Improvements of Occupancy Grid Maps by Sonar Data Corrections

In this paper we address one of the major problems of mobile robots navigation, the creation of a map from local sensor data collected as the robot moves around an unknown environment. Map building is the problem of generating models of robot environments from sensor data and can be often referred as a concurrent mapping and localization problem. That is to build a consistent map, the mobile robot has to know its pose. We present here three approaches to create occupancy grid maps from sonar's data and suggest a simple solution to improve the mapping quality in cases of irregular disposition of the sonars. The proposed solution has been tested on the mobile robot Pioneer 2DX.