Sensor fusion in Robot localization using DS-Evidence Theory with conflict detection using Mahalanobis distance

Abstract

This article contains a new approach for combining sensory information for sensors with compatible information. Using Dempster-Shafer evidence theory, we show how to combine sensory information by Yager combination rule. We also present a method to distinguish contrasts between robots sensors outputs based on Mahalanobis distance, and find the sensor with irrelevant outputs. Experiments on a robot facilitated with vision and encoder shows that this approach reduces localization error significantly, having better performance than Kalman filter.