The robustness of an advanced adaptive controller against system noise

This paper examines the robustness of an advanced adaptive controller against system noise and abrupt parameter changes. Recursive extended least squares (RELS) is incorporated in a self-tuning adaptive controller which uses an enhanced linear-quadratic control algorithm. Simulations are given to demonstrate the effectiveness of the controller in a noisy environment. The system that is being controlled is a piecewise linear system with an independent noise input and with time-scheduled parameter changes unknown by the controller. Effects of noise level are examined along with trade-offs that must be considered regarding probing-signal strength and noise level. The robustness of this self-tuning control is shown to benefit from the use of RELS in place of RLS