Nonlinear and LPV Decoupling Control Strategies for Robotic Manipulators

The thesis arises from the need of the robotics company Comau to address the issue of dynamic inertial coupling between the links of a robotic manipulator. This study aims to analyze the theory and experimentally validate the application of diagonalization techniques for the dynamic model of an industrial robotic manipulator, based on the design of a pre-compensator acting on the robotic plant. This decoupling compensator allows the implementation of independent joint control, treating each link as a SISO subsystem, thereby simplifying the controller synthesis and improving tracking accuracy. In particular, the thesis explores different decoupling strategies in both the time domain and the frequency domain. The study begins with the modeling of robotic manipulators, considering various dynamic representations, including rigid-body models, elastic joint models, and damped joint models, leading to the formulation of the model used in Comau's industrial simulations. For time-domain decoupling, the nonlinear feedback linearization technique is applied to derive a nonlinear control law that effectively decouples the system for different manipulator models. On the other hand, the frequency-domain decoupling approach relies on the design of linear compensators, investigating both static and dynamic techniques, including static, SVD-based decoupling, ideal decoupling, simplified decoupling, and inverted decoupling, with the goal of minimizing interaction effects and enhancing closed-loop performance. Extensive simulations and experimental validations are carried out on a Comau robotic manipulator model to evaluate the effectiveness of each method, highlighting the trade-offs between decoupling, computational complexity and robustness to modeling uncertainties.