A generalized approach for the control of constrained robot systems

Jae Wook Jeon, Purdue University

Abstract

This thesis presents a generalized approach for controlling various cases of the constrained robot system. To accomplish specific tasks successfully by a constrained robot system, both the constrain forces/torques and the motion of the manipulator end-effector must be specified and controlled. Using the Jacobian matrix of the constraint function, the generalized coordinates of the constrained robot system can be partitioned into two sets; this leads to partitioning the constrained robot system into two subsystems. This decomposition of the constraint robot system into subsystems leads to the design of a nonlinear position-force controller with a simple structure which takes the constraints into consideration. Applying the proposed nonlinear position-force controller to each subsystem and using the relation between the motion-independent forces/torques in the subsystems, we can show that both the errors in the manipulator end-effector motion and the constraint forces/torques approach zero asymptotically. Typical examples of the constrained robot systems are analyzed and discussed. Computer simulations are conducted to verify the validity and performance of the proposed approach. The above generalized approach is extended to controlling constrained robot systems in which the manipulators are redundant. To discuss the effect of the redundancy resolution on the manipulator dynamics and its torque bound, the inverse kinematics in the redundant manipulator has to be resolved at the joint acceleration level. In order to avoid instability due to high joint velocities, a kinetic redundancy resolution technique which locally optimizes a kinematic criterion to reflect the changes of the manipulator configuration is proposed. The manipulability measure is selected as the kinematic criterion in the local redundancy resolution. Computer simulations are performed on a three-link planar rotary manipulator to verify the performance of the proposed kinetic redundancy resolution and to compare its performance with three existing local kinetic redundancy resolution techniques for various straight-line trajectories. Then, to control constrained redundant robot systems, two formulations are proposed to combine this redundancy resolution technique with the above generalized approach in the joint-variable space as well as in the Cartesian space. (Abstract shortened with permission of author.)

Degree

Ph.D.

Advisors

Lee, Purdue University.

Subject Area

Electrical engineering

COinS