A FastSLAM Based on the Smooth Variable Structure Filter for UAVs

Abstract

Three-dimensional (3-D) simultaneous localization and mapping (SLAM) is an effective technique of autonomous navigation for the Unmanned Aerial Vehicles (UAVs), in which the UAV can estimate the vehicle's pose and build a map in GPS-denied environments simultaneously. FastSLAM algorithm can solve the quadratic computational complexity and single-hypothesis data association problem of the classical EKF-SLAM algorithm, in which the SLAM problem can be factored into a product of a UAV's path posterior estimated by the particle filter (PF) and independent landmark posteriors estimated by the extended Kalman filter (EKF). But the FastSLAM algorithm suffers from lower map accuracy introduced by ignoring correlation information of landmarks. A SVSF-FastSLAM algorithm for UAVs is presented, in which the SVSF is adopted to estimate the landmarks' position. The simulation results are given to show the effectiveness of the proposed algorithm. Compared with the conventional FastSLAM algorithm, the SVSF-FastSLAM algorithm shows that a more accurate estimation of trajectory and environment can be achieved.