Stablewise absolute output natural tracking control with finite reachability time: MIMO Lurie systems (Q2468036)

The authors study a linear system \(\dot{x} = Ax + B f(u) + Dd, y = Cx + Ed\) with an input nonlinearity \(f\) which is assumed to be invertible and unknown. They define the system property ``(natural) global absolute trackable'' which basically means that for every initial condition, for every non-linearity, for every reference signal \(y_d(\cdot)\), and for every settling time \(\sigma>0\) there exists an input such that the system's output equals the reference signal after the finite settling time, i.e.\ \(e(t):=y_d(t)-y(t)=0\) for all \(t\geq\sigma\). This trackability property is called ``natural'' if additionally the controller does not depend on the matrix \(A\), the non-linearity \(f\), and the disturbance \(d\). All results of the paper seem to be based on ``control laws'' of the form \(u(t) = u(t-) + g(t)\), where \(u(t-):=\lim_{\varepsilon\to 0+} u(t-\varepsilon)\) and \(g(\cdot)\) is some function depending on the observed error \(e(\cdot)\). However, only continuous control inputs are considered in the paper which always yields \(u(t)=u(t-)\), hence the considered control law is mathematically meaningless or even contradictory. Furthermore the authors claim that ``in case \(e(t)=0\) over some interval \([\tau_1,\tau_2]\subseteq\mathbb{R}_+\) then \(u(t)=u(\tau_1)\) for all \(t\in[\tau_1,\tau_2]\)''. Assuming that for a given system a controller is found which ensures exact tracking after a finite time, i.e.\ \(e(t)=0\) for all \(t\geq \sigma\), this statement would yield that \(u(\cdot)\) is constant on \([\sigma,\infty)\). But clearly, a constant input signal can't assure exact tracking of a general reference signal. Altogether, it seems highly questionable whether the results of the paper are correct. The result might hold in a time-delay setup (which was used in the simulations) but for this a complete reformulation of the results is necessary.