Fully Decentralized Cooperative Localization of a Robot Team: An Efficient and Centralized Equivalent Solution

This paper presents an efficient, centralized equivalent and fully decentralized solution to the cooperative localization of mobile robot teams. Formulating the cooperative localization problem in the framework of Bayesian estimation, the decentralized solution is designed by interlacing the calculation steps of prediction and update in a proper sequence. In the proposed solution, each robot fuses only the sensor data relevant to itself; information is shared among the robots by a chain communication topology. The solution yields linear minimum mean-square error estimates, equivalent to a centralized extended Kalman filter. There is no information redundancy and computation duplication among the robots. The solution can also be viewed from the perspective of implementing inference on a specific junction tree. The performance of the proposed algorithm is evaluated with simulation experiments.