ABSTRACT
The goal of kinematic identification is to estimate a kinematic parameter vector that accounts in the least-squares sense for positioning and orientation errors of the robot. An iterative least squares kinematic identification procedure can go as follows: A number of robot end-effector pose measurements, together with the joint variables at each robot measurement configuration, is collected. In order to apply a least-squares algorithm for parameter estimation, one needs to derive an error model by linearizing the kinematic model. A finite difference approximation to a kinematic error model is straightforward to construct and for many practical applications can be used to replace a differential error model without loss of accuracy or numerical stability. This chapter provides the necessary tools for a robot calibration user to implement such an error model in any kinematic modeling convention. It focuses on the portion of the manipulator kinematic error model that is independent of the choice of a specific modeling convention.
