Fuzzy ART based approach for real-time map building

Abstract

In order to perform complicated tasks in previously unknown indoor environments, autonomous mobile robots must be able to acquire models of their environments. The model or map, is often used for planning robot motions which perform a task while avoiding obstacles in the environment. Applications where there is no previously available world model usually rely on sensor systems, such as laser range finders, and infrared and sonar range sensors, to collect thousands of range points which are difficult to handle and interpret. The paper introduces an approach, based on the fuzzy ART neural architecture, for online map building. The map is incrementally constructed in real-time by fitting rectangular geometrical primitives to actual sensor data collected with a mobile robot. This results in a compact world model that is easy to integrate into planning approaches. A sensor data filtering approach is proposed to reject noisy range points. It is assumed that the robot knows its own current world location. Results of experiments with a real Nomad 200 mobile robot are presented, demonstrating the operation of the proposed map building method.