Abstract:

A method of constructing grid map using improved Dempster-Shafer (D-S) evidence theory was proposed for unstructured environment mapping. A mobile robot was used to explore environment with ultrasonic sensors. The problems that the D-S evidence theory can not be applied to information fusion under certain circumstances and that it has counterintuitive behaviors in some cases were discussed. An approximate process algorithm was developed to avoid the above problems and improve the D-S evidence theory. This improved D-S evidence theory was used to fuse multi-sensor information; then a two-dimensional D-S grid map was built. With D-S evidence rule of decision making, a 0 1 grid map was built. The D-S grid map was analyzed, and an algorithm to evaluate the environment and environment exploration effect based on D-S grid map was proposed. Simulation results show that the improved D-S evidence theory is appropriate to unstructured environment mapping.

Key words: automatic control technology, approximate process algorithm, D-S evidence theory, multi-sensor information fusion, grid map