On Feedforward Adaptation

Abstract: Adaptation of a feedforward term in a two-degrees-of-freedom servo controller for single-input single-output systems is investigated in this thesis. The first part ofthe thesis gives a clarifying analysis of feedback-error learning in a linear and time-invariant framework. It leads to the suggestion of a novel scheme that guarantees boundedness of all signals in the system; something that previous considered schemes cannot guarantee. The second part is devoted to the adaptation of a feedforward gain in a linear and time-invariant structure. Stability and convergence are here treated for an adaptive law without the assumption of a model-matching condition, and possible adaptive techniques are discussed and examined in simulations. The last part of the thesis contains a preliminary investigation that treats the problem to adapt a nonlinear feedforward function. The function corresponds to the steady-state relation between the input and the output for a nonlinear plant. The update law in this last part is inspired from earlier parts, whilde the conclusions rely heavily on simulations and intuitive reasoning.