Deep Reinforcement Learning algorithms for Low Latency Edge Computing Systems

Nowadays, due to the increase of technological development in smart devices, more computational capabilities are needed with better performance. The maximization of the cloud computing resources used in mobile networks and Internet of Things makes good results. The data processing along with delay in the communication network will be easily handled by edge computing technology. The edge server brought as much as close to the end devices to avoid the frequent cloud access. In edge servers, the processing of the computational task is time consuming, which causes latency. The inference of deep learning along with reinforcement models makes the better resource allocation and task scheduling for complex systems. In this work, deep reinforcement models like Q learning, Double Q learning, Deep Q network, Double Deep Q learning has been compared. The developed Deep Reinforcement learning model has been deployed in the edge-based system. As a result, it is inferred that Double Deep Q learning is having the ability to make better decisions in resource allocation by maximizing the gains and rewards. The simulation result shows that Double Deep Q learning performs better with decreased latency and increases the performance of the edge systems.