Simultaneous localization and mapping (SLAM) is vital for autonomous robot navigation. The robot must build a map of its environment while tracking its own motion through that map. There are many ways to approach the problem, mostly based on the sequential probabilistic approach, based around extended Kalman filter (EKF) or the Rao-Blackwellized particle filter. In order to improve the SLAM solution and to overcome some of the EKF and PF limitations, especially when the process and observation models contain uncertain parameters, we propose to use a robust approach to solve the SLAM problem based on variable structure theory. The new alternative called Smooth Variable Structure Filter SVSF is a predictor corrector estimator based on sliding mode control and estimation concepts. It has been demonstrated that the (SVSF) is stable and very robust face modeling uncertainties and noises. Visual SVSF-SLAM is implemented, validated and compared with EKF-SLAM filter. The comparison confirms the efficient and the robustness of localization and mapping using SVSF-SLAM.
展开▼
