Reinforcement Learning for Resource Allocation in Edge Computing: Challenges and Future Directions

Real time IoT and 5G applications are dependent on edge computing to reduce latency and increase Bandwidth efficiency. The main challenge for one of them is dynamic resource allocation, like CPU and band width, that traditional methods cannot cope with. Adaptive, real time resource management in dynamic edge environments can be achieved with reinforcement learning (RL). The resource allocation in edge computing based on resource type, application domain, and algorithm is reviewed in this survey using the perspective of using RL in the area. This presents discussion of the performance, advantages, and limitations of these methods, as well as key challenges such as scalability, training overhead, and real time decision making. Also, it explores the use of emerging techniques, namely multia agent RL and hybrid models. Finally, future research directions include the usage of transfer learning, energy efficient RL Models and deployment of RL in real world scenario in edge computing environments. It turns out to be a guide to the improvement of subsequent RL approaches for resource management.