Robust Control of Robots Using Only Position Measurements

摘要

In robot applications, the performance of state-feedback controllers maysignificantly be degraded by the poor quality of velocity measurements. The paper considers the control problem of robots using position measurements only. For this purpose, a model-based controller-observer combination is presented that originates from the passivity approach to robot control. Characteristic for this approach is that it exploits the physical structure of the robot system, and consequently, yields control methods that possess certain robustness properties. A robustness analysis indeed shows that the proposed controller-observer combination is able to face both inaccuracies in the rigid robot model, and bounded input disturbances. In particular, in the presence of such uncertainties the authors prove the closed-loop error dynamics to be locally uniformly ultimately bounded. Experimental tests on a two DOF manipulator system illustrate the robustness properties of the proposed controller-observer system, and show the improved tracking performance compared to the in practice frequently used numerical position differentiation technique for obtaining a velocity estimate.