In the structural design of forestry robot harvesting machine, there are very few planning studies that can avoid obstacle movement. In this regard, this paper uses the VelodyneHDL-32E sensor carried by the UAV of DJI FC6320 to scan the trees of the four test sample plots,for the purpose of obtaining the tree point cloud data, and through the method of data pre-processing,so that the data set is free of interference information. With reference to the current mainstream point cloud segmentation methods, the PointNet network model was finally selected for high-precision tree point cloud segmentation,the loss function of the PointNet model was designated as the Softmax cross – entropy loss function, and in order to achieve a more desirable point cloud segmentation performance, the optimal parameters of the model were also determined through the calculation of the gradient. Combined with the characteristics of point cloud segmentation, we propose to use ACO-RRT* to complete the intelligent path planning of two-machine collaborative forestry robot based on point cloud segmentation.The path planned by ACO-RRT* algorithm is smoother than the RRT algorithm and RRT* algorithm, the path cost is smaller, and the optimization effect is better, and the specific enhancement is 67.14% and 15.16%, which verifies the performance of two-machine collaborative forestry robot based on point cloud segmentation. The effectiveness of intelligent path planning of two-machine collaborative forestry robot based on point cloud segmentation is verified. This paper can automatically plan the paths of target stumps and non-target stumps so as to avoid collision, which is of great theoretical and practical significance to improve the overall operational efficiency of forestry machinery.
