Software-Defined Heterogeneous Edge Computing Network Resource Scheduling Based on Reinforcement Learning

With the rapid development of wireless networks, edge computing networks have been widely considered. The heterogeneous characteristics 6G network bring new challenges to resource scheduling. In this work, we consider a with nodes and task requirements. We design software-defined architecture separate control layer data layer. According different requirements, tasks in are decomposed into multiple subtasks at layer, node alliance responding is established perform subtasks. order optimize both energy consumption load balancing, model scheduling problem as Markov Decision Process (MDP), Proximal Policy Optimization (PPO) algorithm based on deep reinforcement learning. Simulation analysis shows that proposed PPO can achieve low ideal balancing.