基于遗传粒子群算法的Stewart型并联机器人位姿优化及标定

Position Optimization and Calibration of Stewart-Type Parallel Robot Based on Genetic Particle Swarm Algorithm

摘要: 并联机器人在加工与装配过程中不可避免地产生误差，影响其精度。目前，国内外标定主要依赖高精度设备，存在成本高、操作复杂、人工参与多等问题，导致标定效率低。因此，开发高效且精度可靠的标定方法尤为重要。本文分析了Stewart型并联机器人运动学，建立了其正逆解模型，并推导了位姿误差模型。通过最小二乘法辨识结构参数误差，针对标定过程中的测量位姿优化问题，采用遗传粒子群算法在工作空间内优化选取最优位姿。与经典粒子群算法相比，改进算法提高了辨识精度，最后进行了位姿误差补偿实验。经过标定后，平均位置误差由标定前的0.363 mm降低至0.039 mm；平均姿态误差由标定前的5.82 × 10⁻⁴ rad降低至4.51 × 10⁻⁵ rad。

Abstract: Parallel robots inevitably introduce errors during machining and assembly, which affect their accuracy. Currently, calibration methods worldwide primarily rely on high-precision equipment, which is costly, complex to operate, and requires significant manual intervention, resulting in low calibration efficiency. Therefore, developing efficient and reliable calibration methods is crucial. This paper analyzes the kinematics of the Stewart parallel robot, establishing its forward and inverse kinematic models, and deriving a pose error model. Using the least squares method, structural parameter errors are identified. To optimize the selection of measurement poses during the calibration process, a genetic particle swarm algorithm is employed to optimize the selection of optimal poses within the workspace. Compared to the classical particle swarm algorithm, the improved algorithm achieves higher identification accuracy. Finally, pose error compensation experiments are conducted. After calibration, the average position error decreased from 0.363 mm to 0.039 mm, and the average orientation error reduced from 5.82 × 10⁻⁴ rad to 4.51 × 10⁻⁵ rad.