Multi-ultrasonic sensor fusion for mobile robots

To learn the environment using multisensory information, we need both an accurate sensor model and a reasonable sensor fusion methodology. Ultrasonic sensors provide good range information. However, uncertainties in ultrasonic sensors caused by the specular reflection from environments make them less attractive. We have used the Dempster-Shafer evidence method in sensor fusion with the specially designed sensor model. By applying a filtering factor to the sensor model, uncertainties in sonar responses can be successfully reduced. In this paper, we introduce a novel way of using the conflict value in Dempster-Shafer evidence theory in the sensor model. Through this new method, our robot would be able to modify its sensor model dynamically during its navigation. Experimental results have shown that the new method has improved the performance of the modified ultrasonic sensor model in dealing with specular reflections.