Efficient State Space Composition Technique for Intelligent Wheelchair Agents

Abstract

Acquiring environmental information and using it for learning or evolutions are significant issues for an intelligent wheelchair with autonomous, cooperative and collaborative behavior. However, the learning with too many dimensional data derives "curse of dimensionality," so that the learning becomes inefficient. In order to resolve the issues, this study proposes a new technique by which continuous environmental data is converted in arbitrary number of dimension discrete data.