Hierarchical Reinforcement Learning for Robot Navigation using the Intelligent Space Concept

Abstract

Navigation in an unknown environment is a difficult task, because mobile robots need topological maps in order to operate in the environment. Another fundamental problem is that robot programming is a time-consuming process, so it is better to use a learning method with reinforcement. In previous work we proposed a learning framework, which used the capability of the Intelligent Space in order to build a topological map of the environment. In this paper we present an extension of this framework to decompose the learning problem into sub-problems, which can be learned faster.