A Scan Matching Method For Quadruped Robots In Outdoor Environment

Abstract

In the outdoor unstructured environment, the relative displacement and rotation angle of the continuous laser scans are large due to the dynamic motion of the quadruped robots. Without good initial value, widely-used scan matching method ICP is prone to converge to a local optimum. In this paper, an improved Super 4PCS method based on fast and robust point cloud segmentation is proposed. Through point cloud segmentation preprocessing, unstable obstacles such as vegetation and leaves are removed and only large object point clouds with obvious features are extracted. More efficient point clouds to be registered are provided to Super 4PCS registration algorithm. Experiments prove that the improved method can be applied to quadruped robots in outdoor unstructured environment, speed up the matching and improve the accuracy of point cloud registration.