Asymptotic solution of a singularly perturbed nonlinear state regulator problem (Q1898068)

This paper is concerned with the singularly perturbed system \[ \begin{cases} {dx\over dt}= f(t, x, \varepsilon z, \varepsilon)+ b_1 u\\ \varepsilon {dz\over dt}= q(t, x, \varepsilon z, \varepsilon)+ b_2 u\end{cases} \] on the interval \(0\leq t\leq 1\) with variable initial conditions \(x(0, \varepsilon)= x^0(\varepsilon)\), \(z(0, \varepsilon)= z^0(\varepsilon)\) and the quadratic cost functional \[ J(\varepsilon)= \textstyle{{1\over 2}} (E_1 x^2(1, \varepsilon)+ 2\varepsilon E_2 x(1, \varepsilon) z(1, \varepsilon)+ \varepsilon E_3 z^2(1, \varepsilon))+ \] \[ + \textstyle{{1\over 2}}\displaystyle{\int^1_0} [q_1 x^2(t, \varepsilon)+ 2q_2 x(t, \varepsilon) z(t, \varepsilon)+ q_3 z^2(t, \varepsilon)+ u^2(t, \varepsilon) r]dt, \] where \(\varepsilon\) is a small positive parameter, \(E_i\) and \(q_i\) \((i= 1,2,3)\) are constants, \(r\) is a positive constant and the functions \(f\), \(g\) are sufficiently smooth for every variable and have asymptotic power series expansions as \(\varepsilon\to 0\) uniformly in \(0\leq t\leq 1\). The behaviour of the optimal control and corresponding trajectories of the singularly perturbed nonlinear state regulator problem is investigated. Under appropriate hypotheses, it will be possible to complete an asymptotic solution which is uniformly valid when \(\varepsilon\to 0\).