propose a two-layers computing task offload architecture for cloud game scene, including edge servers and cloud server. In order to better simulate the real environment, we modeled the uplink and downlink communication channels for the first time. In order to reduce the energy consumption and time delay of the system, this paper adopts Multi-Agent Reinforcement Learning Algorithm - Bidirectionally-Coordinated Nets(BiCNet) algorithm to realize it. Each edge server is regarded as an agent. Compared with the single agent reinforcement learning algorithm, the decision made by BiCNet saves 300% of the time delay and 200% of the system energy consumption. The experimental results show that the offloading decision made by BiCNet algorithm can most effectively balance the consumption of energy cost and time delay, and can achieve the effect of energy saving for the system while ensuring the user experience.