A linear matrix inequality approach for robust control of systems with delayed states (Q1577767): Difference between revisions

Consider uncertain time-delay systems of the form \[ \dot x(t)= [A+\Delta A(t)]x(t)+ [A_d+\Delta A_d(t)] x(t- d(t))+ [B+\Delta B(t)] u(t),\tag{1} \] \[ \partial\leq d(t)\leq\tau, \] \[ \Delta A(t)= \sum^k_{i=1} A_i r_i(t),\;\Delta A_d= \sum^{k_d}_{j=1} A_{dj} r_{dj}(t),\;\Delta B(t)= \sum^{k_b}_{e=1} B_e s_e(t), \] \[ |r_i(t)|\leq\overline r_i,\;i= 1,\dots, k,\;|r_{dj}|\leq\overline r_{dj},\;j= 1,\dots, k_d,\;|s_e(t)|\leq\overline s_e,\;e= 1,\dots, k_b. \] The following inequalities are obtained \[ \Delta A(\Delta A)^T\leq\widehat A^2,\;\Delta A_d(\Delta A_d)^T<\widehat A^2_d,\;\Delta B(\Delta B)^T\leq\widehat B^2, \] \[ \left[\begin{matrix} S_0 & S^T_1 & S^T_2\\ S_1 & -Q_1 & 0\\ S_2 & 0 & -Q_2\end{matrix}\right]< 0,\;\left[\begin{matrix} X & XA^T+ Y^T B^T & 2X\overline A & 2Y^T\overline B\\ AX+ BY & P_{d_1}- \beta_1 I & 0 & 0\\ 2\overline AX & 0 & 2\beta_1 I & 0\\ 2\overline BY & 0 & 0 & 2\beta_1I\end{matrix}\right]\geq 0, \] \[ \left[\begin{matrix} X & XA^T_d & X\overline A_d\\ A_d X & P_{d_2}- \beta_2 I & 0\\ \overline A_d X & 0 & \beta_2 I\end{matrix}\right]\geq 0,\quad i= 1,\dots, N,\tag{3} \] where \[ \begin{multlined} S_0= (A+ A_d)X+X(A+ A_d)^T+ BKX+ XK^T B^T+\\ 2\tau X+ \alpha\widehat A^2+ \beta\widehat A^2_d+ \gamma\widehat B^2+ \sum^2_{i=1} (A_d P_{d_1} A^T_d+ \alpha_i\widehat A^2_d),\end{multlined} \] \[ S_1[X,X,Y^T]^T,\quad Q_1= \text{diag}(\alpha I,\beta I,\gamma I), \] \[ S_2= [A_d P_{d_1}, A_d P_{d_2}]^T,\quad Q_2= \text{diag}(\alpha_1 I- P_{d_1},\alpha_2I-P_{d_2}). \] Theorem: The system (1) is robustly stabilizable if there exist symmetric positive definite matrices \(X\), \(P_{d_1}\), \(P_{d_2}\), and a matrix \(Y\) with appropriate dimension and scalars \(\alpha\), \(\beta\), \(\gamma\), \(\alpha_i\), \(\beta_i\), \(i= 1,2\), satisfying (3). Moreover, a suitable stabilizing control law is given by \(u(t)= YX^{-1} x(t)\).