A vector space approach to the identification of the dynamic parameters of robot manipulators and interacting tasks.

Abstract

An exact dynamic model and the real-time computation of actuating torques of a manipulator are required in model-based control schemes for implementing trajectory control. However, the dynamic parameters (mass, center of mass, moment of inertia, and joint frictional forces) of the dynamic model are not always exactly known. Furthermore, even if the dynamic model is identified, simplification of the dynamic model is required for computing the actuating torques in real-time. This dissertation offers a new perspective on the identification and simplification of the dynamic model of manipulators and interacting tasks by using a vector space approach. The dissertation first presents an estimation method for identifying the unknown dynamic parameters. The method is based on the theorem of conservation of energy, which makes it equally applicable to serial-link manipulators as well as manipulators containing closed kinematic loops. The method is shown to be conceptually simple, computationally efficient, and easy to implement. Motor torques, joint positions, and joint velocities are the only inputs required by the estimation algorithm. The second part of the dissertation deals with the identification of the independent parameters. Due to the geometric constraints of the manipulator mechanism, not all dynamic parameters contribute to manipulator dynamics; therefore, they are not completely identifiable. By finding the essential and the independent parameter spaces, the estimation method is adapted to identify the dynamic parameters in these parameter spaces. Although both essential and independent dynamic parameters are completely identifiable, the independent dynamic parameters are used to formulate dynamic equations which require less computational load. By applying persistently exciting input trajectories, the independent dynamic parameters are completely identified. Finally, the parameter space analysis is extended to further simplify the dynamic model and to improve manipulator control. The essential parameter space basis vectors which have insignificant effect on manipulator dynamics are eliminated to simplify the manipulator model. To improve manipulator control, we present a method for determining a set of dynamic parameters to create a manipulator with nearly decoupled and configuration independent dynamics. The manipulator can then track desired trajectories accurately at high speed with a simple control scheme.