Path Planning Algorithm for Orchard Scene Based on Road Edge Fitting
Aoze Wang, Junding Xiong, Longfei Gao, Zhigang Sun, Xiao Li, Gang Peng · 2025
Autonomous navigation of agricultural equipment is a key link to achieve smart agriculture and agricultural modernization, but the irregular distribution of fruit trees in the unstructured orchard scene will lead to an inconspicuous distribution of paths in the orchard scene, and the ground vegetation and fallen leaves will lead to a reduction in the accuracy of the traditional path edge detection algorithm, which can not give an accurate path driving direction. In order to complete the task of robot autonomous navigation in the orchard scene, this paper combines LiDAR and camera data to design a path planning algorithm based on road edge fitting. Firstly the algorithm performs ground segmentation and voxel filtering on the LiDAR point cloud data to retain the trunk features in the local point cloud map, then it uses the trunk point cloud clusters to calculate the trunk centroid, and fits the trunk centroid to obtain the trunk centerline. Then, the trunks in the image are identified using the YOLOv8 algorithm, and the obtained coordinates of the identification frame are used as road edge points and fitted to the road edge line. Finally, the trunk center line is used to constrain the road edge line to complete the correction of the path edge line, and the path center line equation is calculated according to the road edge line on both sides to obtain the target path and heading. The experimental results show that the algorithm can well extract the path center line from the orchard dataset, and the extracted path center line has good accuracy.