No label defined (Q3836710)

The interest of the author is focused on the control problem of characteristic vectors for the linear systems, denoted by \((A,b)\), having the form: \[ \dot x= A(t)x+ b(t)u,\quad t\geq t_0, \] where \(A(t)\) is a real continuous \(n\times n\)-matrix, \(b(t)\) is a real \(n\)-vector column with continuous entries and \(u= u(t)\) is a control ranging in \(\mathbb{R}\). For a continuous function \(f(t)\) ranging in \(\mathbb{R}^n\) with the finite characteristic exponent \(\alpha_0= \varlimsup_{t\to\infty} (1/t)\ln|f(t)|\), there are defined the quantities: \[ \alpha_1= \varlimsup_{t\to\infty}\ln\{|f(t)|e^{-\alpha_0 t}\}/\ln t,\;\alpha_2= \varlimsup_{t\to\infty} \ln\{|f(t)|e^{-\alpha_0 t}t^{-\alpha_1}\}/\ln\ln t,\dots, \] \[ \begin{multlined} \alpha_s= \varlimsup_{t\to\infty} \ln\{|f(t)|e^{-\alpha_0t} t^{-\alpha_1}(\ln t)^{-\alpha_2}\cdots (\ln_{s-2} t)^{-\alpha_s- 1}\}/\ln_s t,\\ \ln_s t= \underbrace{\ln\ln\cdots\ln t}_s\quad (s= 3,\dots, p).\end{multlined} \] The vector \(\alpha= \alpha[f]= (\alpha_0,\alpha_1,\dots, \alpha_p)\) is referred to as the \(p\)th-order characteristic vector of \(f(t)\) and there is introduced an order relation \(\succ\) on \(\mathbb{R}^{p+1}\) as follows: \(\alpha\succ\beta\) if there exists an index \(s\), \(1\leq s\leq p\), such that \(\alpha_s> \beta_s\) and \(\alpha_i= \beta_i\), \(i= 1,\dots, s-1\). It is said that the problem of control of characteristic vectors is solvable for system \((A,b)\) if for any integer \(p\geq 0\) and any set \(\alpha^1,\alpha^2,\dots, \alpha^n\) consisting of \(n\) elements of \(\mathbb{R}^{p+1}\) such that \(\alpha^1\prec \alpha^2\prec\cdots\prec \alpha^n\), there exists a control \[ u= f(t)x\tag{1} \] with a continuous row \(n\)-vector \(f(t)\) such that the closed equation \[ \dot x= (A(t)+ b(t) f(t))x\tag{2} \] has a principal solution system \(x_1(t),\dots, x_n(t)\) with \(\alpha[x_j]= \alpha^j\), \(j= 1,\dots, n\). On the other hand, it is defined a special class \({\mathcal B}_n\) consisting of all pairs \((A,b)\in B_n\) such that the coordinates \(g_i(A,b)(t)\), \(i= 1,\dots, n\) of the vector column \(g(A,b)(t)\) are of class \(C^{i-1}\), respectively. Here, \(B_n\) is the set of pairs \((A,b)\) for which the vector functions \(q_0(t)= b(t)\), \(q_i(t)= A(t)q_{i-1}(t)- \dot q_{i-1}(t)\), \(i= 1,\dots, n-1\), are defined and continuous, \(Q(t)\) is the so-called controllability matrix consisting of the columns \(q_0(t),q_1(t),\dots, q_{n-1}(t)\), and \(g: B_n\to C(\mathbb{R}^n)\) is defined by \(g(A,b)(t)= Q(t)^{-1}q_n(t)\), with \(q_n(t)= A(t) q_{n-1}(t)- \dot q_{n-1}(t)\). The main result states that if \((A,b)\in{\mathcal B}_n\), then the problem of control of the characteristic vectors is solvable. Moreover, the vector coefficient \(f(t)\) occurring in the control (1) solving this problem can be chosen in a way such that the closed system (2) is Lyapunov reducible.