Kinematic Modeling, Identification, and Control of Robotic Manipulators

The objective of this dissertation is to advance the state-of-the-art in the kinematic modeling, identification, and control of robotic manipulators with rigid links in an effort to improve robot kinematic performance. The positioning accuracy of commercially-available industrial robotic manipulators depends upon a kinematic model which describes the robot geometry in a parametric form. Manufacturing error in the machining and assembly of manipulators led to discrepancies between the design parameters and the physical structure. Improving the kinematic performance thus requires the identification of the actual kinematic parameters of each individual robot. The identified kinematic parameters are referred to as the arm signature. Existing robot kinematic models, such as the Denavit-Hartenberg model, are not directly applicable to kinematic parameter identification. In this dissertation we introduce a new kinematic model, called the S-Model, which is applicable to kinematic parameter identification, and use it as the foundation for our development of a general technique for identifying the kinematic parameters of any robot with rigid links. The objective of our S-Model identification algorithm is to estimate the S-Model kinematic parameters from a set of mechanical features which are inherent to the manipulator. Each revolute joint possesses two such features and each prismatic joint possesses one. These features contain the essential information to model completely the kinematics of a manipulator. The initial step of the algorithm involves the explicit identification of the feature parameters. Each feature is identified in an independent procedure and is based upon measurements of the three-dimensional Cartesian positions of target points mounted on each of the links of the manipulator. A relatively simple and systematic method for collecting these measurements is one of the practical advantages of our approach. The identified feature parameters are then used to establish the positions and orientations of Cartesian coordinate frames fixed relative to each link of the manipulator in accordance with the definition of the S-Model. The parameters of the S-Model are then computed from the estimated link coordinate frame locations. Finally, the Denavit-Hartenberg parameters for the manipulator are extracted from the identified S-Model parameters. We have implemented a complete prototype arm signature identification system and have applied it to identify the signatures and control the end-effector of seven Unimation/Westinghouse Puma 560 robots. Evaluation of the experimental results has demonstrated consistent and significant improvements in the kinematic performance of all the robots tested.