Experimental Evaluation of Occupancy Grid Map Improvement by Sonar Data Corrections

This paper addresses the creation of a consistent occupancy grid map from local sensors data collected during mobile robot movement. The map building problem, together with the mobile robot localization problem, is one of the major problems of mobile robots navigation. It can be defined as an environment modelling problem using different algorithms for sensor data integration. Bayesian and Dempster-Shafer algorithm combined with sonar sensors were used in our approach. A simple solution to deal with irregular disposition of sonar sensors on mobile robot body has been proposed. It increases the robustness of mapping, which is experimentally evaluated for both Bayesian and Dempster-Shafer on-line mapping algorithms. Experiments were conducted on Pioneer 2 DX mobile robot