Necessary and sufficient conditions for adaptive stabilizability of jump linear systems (Q1429278): Difference between revisions

The authors consider the problem of stochastic stabilizability of a class of linear discrete-time systems with Markovian jumps and mode-dependent time delays described by \[ x_{k+1}= A(r_k) x_k+ B(r_k) u_k+ w_{k+1},\tag{1} \] where \(x_k\in\mathbb{R}^n\), \(u_k\in \mathbb{R}^m\) and \(w_{k+1}\in \mathbb{R}^n\) are the state, control and noise vectors, respectively, at instant \(k> 0\); \(\{r_k, k\geq 0\}\) is an unobservable, non-reducible, non-periodic and homogeneous Markov chain that takes values in a finite state space \(\{S= 1,2,\dots, N\}\) with transition probability matrix function \(p_{ij}= P\{r_{k+1}= j\mid r_k= i\}\), where \(p_{ij}\geq 0\) and \(\sum^N_{j=1} p_{ij}= 1\) for all \(i,j\in S\). The matrices \(A(r_k)\) and \(B(r_k)\) are constant matrices of appropriate sizes for any fixed values of \(r_k\) in \(S\) such that there exists a matrix \(L\in\mathbb{R}^{m\times n}\), \[ \text{det}[(A(i)- A(j))- (B(i)- A(j))L]\neq 0,\;\forall i\neq j,\;1\leq i,\,j\leq N \] and \(\{w_k\}\) is a martingale difference sequence which is independent of \(\{r_k\}\) and satisfies \[ aI\leq E[w_k w^T_k],\;E[w^T_k w_k]\leq b,\quad\forall k\geq 0, \] where \(a\) and \(b\) are two positive constants, \(I\) is the identity matrix. For system (1) the authors use the adaptive linear feedback control law \(u_k= K(r_k)x_k\), where the control gain \(K(r_k)\) is determined for each system mode \(r_k\in S\). To obtain necessary and sufficient conditions for adaptive stabilizability, the authors use algebraic coupled Riccati matrix equations techniques. They also propose a constructive method for designing the stabilizing controller. Unfortunately, there are not illustrative examples.