Use of Disturbance Measurement Feedforward in LQG Self-tuners

Mikael Sternad

International Journal of Control, vol 53, pp 579-596, March 1991.

Paper available in Pdf

Outline:

Disturbances which act on a process can sometimes be measured and used by a controller. The use of such disturbance measurement feedforward can lead to large performance and robustness improvements, as compared to the use of feedback only. The paper discusses the design of indirect self-tuning scalar LQG control laws, which utilize feedforward.

Abstract:

An explicit adaptive regulator with disturbance measurement feedforward is presented, based on a polynomial LQG design. The addition of an optimized feedforward filter to a feedback regulator involves the solution of only one additional linear polynomial equation.

The regulator is designed to handle shape-deterministic disturbances, such as steps, ramps and sinusoids, as well as stochastic disturbances. The properties of the off-line solution in the case of unstable disturbance models are explained.

Computational aspects, the computational complexity and the robustness against unmodelled dynamics are discussed. It is argued that the use of feedforward can improve not only the disturbance rejection, but also the stability robustness of an LQG feedback regulator.

Related publications:

Book Chapter which includes the material of the paper, available in Postscript.
Conference paper in IFAC ACASP'89, on the above theme.
Paper in Automatica 1988, on the polynomial (non-adaptive) design of these regulators.
Conference paper on indirect adaptive LQG, explicit criterion minimization and self-tuners for adaptive feedforward control.
Lecture notes of an undergraduate course, where these adaptive algorithms are utilized successfully.