In a partially known navigation and localization environment

An algorithm for autonomous navigation of a vehicle using sonar sensors is presented. The main feature is that it is designed to operate in mapped environments in the presence of unmapped obstacles. Key requirements are: 1) robust localization in the presence of unknown features and vehicle motion, 2) estimation of deterministic disturbances such as currents, and 3) localization of unknown objects. The approach is based on a suitable potential function describing the environment and on extended Kalman filtering techniques to provide recursive estimation of the vehicle location.