Plane-feature based 3D outdoor SLAM with Gaussian filters

Abstract

In this paper, a novel method extracting 3D planar features from laser range data (LiDAR) and its adaptation to outdoor SLAM using respective Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) is proposed. The paper is mainly divided into two parts such that the feature extraction algorithm and its application to SLAM problem are given. Firstly, the feature extraction from 3D LiDAR data using a probabilistic plane extraction method and merging criteria with a region growing algorithm is explained. Secondly, the extracted 3D planar features are used with the well-known Gaussian local filters such as EKF or UKF to solve SLAM problem. The plane-feature based SLAM method estimate the robot pose in 6D as well as the plane parameters in 4D, which are the normal of plane and its minimum distance to origin represented in the world frame. Although the feature extraction method is proposed for outdoor SLAM, since it is developed from an indoor feature extraction method it can be safely used in indoor, outdoor, and even in the complex environments. The method is evaluated with the real datasets, and the EKF and UKF based SLAM performances are compared. The results show that UKF has better performance than EKF and can be successfully used in SLAM problems.