Single-Sensor Probabilistic Localization on the SeReS Self-Reconfigurable Robot

This paper proposes a novel method for localizing a stationary infrared source of unknown orientation relative to a static docking sensor. This method uses elliptical approximations of likely positions of the infrared source and computes the intersections to find the most probable locations. It takes only a few samples to localize, is easily computed with inexpensive microcontrollers, and is robust to sensor noise. We then compare our approach with two other methods. The first uses a Bayesian filter across a map of discrete locations in the robot’s operational workspace to determine the suspected source position. The second also uses a probability distribution map but uses the method described by Elfes in his paper on probabilistic sonar-based mapping and navigation [1]. We show that our approach localizes quickly with a single sensor and is no more computationally demanding than other methods.