Direct Adaptive Control Methodologies for Flexible-Joint Space Manipulators with Uncertainties and Modeling Errors

This work addresses the direct adaptive trajectory tracking control problem associated with lightweight space robotic manipulators that exhibit elastic vibrations in their joints, and which are subject to parametric uncertainties and modeling errors. Unlike existing adaptive control methodologies, the proposed flexible-joint control techniques do not require identification of unknown parameters, or mathematical models of the system to be controlled. The direct adaptive controllers developed in this work are based on the model reference adaptive control approach, and manage modeling errors and parametric uncertainties by time-varying the controller gains using new adaptation mechanisms, thereby reducing the errors between an ideal model and the actual robot system. More specifically, new decentralized adaptation mechanisms derived from the simple adaptive control technique and fuzzy logic control theory are considered in this work.;Finally, this work considers the partial state feedback problem related to flexible-joint space robotic manipulators equipped only with sensors that provide noisy measurements of motor positions and velocities. An extended Kalman filter-based estimation strategy is developed to estimate all state variables in real-time. The state estimation filter is combined with an adaptive composite controller, to provide a closed-loop adaptive partial state feedback control scheme for flexible-joint manipulators. In addition to noise, measurement bias is a detrimental characteristic associated with motor encoders and tachometers. Therefore, this work also demonstrates that the state observation approach developed herein can be modified to compensate for unknown sensor biases.;Numerical simulations compare the performance of the adaptive controllers with a nonadaptive and a conventional model-based controller, in the context of 12.6 m ×x 12.6 m square trajectory tracking. To validate the robustness of the controllers to modeling errors, a new dynamics formulation that includes several nonlinear effects usually neglected in flexible-joint dynamics models is proposed. Results obtained with the adaptive methodologies demonstrate an increased robustness to both uncertainties in joint stiffness coefficients and dynamics modeling errors, as well as highly improved tracking performance compared with the nonadaptive and model-based strategies.