Suboptimal control of rigid body motion with a quadratic cost (Q1384042)

This paper deals with the attitude control of a rigid body in \(R^3\), subject to three control torques exerted about its principal axes of inertia. The control system assumes the following form \[ \begin{cases} \dot\rho & = G(\rho)\omega,\\ \dot\omega & = F(\omega)\omega+ J^{-1}u,\end{cases} \] where \(\rho= (\rho_1, \rho_2, \rho_3)\) are Cayley parameters of the special orthogonal group, \(\omega= (\omega_1, \omega_2, \omega_3)\) denotes the vector of angular velocities, \(u= (u_1, u_2, u_3)\) stands for the control, and matrices \(G\), \(F\), \(J\) are given. When defining, for suitable matrices \(C\), \(D\), a performance output \[ z= C{\rho\brack \omega}+ Du, \] the control problem addressed in the paper is formulated as follows: For any bounded set \(\Delta\subset R^6\) containing the origin, and any positive number \(\gamma\), find a static feedback \(u= k(\rho,\omega)\) such that the closed loop system is asymptotically stable within the region \(\Delta\), and for every initial state \((\rho_0, \omega_0)\in\Delta\) the performance index remains bounded by \(\gamma\), i.e. \[ J(\rho_0, \omega_0, u(\cdot))= \int^\infty_0 \| z(t)\|^2 dt\leq\gamma. \] The main result (Theorem 5.1) provides sufficient conditions for the solvability of this problem. Computational techniques for obtaining the best suboptimal controller require solving a sequence of linear matrix inequalities.