Decoupled motion planning method for 7-DOF manipulator and lifting joint in automated tomato harvesting

Abstract

Robotic harvesting tasks must contend with complex operation environments, where varying grasping poses, fruit heights, and obstacle distributions impose higher demands on the robot’s effective working space and dexterity. A fixed robotic arm base placement for the robotic arm can result in targets being outside the arm’s reachable space or unreachable with the given approach pose, which reduces the robot’s harvesting success rate. In this work, we propose a step-by-step path planning method that decouples the robotic arm and the lift joint. This method is based on the reachability map of the robotic arm and takes into account the arm’s manipulation costs (including similarity cost and gravity cost) as well as the lift joint’s movement cost. After performing a fast non-dominated sorting of the costs for the discrete point candidates, it generates sorted lift height candidates. Finally, improved RRT* for the robotic arm is executed at the optimal position of the lift joint. Field test results conducted on 36 samples validate the effectiveness of the proposed method. In observation tasks, the system achieved a success rate of 35 out of 36 (97.2 %), with a lift joint planning time of 3.09 ms. For grasping tasks, the decoupled approach achieved an overall success rate of 83.3 % (30/36) with a lift joint planning time of 5.95 ms, significantly outperforming the fixed-set method (36.1 %, 13/36) and Reuleaux method (77.8 %, 28/36). This indicates that the method can significantly improve the success rate of tomato harvesting by robots in vertical farms, with an acceptable time cost. The findings establish a robust foundation for improving robotic harvesting systems by addressing the limitations of conventional approaches. The proposed method not only achieves high success rates and motion efficiency but also provides a scalable and practical solution for modern agricultural robotics.