Improved Unscented Kalman Filter Algorithm to Increase the SLAM Accuracy

Abstract

Determining the position of autonomous mobile robots, especially when an external positioning reference like a satellite-based positioning signal is unavailable, poses a significant challenge. In such situations, to navigate through unknown environments effectively, robots need to create detailed maps of their surroundings while simultaneously determining their own location within those maps. This process is commonly achieved through the utilization of Simultaneous Localization and Mapping (SLAM) techniques, which enable robots to accurately ascertain their position and construct maps in real-time. In this paper an Improved Unscented Kalman Filter SLAM (IUKF-SLAM) is presented for improving the accuracy, consistency and convergency of the Unscented Kalman Filter (UKF) applied in the SLAM. In this method, the information matrix and the information state vector are calculated by the inverse of covariance matrix and transformation of the state vector, respectively, in the prediction step. Then they are updated in the update step. Simulation results indicate the superior accuracy of the proposed method compared to both EKF and UKF SLAM methods.