This article describes the design of an observer based robust linear output feedback controller for the regulation and output reference trajectory tracking tasks in switched ‘buck’ converter circuits feeding a completely unknown time-varying load. The state-dependent perturbation effects of the unknown load resistance are on-line estimated by means of a generalised proportional integral (GPI) observer, which represents the dual counterpart of GPI controllers introduced in Fliess, Márquez, Delaleau and Sira-Ramírez (Fliess, M., Márquez, R., Delaleau, E., and Sira-Ramírez, H. (2002), ‘Correcteurs Proportionnels-intégraux Géneralisés’, ESAIM: Control, Optimisation and Calculus of Variations, 7, 23–41). The reconstructed perturbation complements the controller in a cancellation effort which allows the core of the feedback controller to become a traditional proportional derivative (PD) controller. The designed average feedback controller is then implemented via a sigma–delta-modulator, which effectively translates the designed continuous average feedback control input signal into a discrete valued switched input signal driving the converter's input switch and preserving all relevant features of the average design. The Appendix collects some generalities about GPI observers.

In this article, the problem of robust trajectory tracking, for a parallel robot is tackled via an observer‐based active disturbance rejection controller. The proposed design method is based on purely linear disturbance observation and linear feedback control techniques modulo nonlinear input gain injections and cancellations. The estimations are carried out throughGeneralizedProportionalIntegral (GPI) observers, endowed with output integral injection to ease the presence of possible zero mean measurement noise effects. As the lumped (both exogenous and endogenous) disturbance inputs are estimated, they are being used in the linear controllers for on‐line disturbance cancellation, while the phase variables are being estimated by the sameGPIobserver. The estimations of the phase variables are used to complete a linear multivariable output feedback controller. The proposed control scheme does not need the exact knowledge of the system, which is a good alternative to classic control schemes such as computed torque method, reducing the computation time. The estimation and control method is approximate, ensuring small as desired reconstruction and tracking errors. The reported results, including laboratory experiments, are better than the results provided by the classical model‐based techniques, shown to be better when the system is subject to endogenous and exogenous uncertainties.