Trading robustness with optimality in nonlinear control (Q5949060)

The authors consider a class of nonlinear systems described by equations of the form \[ x'=f(x)+g_1(x)w+g_2(x)u,\quad z_2=h_2(x)+k_2(x)u,\quad z_{\infty}=h_{\infty}+k_{\infty}u. \] The first equation describes a plant defined in \(\mathbb R^n\) with state \(x\), control input \(u\in\mathbb R^{m_1}\), and exogenous input \(w\in\mathbb R^{m_2}\). The second and the third equations define two performance variables \(z_2\in\mathbb R^{p_1}\) and \(z_{\infty}\in\mathbb R^{p_2}\), which do not depend on the exogenous input. The mixed \(H_2/H_{\infty}\) control problem consists in finding an admissible state-feedback control law \(u=u(x)\) solving the constrained optimization problem \[ \min_{u} \int_0^{\infty}\|z_2(x(t))\|^2dt\quad\text{subject to} \int_0^{\infty}\|z_{\infty}(x(t))\|^2dt-\gamma_2 \int_0^{\infty}\|w(t)\|^2dt\leq 0. \] A parameterized family of stabilizing control laws is constructed in terms of the solution to a parameterized Hamilton-Jacobi equation. A procedure to construct the sub-optimal solution is presented and studied in detail. A simple one-dimensional example is discussed also in detail, and applications to the optimal robust stabilization of the angular velocities of a rigid body are presented.