Reinforcement Learning for Robots Path Planning with Rule-based Shallow-trial

Abstract

A key skill for mobile robots is the ability to navigate efficiently through their environment, and reinforcement learning is widely used in path planning for mobile robots. However, this algorithm has a slow convergence speed and a large number of iterations. There are few studies on how to improve learning efficiently from the perspective of acquisition in rule-based shallow-trial strategy. In biological world, animals depend on their own empirical knowledge when making path planing. Humanity has transcendental knowledge, which is of great help to peoples navigation. We take the transcendental knowledge of human behavior, and express it acts as shallow-trial rules, then apply the rule-based shallow-trial reinforcement learning(RSRL) to the navigation learning of robot and improve learning efficiently.