A note on overall observability: The joint problem of state estimation and parameter identification (Q1328461)

Nonlinear systems are considered having the form \[ \dot x(t)= Ax(t)+A_ 1 \alpha+ N(x(t), \alpha), \quad x(0)= x_ 0\in \mathbb{R}^ n, \quad y(t)= Cx(t), \tag{1} \] where \(A\in {\mathcal L} (\mathbb{R}^ n)\), \(C: \mathbb{R}^ n\to \mathbb{R}^ q\) and \(A_ 1: \mathbb{R}^ p\to \mathbb{R}^ n\) are linear operators and \(N\) is nonlinear. The joint problem of state estimation and parametric identification is to construct the state \(x(t)\), \(t\in [0,T]\) and identify the \(p\) unknwon constant parameters \(\alpha= (\alpha_ 1,\dots, \alpha_ p)^ T\in \mathbb{R}^ p\) for a given observation \(y\in {\mathcal L}^ 2 (0,T; \mathbb{R}^ q)\). This problem is transformed into just state estimation of the overall system obtained from (1) by considering the enlarged state \(\overline {x}(t)= [x(t)^ T \alpha^ T]^ T\). Necessary and sufficient conditions for linear overall observability of the system (1) (i.e. observability of the overall system) are given. For overall observable systems the nonlinear state estimation problem is converted into one of obtaining a fixed-point for an associated mapping and this approach is extended to solve the initial problem.