Low-cost and High-accuracy LIDAR SLAM for Large Outdoor Scenarios

Abstract

Be applied in the large outdoor scenarios, we keep the high-accuracy of robot's mapping and location. Relied on lightweight hardware, we have improved an algorithm of real-time LIDAR SLAM (simultaneous localization and mapping). Compared to vision sensors, LIDARs have higher accuracy and larger range for outdoor scenarios. In the situation of robot moving quickly, LIDARs can keep the features no missing in the scan range, meanwhile camera maybe happen feature fuzzy or missing. Velodyne VLP-16 LIDAR can provide 3D point cloud with 16 channels' scan. By segmenting for LIDAR point clouds, and extracting more special edge or corner feature, we compute robot' motion model that relied on the methods like Point-to-line Iterative Closest Points (PL-ICP) or Point-to-Plane Iterative Closest Points (PP-ICP). Based on robots' motion model, matching point cloud features in a certain range to fix the rigid body motion model. In every certain distant, optimized robot's state information and map based on the method of graph optimization. We use this method to achieve lightweight and high-accuracy LIDAR SLAM for large outdoor, and computed the state of robot. By making experiments in our school, we tested our method for building high accuracy map and providing robot's state.