A class of robust low-order controllers: a synthesis method (Q1882014)

In this important paper a new method of reduced order robust dynamical output feedback is proposed. By virtue of this method, robust controllers are described by linear differential equations dependent on two-positive-definite matrices (the solutions to two Riccati algebraic equations). A class of robust controllers for \(H^\infty\)-control and absolute stabilization for the system \[ \dot{\mathbf x}= A{\mathbf x}+ B_1{\mathbf u}+ B_2\xi,\quad {\mathbf y}= Q{\mathbf x}, \] is described. Here \({\mathbf x}\in\mathbb{R}^n\) is the state vector, \({\mathbf u}\in\mathbb{R}^m\) is the control, \(\xi\in\mathbb{R}^e\) is the disturbance, and \({\mathbf y}\in\mathbb{R}^k\) is the output vector. The process \(\overline\xi(t)\) may depend on some variable \(z({\mathbf x})\). The object is assumed to be stabilized when there is a matrix \(S\) such that the matrix: \(A_1= A+ B_1S\) is stable. Main result: Under given stability conditions, there exists a quadratic function \(V({\mathbf x})\) (the Bellman function) for the conditional minimax problem \[ V(h)= \max_\xi\,\min_{\mathbf v}\,\int^\infty_0 F({\mathbf x}(t),{\mathbf v}(t), \overline\xi(t))\,dt,\quad \xi\in L^1_2,\quad{\mathbf v}\in L^m_2 \] under the condition \[ \dot{\mathbf x}= A_1{\mathbf x}+ B_1{\mathbf v}+ B_2\overline\xi,\;{\mathbf x}(0)={\mathbf h},\text{ where }{\mathbf v}(t)={\mathbf u}(t)- S{\mathbf x}(t)={\mathbf u}(t)- S_1{\mathbf x}(t)- D{\mathbf y}(t). \] The proposed method, which the author applies to a system of two elastically connected masses, shows that the equations of such a controller contain independent parameters (not dependent on the initial parameter of the problem and on the solutions to the Riccati equation), owing to which the frequency conditions defining the class of robust controllers are satisfied. Furthermore, the given examples illustrate that the parameters in the controller equation depend on the decomposition of the \(n\)-dimensional space into a direct sum of spaces, one of which is the space of output variables.