Fast Grid-Based Position TRacking for Mobile Robots

One of the fundamental problems in the field of mobile robotics is the estimation of the robot's position in the environment. Position probability grids have been proven to be a robust technique for the estimation of the absolute position of a mobile robot. In this paper we describe an application of position probability grids to position tracking. Given a starting position our approach keeps track of the robot's current position by matching sensor readings against a metric model of the environment. The method is designed to work with noisy sensors and approximative models of the environment. Furthermore, it is able to integrate sensor readings of different types of sensors over time. By using raw sensor data, the method exploits arbitrary features of the environment and, in contrast to many other approaches, is not restricted to a fixed set of predefined features such as doors, openings or corridor junction types. An adaptable sensor model allows a fast integration of new sensings. The results described in this paper illustrate the robustness of our method in the presence of sensor noise and errors in the environmental model.