Integrating a geometric-primitive map into a multiresolution motion planning approach

Abstract

Presents a method for mobile robot navigation that integrates (1) the fuzzy ART neural architecture for online map-building, into (2) the parti-game learning approach. Using the improved world model, a predictive online trajectory filtering method, is introduced in the learning approach for increasing the effectiveness of robot exploration and planning. This results in an overall powerful method for simultaneous construction of a world model, and learning to navigate from an initial position to a known goal region in an unknown world. It is assumed that the robot knows its own current world location, is able to perform sensor-based obstacle detection (not avoidance), and straight-line motions. Simulation and real-robot navigation results with a Nomad 200 mobile robot are presented. Quantitative results demonstrate (1) the improvements of the navigation approach, and (2) that the constructed world model (original or improved) is general purpose in the sense that its usefulness is not restricted to be used on learning a particular path, but is valuable for learning paths with different (start, goal) pairs.