Acoustic-based position discrimination of a moving robot

A real-world experiment is described, demonstrating the possibility of discriminating positions of a mobile robot moving through an unknown, unprepared office room, solely based on broadband audible acoustic signals. The average distance of distinguishable positions was found to be less than 15 cm. Maximum length sequence measurements are used to obtain the impulse response of the room at the present position, and a modified vector distance measure is used to cluster the data via the Neural Gas algorithm. No further interpretation of the data is performed. The data analysis method is independent of the environment or the recording system (the robot), therefore it is potentially applicable to other environments and vehicles. The method is especially interesting for the navigation of AUVs in unknown environments, as acoustic signals represent a major source of information in the underwater world, being available under the broadest range of circumstances.