基于Sylvester方程变形的荔枝采摘机器人手眼标定方法

针对视觉荔枝采摘机器人的Eye-in-Hand视觉与机器人关联方式的手眼标定问题,该文提出一种基于优化的求解齐次变换矩阵方程的手眼标定方法。该方法通过机器人带动其臂上的双目相机从多个位置观测标定板,使用Sylvester方程变形对手眼标定近似方程线性化,再对简单的初值进行优化计算,最终得到精确的标定结果。该方法的软件用C++/Open CV开发实现,并进行了多个试验。试验结果表明,视觉与机器人关联后,定位误差与机器人运动次数相关,当距目标1 m左右,静态时的视觉系统误差均值为0.55 mm;动态工作时,视觉关联机器人重复定位误差的均值为2.93 mm,标准差为0.45 mm,符合具有容错功能的视觉荔枝采摘机器人的实际使用需求。使用基于Sylvester方程变形的手眼标定方法标定的视觉荔枝采摘机器人,在野外环境下,总体采摘成功率达到76.5%,视觉系统成功识别、定位采摘点的情况下,采摘成功率达92.3%。

Hand-eye calibration method based on Sylvester equation deformation for lychee harvesting robot

Abstract: Aiming to realize the obstacle avoidance of fruit harvesting robot manipulator in unstructured environments, a kind of obstacle avoidance path planning algorithm of hybrid harvesting manipulator based on joint configuration space was proposed in this study. The structure of the 2P4R hybrid camellia oleifera fruit harvesting robot manipulator composed of a palletizing manipulator and a spherical wrist serial manipulator was simplified, and the link connecting the palletizing manipulator and serial manipulator was named as moving platform. The manipulator could accomplish six kinds of movements, including waist rotation, translational motions of vertical slider and horizontal slider, as well as three kinds of rotational motions of the wrist part. In this study, only five joint motions were taken into consideration. It meant that the last link was considered as a fixed part of its previous link when a collision-free path planning operation was conducted. This was because the motion of the last link was not related to the position of end effector which could only affect its posture. Firstly, a goal point for the moving platform of the manipulator in Cartesian space was selected with the proposed algorithm. The goal point should be located in workspace of the parallel manipulator, rather than inside obstacles. The initial and goal configurations of serial manipulator were determined by the moving platform of the goal point, the goal position of end effector and initial posture of serial manipulator. Secondly, the obstacle mapping model in serial manipulator joint configuration space was built with traversal method. Then, an attempt was made to search for a collision-free route from initial point to the goal point in this space with each point uniquely corresponding to a configuration of the serial manipulator by using rapid-exploring random tree (RRT) algorithm. Thirdly, if the algorithm failed to find such a route in previous step, the selection of goal point and collision-free route searching operation would be repeated. Otherwise, joint configuration space of parallel manipulator would be established. Fourthly, the unique posture of parallel manipulator was determined based on the mapping relationship between driving joint value and position coordinates of moving platform. The obstacle model was built in the joint configuration space of parallel manipulator. Subsequently, the goal point of the moving platform would be selected again, if the point in parallel manipulator configuration space corresponding to the goal configuration was located in mapping model of obstacle. Otherwise, a collision-free route from the start point to the goal point in configuration space corresponding to initial posture and goal configuration of the parallel manipulator respectively should be searched by using RRT algorithm. The obstacle avoidance path of hybrid manipulator was obtained from synthesized results of collision-free paths of serial manipulator and parallel manipulator. In order to verify the feasibility and effectiveness of the proposed algorithm, path planning simulation of a hybrid manipulator using Matlab was carried out. Proposed algorithm was also applied in obstacle avoidance path planning experiment on camellia oleifera fruit harvesting manipulator. According to the simulation and experiment results, path planned by the proposed algorithm could successfully drive the end effector from its initial position to the goal position without any collision. That is to say, these results can validate the feasibility and effectiveness of the proposed algorithm.