Resource Offload Consolidation Based on Deep-Reinforcement Learning Approach in Cyber-Physical Systems

In cyber-physical systems, it is advantageous to leverage cloud with edge resources distribute the workload for processing and computing user data at point of generation. Services offered by are not flexible enough against variations in size underlying data, which leads increased latency, violation deadline higher cost. On other hand, resolving above-mentioned issues devices limited also challenging. this work, a novel reinforcement learning algorithm, Capacity-Cost Ratio-Reinforcement Learning (CCR-RL), proposed considers both resource utilization cost target systems. CCR-RL, task offloading decision made considering arrival rate, device computation power, transmission capacity. Then, deep model created allocate based on communication rate. Moreover, new algorithms regulate allocation among servers. The simulation results demonstrate that method can achieve minimal latency reduced compared state-of-the-art schemes.