A DSmT-based Fusion Machine for Robot’s Map Reconstruction

Characteristics of uncertainty and imprecision, even imperfection is presented from knowledge acquisition in map reconstruction of autonomous mobile robots. Especially in the course of building grid map using sonar, this characteristic of uncertainty is especially servere. Jean Dezert and Florentin Smarandache have recently proposed a new information fusion theory (DSmT), whose greatest merit is to deal with uncertainty and conflict of information, and also proposed a series of proportional conflict redistribution rules (PRC1∼PCR5), therein, presently PCR5 is the most precise rule to deal with conflict factor according to its authors, though the complexity of computation might be increased correspondingly. In this chapter, according to the fusion machine based on the theory of DSmT coupled with PCR5, we not only can fuse information of the same reliable degree from homogeneous or heterogeneous sensors, but also the different reliable degree of evidential sources with the discounting theory. Then we established the belief model for sonar grid map, and constructed the generalized basic belief assignment function (gbbaf). Pioneer II virtual mobile robot with 16 sonar range finders on itself served as the experiment platform, which evolves in a virtual environment with some obstacles (discernable objects) and 3D Map was rebuilt online with our self-developing software platform. At the same time, we also compare it from other methods (i.e. Probability theory, Fuzzy theory and Dempster-Shafer Theory (DST)). The results of the comparison shows the new tool to have a better performance in map reconstruction of mobile robot. It also supplied with a foundation to study the Self-Localization And Mapping (SLAM) problem with the new tool further. Supported by National Natural Science Foundation of China (No.69585003). The corresponding author: Xinde Li ([email protected]).