| The objective of this thesis is to develop, implement and analyze the performance of force control algorithms as applied to hydraulic manipulators. This work, which comprises both experimental and simulation studies, is built upon the concept of impedance control technique; this technique accommodates stable transition between unconstrained and constrained motions. Position-based formulation of impedance control is employed here because of the difficulty in implementing controlled torques on hydraulic actuators. First, a nonlinear proportional-integral position controller is developed. It is shown that the controller accurately tracks position trajectories to within 0.2° of error, and regulates static trajectories to within 0.05° on a Unimate MK-II hydraulic manipulator. The controller has these capabilities despite large manipulator inertia, joint suction, valve deadband and other nonlinearities. The positioner is then used to implement the position-based impedance controller (PBIC) to the manipulator. Various tests are performed to show that the PBIC can successfully replace the actual manipulator dynamics with those of the target impedance. Regulation of the end-effector force, however, cannot be achieved using the PBIC alone. Various available methods that potentially enable a PBIC to regulate the force are identified, studied, and modified as required. They are then applied, for the first time, to the hydraulic manipulator to improve the force control capability of the PBIC. The improvements thus made are twofold: through estimation of environmental parameters and through modification of the position reference. Accurate estimation of the environmental parameters is shown to be possible through an off-line identification procedure. The on-line identification method, which is built upon minimization of the force prediction error, gives estimates that do not truly reflect the nature of the environment. The combination of these estimates, however, can modify the position reference toward accurate force regulation. It is also shown that the estimation part can be avoided by directly modifying the position reference to minimize the force error. |
