Vision-based Localization of Leader-Follower Formations

This paper focuses on the localization problem for a mobile camera network. In particular, we consider the case of leader-follower formations of nonholonomic mobile vehicles equipped with vision sensors which provide only the bearing to the other robots. We prove a sufficient condition for observability and show that recursive estimation enables a leader-follower formation if the leader is not trapped in an unobservable configuration. We employ an Extended Kalman Filter for the estimation of each follower position and orientation with respect to the leader and we adopt a feedback linearizing control strategy to achieve a desired formation. Simulation results in a noisy environment are provided.