Sensor fusion for ultrasonic and laser arrays in mobile robotics: a comparative study of fuzzy, Dempster and Bayesian approaches

In any autonomous mobile robot, if not the most, one important issue to be designed and implemented on the robot, is environment perception and its role in autonomous navigation. There are many grid-based and topological methods for environment mapping. Among the grid-based methods the main difference is about the method of data integration that is applied to mapping. In this paper, three different approaches are formulated to perform sensor data integration in the process of generation of a generalized version of occupancy grids map of the environment. The methods are formulated based on Bayesian, Fuzzy and Dempster-Shafer approaches to data fusion/integration. Although, they are famous for data fusion applications, in this research work they have been applied, formulated and simulated to solve a unique problem: map building for the same mobile robot, equipped with 8 Polaroid ultrasonic range finder sensors and operating in the same environment. The simulation results are applied for comparative study of the merits of the methods and their applicability in the map building and environment perception for autonomous mobile robots. They show that the Bayesian approach gives more appropriate maps, by which, A* path planning algorithm leads to shorter and safer routes for the mobile robot to navigate.