A guiding bidirectional rapidly exploring random-tree*-based trajectory planning method for tobacco picking robot

China occupies a leading position in global tobacco cultivation, and automated picking is become the direction of future development. Aiming at the demand for fast and smooth movement in the trajectory planning problem of tobacco picking robot, a trajectory planning method based on the Guiding Bidirectional Rapidly-exploring Random Tree* (GBi-RRT*) algorithm is proposed. The method decouples motion planning into three parts: spatial model construction, path generation, and path optimization. The method first performs point cloud preprocessing and octree map spatial model construction, uses the GBi-RRT* algorithm to carry out trajectory planning between picking points, uses the greedy idea to optimize the planned paths, and uses the B-spline curve to carry out path smoothing. After experimental verification, the proposed algorithm, compared with the original RRT average sampling times, was reduced by 83.18%, the average path length was reduced by 55.24%, and the planning success rate increased by 16.67%, which can be better for the trajectory planning of the picking points.