An improved FastSLAM2.0 algorithm using Kullback-Leibler Divergence

Abstract

The ability to simultaneously localize a robot and map its surroundings is considered by many to be a key prerequisite of truly autonomous robots. However, there is a dilemma between accuracy and computational complexity in existing SLAM algorithms. EKFSLAM algorithm, developed by Smith R in 1988, was first applied in SLAM. Nevertheless, high computational complexity became one of the main barriers for wide spread usage. To reduce the computational consumption, a new method based on conditional probability decomposition was used in FastSLAM, which makes the running time a logarithmic function of landmarks. Then the following FASTSLAM2.0 algorithm fused the proposed distribution with observation information, and it raised algorithm accuracy effectively. Aiming at the degeneracy problem in FastSLAM2.0, an improved resampling method using Kullback-Leibler Divergence is put forward, which contains particle degeneration largely. Simulation results show that this approach accelerates the convergence of particles set and restrains particle depletion as well.