Putting the 'I' in 'team': an ego-centric approach to cooperative localization

This paper describes a method for localizing the members of a mobile robot team relative to each other; that is, we describe a method whereby each robot may determine the pose of every other robot in the team, relative to itself. This method does not require GPS, landmarks, or any kind of environment model. Instead, robots make direct measurements of the relative pose of nearby robots, and broadcast this information to the team as a whole; each robot processes this information to generate ego-centric estimates of the pose of other robots. Our method uses a Bayesian formalism and a particle filter implementation, and is, as a consequence, extremely robust. The system will both self-initialize (it does not need to know the initial robot poses) and self-correct (it will recover quickly from tracking failures). It thus solves both the weak and the strong form of the kidnapped robot problem [17]. The method is, in addition, completely distributed, yet requires minimal communication bandwidth. This paper describes the basic formalism, sketches the implementation, and presents experimental results obtained using a team of four mobile