Locally optimal stabilization under incomplete information (Q1290216)

The paper deals with nonlinear discrete-time systems of the form \( x_{t+1}=Ax_t+Bu_t+\xi _t+\Psi (x_t,u_t,\xi _t)\), \(y_t=Cx_t+\Gamma (x_t)\), \(t=0,1,2,\ldots\), where \(x_t\in R^n\) is the state, \(u_t\in R^m\) is the control, \(y_t\in U\subset R^s\) is the measured output and \(\xi _t\in \Theta \subset R^n\) is an unknown disturbance. \(x=0\), \(u=0\), \(\xi =0\) is supposed to be an equilibrium. For the designer of the controller, only an approximate linear model \(x_{t+1}=Dx_t+Fu_t+\xi _t\), \(y_t=Gx_t\) is available. A notion of asymptotic robustness is introduced. The proposed algorithm of stabilization is based on certain estimations of the disturbance and the state. It is proved that, under suitable conditions, this algorithm is asymptotically robust. In order to weaken the conditions, a modified \(k\)-step version of the algorithm is also developed.