Dense Normal Based Degeneration-Aware 2-D Lidar Odometry for Correlative Scan Matching

Abstract

Degeneracy detection and pose estimation of mobile robots have long been difficult task for laser scan matching techniques. Since the information perceived by lidar is underconstrained or deficient in degeneration scenes such as long corridors and lobbies, it is challenging for common hardware-limited lidar odometry to operate safely and stably. In this article, we propose a novel degeneration-aware correlative scan matching (CSM) algorithm for such situations. First, we carry out a low-dependency degeneration scene detection and analysis approach to determine the current degeneration situation accurately and instantly. Then, an environmentally sensitive CSM approach is designed based on real-time results, which dynamically adjusts the motion weight to achieve the autonomous adaptability to a variety of environments. Extensive experiments on both datasets and real-world scenarios demonstrate the superiority of the proposed algorithm in both degeneration scene detection and lidar odometry. In comparison to the CSM algorithm, the proposed algorithm not only minimizes the need of manually weights setup under dynamic environments but also ameliorates the accuracy of lidar odometry.