A mutated FastSLAM using soft computing

Abstract

Purpose Simultaneous localization and mapping (SLAM) is the problem of determining the pose (position and orientation) of an autonomous robot moving through an unknown environment. The classical FastSLAM is a well-known solution to SLAM. In FastSLAM, a particle filter is used for the robot pose estimation, and the Kalman filter (KF) is used for the feature location’s estimation. However, the performance of the conventional FastSLAM is inconsistent. To tackle this problem, this study aims to propose a mutated FastSLAM (MFastSLAM) using soft computing. Design/methodology/approach The proposed method uses soft computing. In this approach, particle swarm optimization (PSO) estimator is used for the robot’s pose estimation and an adaptive neuro-fuzzy unscented Kalman filter (ANFUKF) is used for the feature location’s estimation. In ANFUKF, a neuro-fuzzy inference system (ANFIS) supervises the performance of the unscented Kalman filter (UKF) with the aim of reducing the mismatch between the theoretical and actual covariance of the residual sequences to get better consistency. Findings The simulation and experimental results indicate that the consistency and estimated accuracy of the proposed algorithm are superior FastSLAM. Originality/value The main contribution of this paper is the introduction of MFastSLAM to solve the problems of FastSLAM.