Although calibration of a robotic manipulator's kinematic parameters is well-established as a technique for improving end-effector positioning accuracy, little attention has been given to the relationship between the parameters of a kinematic model and the corresponding physical features for a particular manipulator. If each kinematic parameter of a robot corresponds to one and only one of its independent physical features, then a change in the value of an optimal parameter would reflect a change in the corresponding physical feature. This knowledge is of potential value in isolating the sources of wear and damage (which alter the physical features) in a manipulator, and more generally, in predictive maintenance. I have shown that the choice of kinematic model has a strong effect on the physical feature/kinematic parameter relationship, as well as on the calibration process in general. The most suitable kinematic model for preserving this relationship includes redundant parameters, which may numerically interact among themselves during the parameter estimation process and may complicate the interpretation of results. On the other hand, a kinematic model with no redundant parameter contains too few parameters to establish a comprehensive physical features/kinematic parameters relationship. I have explored multiple site pose measurement as a method of preserving the desired relationship even in the presence of redundant parameters. I have analyzed the calibration process and the effect of redundant kinematic parameters thereon from both geometrical and linear algebra viewpoints.
Robot tasks involving off-line programming crucially depend on the absolute accuracy of the robot. Accuracy issues arise for us in the context of employing a robotic system as a precision positioning device in a medical operating room. A robot may potentially improve the overall accuracy, and hence, the quality of an orthopedic surgery by guiding the surgeon to different pre-operatively-selected poses. Since accuracy requirements are extremely stringent in invasive medical applications such as knee-surgery or brain surgery, it is necessary to develop better error-compensation methods for robots.
The absolute accuracy of a robot depends to a large extent on the accuracy with which its kinematic parameters are known. Many methods have been explored for inferring the kinematic parameters of a robot from measurements taken as it moves. Some require an external global positioning system, usually optic or sonic. We used instead a simple radial-distance linear transducer (LVDT) which measures the distance from a fixed point in the workspace to the robot's endpoint. This incomplete pose information is accumulated as the robot endpoint is moved within one or more spherical shells centered about the fixed point. Optimal values for all the independent kinematic parameters of the robot can then be found. This represents a particularly easy and economic approach to robot calibration.