These Matlab simulations are of EKF-SLAM, FastSLAM 1.0, FastSLAM 2.0 and UKF-SLAM. The intent of these simulators was to permit comparison of the different map building algorithms. However, they might also be useful to the wider research community interested in SLAM, as a straight-forward implementation of the algorithms.

Download

EKF-SLAM version 1. This older version of the EKF-SLAM simulator is probably easier to understand than the 2nd version, as it avoids using global variables.

EKF-SLAM version 2. This version of the EKF-SLAM simulator runs much faster in MatLab as it avoids copying overhead by using global variables. It also comes with an alternative observation model that can replace the 'update' function in 'ekfslam_sim.m'. This alternative is a "global constraint" model devised by Jose Guivant, and may have better linearisation properties than the conventional range-bearing model.

FastSLAM 1.0 and 2.0. This implementation is slow in Matlab due to the overhead of looping constructs etc. However, it can give a good idea of how each algorithm works, and may serve as a starting point for more efficient implementations.

UKF-SLAM. This simulator is a direct adaptation of the EKF-SLAM code, but replaces the EKF with an unscented Kalman filter (UKF). To run this code you must first install MatLab utilities.