SUMMARY

Modern Industrial robots are designed to be highly repeatable (< 0.1 mm) but not as globally accurate (<2 mm). Global accuracy, however, is necessary for tasks where it is not convenient to “teach” the robot the set of poses it needs to run through to perform the task. In addition, some of these tasks, like machining, may involve high time-varying external forces which cause the robot to deflect and its accuracy to suffer further. This dissertation examines modeling and control strategies for the purpose of improving the global accuracy of the robot for manufacturing tasks including machining. First, a comparison of stiffness modeling techniques is performed to examine when it is important to account for the structural dynamics of the robot, versus when static stiffness calibrations are sufficient. Next, a new method of performing a highly accurate state estimation of the robot end-effector by combining instantaneous inertial and pose measurements is presented. Finally, a new way of incorporating the manipulator dynamics into a control system model which allows for offline closed-loop stability prediction is developed and validated.