Robust observer-based nonlinear computed-torque design via linear matrix inequalities

Robustness of the well-known computed-torque technique is twofold improved in this paper: on the one hand, the inner-loop control law is made exclusively dependent on user-generated signals whose accuracy is no longer affected by noise or numerical errors; on the other hand, the outer-loop control law is based on available positions and observer-based estimations of the velocities. Both the controller and the observer are nonlinear structures designed via linear matrix inequalities arising from the application of a recently appeared factorization. Asymptotic convergence of the tracking and estimation errors is guaranteed via Lyapunov-based analysis. The proposal is put at test in a variety of Lagrange-Euler systems where advantages over standard computed-torque techniques are confirmed, both in simulation and real-time setups. © 2024 Universidad Politecnica de Valencia.. All rights reserved.