New Adaptive UKF Algorithm to Improve the Accuracy of SLAM

SLAM (Simultaneous Localization and Mapping) is a fundamental problem when an autonomous mobile robot explores an unknown environment by constructing/updating the environment map and localizing itself in this built map. The all-important problem of SLAM is revisited in this paper and a solution based on Adaptive Unscented Kalman Filter (AUKF) is presented. We will explain the detailed algorithm and demonstrate that the estimation error is significantly reduced and the accuracy of thenavigation is improved. A comparison among AUKF, Unscented Kalman Filter (UKF) and Extended Kalman Filter (EKF) algorithms is investigated through simulated as well as experimental dataset. An indoor dataset is generated from a two-wheel differential mobile robot in order to validate the robustness of AUKF-SLAM to noise of modeling and observation, and to examine the applicability of the method for real-time navigation. Both experimental and simulation results illustrate that AUKF-SLAM is more accurate than the standard UKF-SLAM and the EKF-SLAM. Finally, the well-known Victoria Park dataset is used to prove the applicability of the AUKF algorithm in large-scale environments.