Distributed Task Scheduling for MEC-Assisted Virtual Reality: A Fully-Cooperative Multi-Agent Perspective

In the upcoming 6th generation of mobile networks, wireless virtual reality will become an essential application. Due to the heterogeneity of rendering requirements and edge servers, how to provide adaptable task scheduling capabilities in edge networks becomes an important issue. However, most related studies regard edge servers as self-interested nodes that make scheduling decisions independently, ignoring the potential power of cooperation among edge servers. In this paper, we propose a fully-cooperative task scheduling scheme for virtual reality, taking into account parallel multitasking, the randomness of task arrival and leaving, as well as network load balance. By sharing a common objective among distributed schedulers, tasks can be offloaded to proper edge servers in a cooperative way. We first formulate an optimization problem to maximize the number of delay-satisfied tasks in the entire edge network. Next, we treat the edge network as a fully-cooperative multi-agent system and transform the problem into a decentralized partially observed Markov decision process (Dec-POMDP). Finally, we propose a cooperative task scheduling scheme based on multi-agent proximal policy optimization to solve the Dec-POMDP. Simulation results show that the cooperation among edge nodes brings the network improvement in terms of offloading success rate, delay compliance rate, and load balance.