On Disturbance-to-State Adaptive Stabilization without Parameter Bound by Nonlinear Feedback of Delayed State and Input

TL;DR

This paper introduces a novel nonlinear feedback control method using distributed delays to achieve disturbance-robust adaptive stabilization for scalar linear systems with unknown parameters and unbounded disturbances, without requiring parameter bounds or persistent excitation.

Contribution

It provides the first solution to disturbance-robust adaptive control without parameter bounds using infinite-dimensional nonlinear feedback with delay, ensuring boundedness, stability, and finite-time parameter estimation.

Findings

Global boundedness of the closed-loop state with disturbance

Practical input-to-output stability with assignable gain

Finite-time exact parameter estimation without disturbance

Abstract

We complete the first step towards the resolution of several decades-old challenges in disturbance-robust adaptive control. For a scalar linear system with an unknown parameter for which no a priori bound is given, with a disturbance that is of unlimited magnitude and possibly persistent (not square integrable), and without a persistency of excitation necessarily verified by the state, we consider the problems of (practical) gain assignment relative to the disturbance. We provide a solution to these heretofore unsolved feedback design problems with the aid of infinite-dimensional nonlinear feedback employing distributed delay of the state and input itself. Specifically, in addition to (0) the global boundedness of the infinite-dimensional state of the closed-loop system when the disturbance is present, we establish (1) practical input-to-output stability with assignable asymptotic gain from the disturbance to the plant state; (2) assignable exponential convergence rate; and (3) assignable radius of the residual set. The accompanying identifier in our adaptive controller guarantees (4) boundedness of the parameter estimate even when disturbances are present; (5) an ultimate estimation error which is proportional to the magnitude of the disturbance with assignable gain when there exists sufficient excitation of the state; and (6) exact parameter estimation in finite-time when the disturbance is absent and there is sufficient excitation. Among our results, one reveals a tradeoff between "learning capacity" and "disturbance robustness:" the less sensitive the identifier is to the disturbance, the less likely it is to learn the parameter.