Abstract:

To solve the problem of computational complexity for the regular extended Kalman filter method in cooperative localization, an information filter based on joint distribution state to study cooperative localization for robots is presented. The proposed method consists of three key components to solve the computational complexity problem. Firstly, our approach preserves the historical states in the filter, which avoids the process of time updating that draws lessons from simultaneous localization and mapping. Secondly, the information parameters are sparse, thus the computational complexity of the filter is less. Finally, the special properties of the Cholesky modification algorithm are also used for further decreasing the computational complexity, which is convenient to distribute the work. The simulation result indicates that the method ensures these performance advantages as well as guarantees the estimation precision and the effectiveness of cooperative localization.