An Integrated LiDAR-SLAM System for Complex Environment with Noisy Point Clouds

Kangcheng Liu

DOI: https://doi.org/10.48550/arXiv.2212.05705

IF: 3.7

2022-12-12

Robotics

Abstract:The current LiDAR SLAM (Simultaneous Localization and Mapping) system suffers greatly from low accuracy and limited robustness when faced with complicated circumstances. From our experiments, we find that current LiDAR SLAM systems have limited performance when the noise level in the obtained point clouds is large. Therefore, in this work, we propose a general framework to tackle the problem of denoising and loop closure for LiDAR SLAM in complex environments with many noises and outliers caused by reflective materials. Current approaches for point clouds denoising are mainly designed for small-scale point clouds and can not be extended to large-scale point clouds scenes. In this work, we firstly proposed a lightweight network for large-scale point clouds denoising. Subsequently, we have also designed an efficient loop closure network for place recognition in global optimization to improve the localization accuracy of the whole system. Finally, we have demonstrated by extensive experiments and benchmark studies that our method can have a significant boost on the localization accuracy of the LiDAR SLAM system when faced with noisy point clouds, with a marginal increase in computational cost.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve?

This paper primarily addresses two key issues in LiDAR SLAM (Simultaneous Localization and Mapping) systems with large amounts of noisy point clouds in complex environments:

1. **Noise Removal**: Current LiDAR SLAM systems have limited capabilities in handling noise and outliers when processing large-scale point clouds. To address this, the authors propose a lightweight network (Denoise-Net) for denoising large-scale point clouds to improve the quality of the local map.

2. **Loop Closure Detection**: To enhance global localization accuracy, the authors designed an efficient loop closure detection network (LC-Net) to identify place features and correct drift issues in the global map.

With these two improvements, the proposed method significantly enhances the overall localization accuracy and robustness of LiDAR SLAM systems when dealing with noisy point clouds. Experimental results show that this method is suitable for both mechanical and solid-state LiDARs and performs excellently in various complex environments.