Disturbance compensation in continuous time by sampled data control
The authors consider the linear plant \[ \dot x= Ax+ Bu,\quad x(0)= x_0,\quad y= Cx+v\tag{1} \] which is subject to external disturbances whose effects on the output are represented by \(v\). The objective is to zero the output in continuous time by sampled data control. A disturbance \(v(t)\) can be compensated, if there exists an input \(u(t)\) to the plant such that its output reduces to \(y(t)= Ce^{At}z\) for some \(z\in\mathbb{R}^n\). The set of all disturbances which can be compensated in continuous time is characterized and one defines a disturbance model \[ \dot\alpha= 0,\quad v(t)= V(t)\alpha\tag{2} \] which covers a general subset of such disturbances. A sampled data controller can be composed of the following two parts: a discrete-time observer \[ \widehat z(t_{i+1})= \phi_i\widehat z(t_i)+ \Gamma_i(u)+ H_i[\widehat y(t_i)- y(t_i)],\quad \widehat y(t_i)= \Delta_i\widehat z(t_i)\tag{3} \] whose state is updated at time instants \(t_i\in [kT- T,kT)\) to obtain \(\widehat z=\left[\begin{smallmatrix} \widehat x\\ \widehat\alpha\end{smallmatrix}\right]\) at time \(kT\) and a feedback law for \(\tau\in [0,T)\), \[ u(kT+ \tau)= U(kT+\tau)\widehat\alpha(kT)+ K[\widehat x(kT)- X(kT)\widehat\alpha(kT)].\tag{4} \] The main result states that if \(K\) is chosen such that all eigenvalues of \(F+GK\) are in the open unit disk, where \(F:= e^{AT}\), \(G:= \int^T_0 e^{A(T-\tau)} Bd\tau\), and \(H_i\), \(i\in Z^+_0\) is chosen such that \[ (\widehat z-z)_{t_{i+ 1}}= [\phi_i+ H_i\Delta_i] (\widehat z- z)_{t_i} \] is asymptotically stable, with transient sufficiently fast such that \(U_{kT+\tau}(\widehat\alpha- \alpha)_{kT}\to 0\) and \(X_{kT+\tau}(\widehat\alpha- \alpha)_{kT}\to 0\) on \(\tau\in [0,T]\) as \(k\to\infty\), then the sampled data control system (1), (2), (3), (4) is asymptotically stable, and the disturbance is compensated in continuous time, i.e. \(y(t)\to 0\) as \(t\to\infty\). Summarizing, the authors show that whenever a class of disturbances can be compensated in continuous time, then it can also be compensated in continuous time by means of sampled data control. With a hold function chosen for perfect compensation, the choice of sampling patterns becomes essentially arbitrary and independent of the disturbances to be compensated.
This page was built for publication: Disturbance compensation in continuous time by sampled data control
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4373065)
