Resource Allocation and Migration Strategy of Edge Computing Based on Reinforcement Learning

In the mobile Internet scenario of edge computing, where high volatility and instability in user population and network traffic prevail, addressing uncertainty is crucial for reducing latency and enhancing user experience. This paper presents an Actor Critic-based method for resource allocation, optimizing the number of tasks processed by considering channel gain, bandwidth conditions, computing resources, and user positions. The proposed method incorporates an exploration-exploitation trade-off, effectively balancing the exploration of new actions and the exploitation of the currently known best actions. Experimental results demonstrate the algorithm's rapid convergence, significant improvement in utility value, and superior load balancing performance compared to existing approaches, achieving a 2.4% higher load balancing value. This research contributes to mitigating latency and improving user experience in the mobile edge computing environment, addressing uncertainties and enabling the deployment of more efficient and reliable applications.