Learning probabilistic grid-based maps for indoor mobile robots using ultrasonic and laser range sensors

A new method for learning probabilistic grid-based maps of the environment of a mobile robot is described. New contributions on the three major components of map learning, namely, sensor data fusion, exploration and position tracking, are proposed. In particular, new models of sensors and a way of sensor data fusion that takes advantage of multiple viewpoints are presented. A new approach to control the exploration of the environment taking advantages of local strategies but without losing the completeness of a global search is given. Furthermore, a robot position tracking algorithm, based on polar and rectangular correlations between laser range data and the map is also introduced. Experimental results for the proposed approach using a mobile robot simulator with odometer, ultrasonic and laser range sensors (implemented with laser pointers and a camera), moving in an indoor environment, are described.