Double-Arm Cooperation and Implementing for Harvesting Kiwifruit

Abstract

Double-arm picking robots are widely used in agricultural production for their high collaborative efficiency. While picking, area planning and collision detection between the mechanical arms is a crucial challenge for the double-arm robot, which needs to establish a collision-free path for fruit picking. In this study, we developed a double-arm cooperation method for robotic picking of kiwifruit. Firstly, the problem of dividing the picking area was simplified into a multiple traveling salesmen problem (MTSP) to be solved. The picking sequence of each robotic arm was formulated by the principle of similar picking numbers, and combined with the brainstorming optimization algorithm (BSO). Secondly, a double-arm parameter model was built to solve the forward and backward movements of the robotic arms and to figure out the joint position. The spatial mathematical relationship of the bounding boxes between the robotic arms was used to detect the collision between the two robotic arms, in order to achieve the avoidance between the robotic joints. Then, simulation software was applied to the simulation and analyzed the availability of picking area planning and collision detection. The simulation results showed that the optimized picking sequence planning using BSO was more efficient; the smooth joint trajectory during the movement of the robotic arms met the limits on the range of movement and on the angular velocity of the robotic arm joints. Finally, based on the simulation result, a double-arm collaboration platform was tested. The double-arm collaboration platform harvesting trials showed that the average picking success rate was 86.67%, and collision detection time was 3.95 ± 0.83 s per fruit. These results indicated that the proposed method could plan the operation tasks of the double-arm picking robot system, and effectively implement the collision-free picking operation.