基于距离测量的机器人误差标定及参数选定

为消除机器人运动学标定中的模型奇异问题，研究距离平方差模型的可辨识参数.采用修正DH（Denavit-Hartenberg）模型描述机器人连杆结构，六参数描述末端关节到工具坐标系的变换，推导了连杆参数误差到机器人末端距离平方差的线性模型，从冗余关系式和物理解释角度讨论冗余参数，确定可辨识参数，使用迭代的最小二乘法辨识参数完整且非奇异的模型.结合某6旋转关节机器人，以激光跟踪仪为测量仪器进行标定实验，在设计末端执行器时目标测量点偏离末端关节轴线以保证倒数第二关节参数的辨识.结果显示，距离平方差模型避免了测量坐标系到基坐标系的转换，检验点的距离误差均方根降低93.1%，完整性和非奇异性提高了参数辨识的有效性.

Robot error calibration based on distance measurement with parameter selection

To eliminate singularity in robots' kinematics calibration, identifiable parameters in distance square difference model were confirmed. Robot's link structure was expressed by modified Denavit-Hartenberg (DH) model, while the transformation between last joint and tool frame by six parameters. The linear model from link parameter errors to distance's square difference of robot's end-effector was derived. After redundant parameters eliminated via redundant expressions and physical interpretation, complete and nonsingular model was identified by iterative least square algorithm. Experiment was conducted with a 6 rotational robot and Leica laser tracker. Measured target should be away from the last joint's axis in order to guarantee penultimate joint parameters identifiable when end-effector was designed. The result shows that the model avoids coordinate transformation, the error distance's root-mean-square of inspection data reduced 93.1% and completeness and non-singularity improved identification effectiveness.