Modelling and Handling Uncertainties in Mobile Robotics

Abstract

This paper introduces a method for modelling a Mobile Robot's environment, with its in-built uncertainties, in the Configuration Space using sonar range finders. The sonar uncertainties are modelled using a possibilistic, rather than probabilistic, measure and are represented by Fuzzy Membership grades. The Resistive Grid path planning method, that maps the Fuzzy Configuration Space Map into a grid of resistors and models goals as voltage sources, is used for planning a path. The near-optimal planned path takes into consideration position and modelling uncertainties, kinematic constraints of the vehicle as well as the uncertainty of the goal. The path is produced incrementally, giving inherent interleaving between planning and execution, thus the method is particularly suited for real time applications.