Comparative Evaluation of Occupancy Grid Mapping Methods Using Sonar Sensors

Building occupancy grid maps with sonar sensors is a challenging task due to angular uncertainty, specular reflections and crosstalk. This paper presents a quantitative comparison of two probabilistic and one heuristic approaches to the robotic mapping using real sonar data – inverse and forward sensor models and the CEMAL methods. Moreover, the two probabilistic methods are also tested pre-filtering the sonar data with the CEsp filter, which is part of the CEMAL approach. The results show that, using the pre-filtering, all algorithms present a similar and better performance, while without filtering the inverse method presents the highest error.