Real-time high-precision location and mapping system based on NDT of the large scene

Abstract

The traditional lidar registration algorithm has good stability and high precision when carrying out large-scale simultaneous localization and mapping. However, in complex and changeable environments such as outdoor, severe occlusion, or electronic interference, the signal is easily affected, and the positioning accuracy cannot be guaranteed. To solve the above problems, a high-precision localization method based on laser-point cloud NDT (Normal Distributions Transform) is proposed in this paper, which makes full use of the mean and variance characteristics of NDT registration of the midpoint cloud, and the proposed method is added into the complete framework of SLAM (Simultaneous Localization and Mapping). The algorithm proposed in this paper was verified on the “Wuling Electric Vehicle“ platform. The experimental results were showed that the proposed algorithm could effectively avoid the influence caused by signal weakening, which improved the positioning accuracy. It also had stronger robustness and better tracking performance.