Abstract: In order to open the field of autonomous mobile robotics to new applications such as the provision of assistance to disabled people, the research is being focused upon low-cost solutions. That implies the use of poor perception systems and low computing power. In such a context, the algorithms used have to be simple, if they are to be executed in real time, and proof against the weaknesses of the sensing systems. The localisation approach presented here is based on the fact that the higher the localisation algorithm speed is, the lower the error in the position and the orientation, due to the odometry. Any systematic errors in the relative localisation using odometry are corrected on-line by using a limited set of ultrasonic data. If a non-systematic error occurs, a more complex procedure is necessary.
Both simulation and experimentation show that the systematic odometric errors become bounded, thanks to those algorithms. Moreover, they are robust to a high rate of false ultrasonic measures.

Keywords: Mobile robots, medical robotics, ultrasonic transducers, range data, least-squares algorithm, localisation estimation, localisation errors.