Convergence and consistency analysis for FastSLAM

The main contribution of this paper is an analysis of the FastSLAM algorithm for simultaneous localization and mapping (SLAM) problem. The convergence properties of the landmark uncertainty for FastSLAM are provided. The proofs clearly show that the limit of the uncertainty for the landmark estimation has no relationship with the vehicle's initial pose uncertainty. Furthermore, the consistency of vehicle pose in FastSLAM is analyzed by Monte Carlo tests and finding that through reducing the control noise and measurement noise the consistency of the vehicle pose can't be improved remarkably, whereas, by cancelling the re-sampling process in FastSLAM the consistency of the vehicle pose can be enhanced obviously.