Decentralized Offloading Strategies Based on Reinforcement Learning for Multi-Access Edge Computing

Using reinforcement learning technologies to learn offloading strategies for multi-access edge computing systems has been developed by researchers. However, large-scale are unsuitable learning, due their huge state spaces and behaviors. For this reason, work introduces the centralized training decentralized execution mechanism, designing a model systems. Considering cloud server several servers, we separate in model. The happens devices of system, servers need no communication. Conversely, process occurs at device, which causes lower transmission latency. method uses deep deterministic policy gradient algorithm optimize strategies. simulated experiment shows that our can strategy each device efficiently.