Sonar feature based exploration

Abstract

A mobile robot wishing to explore unknown and dynamic environments, has to build up maps of its environment while at the same time ensuring that it does not become lost in the process. However current feature based localisation techniques can become unstable if the robot drives in such a way that its sensors cannot recognise these features. A novel dynamic path planning approach is presented that supports the data acquisition for the localisation process as well as the extraction of new feature while exploring an a priori unknown indoor environment. Real experiments on the authors' mobile robot Roamer have shown that the overall localisation stability and the feature extraction process were improved significantly while exploring an unknown environment.