A FUSION MACHINE BASED ON DSMT AND PCR5 FOR ROBOT'S MAP RECONSTRUCTION

Abstract

Characteristics of uncertainty, imprecision, and even imperfection are presented from knowledge acquisition in map reconstruction using sonar sensors fixed on autonomous mobile robot. In order to improve the precision of the fusion and performances of map reconstruction, we propose in this paper a new fusion machine based on Dezert-Smarandache Theory (DSmT) coupled with the fifth Proportional Conflict Redistribution rule (PCR5) for dealing with uncertain and conflicting evidences provided by homogeneous or heterogeneous sources of information. We propose a belief model for sonar grid map and show how to construct efficiently generalized basic belief assignment functions for sensors onboard. A Pioneer II mobile robot with 16 sonar range finders serves as the experiment platform. In our experiment, the robot evolves in a real environment with some obstacles and the environment map is rebuilt online with our self-developing software platform. In this study, we also compare our new approach with other ones based on probability theory, fuzzy theory, Dempster-Shafer Theory (DST) and gray system theory. Our results show an improvement of the performances in precision of map reconstruction of mobile robot with respect to those obtained from aforementioned classical approaches.