Disturbance attenuating output-feedback control of nonlinear systems with local optimality (Q5935475): Difference between revisions

An output-feedback control law that achieves local near-optimality and semiglobal inverse \(\mathcal H_\infty\)-optimality is obtained for a class of nonlinear systems with additive uncertainty: \[ \left\{\begin{aligned} \dot x&=Ax + f(x_1) +G(x_1)w+ Bu,\\ y&=x_1, \end{aligned}\right. \] where \(x\in\mathbb R^n\) is the state, \(w\in\mathbb R^q\) is the disturbance, \(u\in\mathbb R\) is the control, \(y\in\mathbb R\) is the output, \(B = [0'_{n-i} 1]'\), \(A\) is such that its \(i\)th row is \([a_{[i]} 1 0'_{n-i-1}]\). The vector functions \(f(x_1)\) and \(G(x_1)\) are supposed to be sufficiently smooth. Such systems have nonlinearities that depend solely on the measured output. The design procedure is presented and followed by an example. The design combines the locally optimal backstepping procedure with the cost-to-come methods for robust estimation. If the derivative of the output measurement is also available for feedback, then the same control law with a reduced-order observer is shown (in the benchmark problem) to achieve local \(\mathcal H_\infty\)-optimality and global inverse \(\mathcal H_\infty\)-optimality. Unlike standard linear \(\mathcal H_\infty\)-designs which penalize the actual state of the system, the considered performance index penalizes the estimation error and the estimate of the state. When the penalty matrices are chosen carefully, the optimal controller which corresponds to the modified performance index is shown to be suboptimal with respect to the standard \(\mathcal H_\infty\)-performance index.