Robot Calibration Using Active Vision-based Measurement

This paper presents an efficient robot calibration method with non-contact vision metrology. Using the coplanar pattern to calibrate camera made the active-vision-based end-effector pose measurement be a feasible and cost-effective way. Kinematic parameter errors were linearized and identified through two-step procedure, thus the singular and non-linear condition was overcome. These errors were then compensated using inverse model method. The whole calibration process is flexible, easy to implement and prevents the error propagation from the earlier stages to the later ones. Calibration was performed on MOTOMAN SV3 industrial robot. Experiment results show that the proposed method is easy to setup and with satisfactory accuracy.     

工业机器人运动学参数误差两步识别法

提出了将转动误差和定位误差分离的参数误差两步线性识别方法,利用刚体相对转动误差识别出仅与转动相关的运动学参数误差;由已校准的参数并利用刚体距离误差识别出其余的运动学参数误差,两步识别法不仅避免了测量坐标系与机器人坐标系的转换以及参数估计误差的传递,而且有效解决了距离方法中存在的奇异和非线性问题,可以快速求出参数误差的线性解,该识别法不仅可以用于串联、并联机器人的校准,而且适用于数控机床、坐标测量机等的误差识别与补偿。     

基于主动视觉的机器人末端姿态测量

采用平面校准形及移动摄像机到不同形位摄取图像的方法来进行摄像机校准，由灭点正交性和单对应矩阵的固有特征估计摄像机内、外参数初值，通过非线性优化来最小化投影误差；利用遗传算法求解机器人手眼矩阵，避免了初值估计，可容易地得到机器人执行器末端相对于绝对坐标系的姿态．实验和计算验证表明，这种测量方法是有效的，可以用于机器人校准、机器人辅助测量以及机器人辅助手术等领域．