Improving consistency of EKF-based SLAM algorithms by using accurate linear approximation

This paper presents a modified EKF-based SLAM algorithm to improve the consistency of the EKF-based SLAM algorithms. The proposed algorithm extracts the exact linear approximation of the measurements, which is considered as a variable, updates it using the new measurements, and finally transforms it back into the original state. The exact linear approximation is achieved by maintaining the point for linearization and updated along with the state. In this way, the structure of the variables being updated is more accurate, and the inconsistency of the EKF-based SLAM is greatly reduced, while at the same time, the computation and memory requirements do not increase too much. Simulation and experiment results demonstrate the advantages of the new algorithm.