Robust adaptive control in Hilbert space (Q914629): Difference between revisions

This paper considers stabilization and regulation problems for infinite- dimensional systems. It is assumed that the system evolves according to the state space model \(\dot x=Ax+Bu\) where A is a possibly unbounded operator on a Hilbert space X (this generality is shown to be necessary in an example). The parameters of the system are not given, so that the control problem is set up as a variant of model reference adaptive control (MRAC). The authors state that previous control strategies rely crucially on the assumption that the state space X is finite. This is only partially correct since many papers have appeared in the robust adaptive control literature which prove stability in the infinite-dimensional case. However, it is true that only this paper and a few of its reference explicitly take into account the infinite-dimensionality of the problem in the construction of a control law, and in the proof. In MRAC the goal is to choose a control law which attempts to match the plant to be controlled with a given known system model. Standard MRAC is not easily analysed in the infinite-dimensional framework since one essentially assumes that the order of the system is finite and known. To overcome this difficulty it is assumed that an infinite-dimensional model exists which produces the same outputs as the given system model (the command generator tracker (CGT)). This is an impossible goal to meet in general. But it is shown in the paper that if the class of inputs is restricted to only time-constant functions, then this matching condition is easily obtained. Hence the results of the paper apply to the regulation problem. It is claimed that the generalization to piecewise constant inputs is straightforward. The main new insight in the paper is that the positive real condition normally imposed is replaced by a new ``near strict positive real condition'', which is the original condition when a fictitious feedthrough term is introduced. Under the assumption that solutions to the resulting differential equations exist (which may be a nontrivial problem due to discontinuities in the RHS of the resulting differential equations) it is shown that the closed loop system is stable in a mean square sense, where the steady state error is proportional to the magnitude of the fictitious feedthrough term introduced. Robustness properties of the algorithm, and an application to the beam equation are also discussed.