Efficient motion planning for chili flower pollination mechanism based on BI-RRT

To achieve obstacle avoidance for a chili flower pollination robotic arm in complex and narrow environments, this paper proposes a pollination action planning algorithm based on Map Preprocessed Step-by-Step Informed Rapidly-exploring Random Trees (PSBI-RRT). The algorithm, which is based on the Bidirectional Rapidly-exploring Random Tree (BI-RRT) algorithm, pre-processes the robot’s task space to obtain local candidate points. These points are used to divide the task space and determine segmented local goal points based on the distribution of obstacles. In the PSBI-RRT algorithm, a segmented dynamic sampling space is used instead of a fixed sampling space to reduce invalid sampling points. By introducing a hybrid expansion method combining greedy expansion with adaptive goal point attraction, the algorithm rapidly generates paths, improving the flexibility of the PSBI-RRT algorithm in exploring unknown spaces and enhancing its adaptability to the environment. 3D simulation experiments in various environmental types show that the PSBI-RRT algorithm reduces search time by over 95%, and it decreases invalid sampling nodes by 70% compared to traditional methods. The quality of the pollination path is optimized, with an average path length reduction of 30%. Pollination experiments with a 6-DOF robotic arm in both simulated and real environments show that the collision-free paths planned by the algorithm successfully guide the robotic arm from the initial to the target position without collisions. Additionally, in several typical pollination tasks, the success rate of path planning remains at 95%, significantly improving the planning success rate.