Lyapunov design of multivariable MRAC via generalized passivation

Abstract

A Lyapunov design of multiple input, multiple output (MIMO) model-reference adaptive control (MRAC) systems is proposed using new generalized passivity concepts based on WSPR and WASPR properties. The new design avoids control overparameterization and leads to the simplest unnormalized gradient algorithm by means of a passivating multiplier. To this end, a necessary and sufficient condition on the high frequency gain matrix to achieve the WASPR property is established. This condition is referred to as Positive Diagonal Jordan (PDJ) form. Based on the robustness WSPR and WASPR properties, a systematic determination of a robust passivating matrix multiplier is then presented. The advantages of the new design are illustrated by numerical simulation including a robotics adaptive visual servoing problem and a 4-input-4-output dynamic system.