A Modified FastSLAM for an Autonomous Mobile Robot

Based on the standard FastSLAM, a new decomposing strategy is proposed in this paper. Although the method involves a little extra computation, it considers fully the relationship between the pose of robot and the individual map feature, and it is better to improve the precision and reliability of SLAM. In addition, a nonlinear adaptive square-root unscented Kalman filter (NASRUKF) is utilized to replace the extended Kalman filter (EKF) and the particle filter (PF) to estimate the pose of robot and the map feature, then several shortcomings of traditional FastSLAM (such as it needs to calculate Jacobian matrix, it is difficult to satisfy the requirement of consistency, and it always yields particle degradation) are overcome, the accuracy and adaptability are improved. In order to verify the concept, one experiment prototype was built. The result proved that the modified FastSLAM algorithm provides a high performance of localization and map building for an autonomous mobile robot (AMR).