Optimized Path Planning in Reinforcement Learning by Backtracking

Abstract

This paper focuses on finding the shortest path in a controlled indoor environment, called "smart lab". This study proposes a faster pathfinding model based on optimizing the decision-making process and fitting the hyper-parameters. At the same time, the study includes a comparison of performing the path planning on the robot through fog computing. The solution components are designed around the four main subcomponents of the reinforcement learning systems [1]: policy, reward signal, value function, and