Stability of controllers with on-line computations (Q1189264)

The paper considers the important problem of stability of multivariable nonlinear control loops, in which the nonlinear discrete-time feedback law requires in real-time the solution of a set of nonlinear algebraic equations. Since the iterative algorithm for solving the algebraic system may be time consuming, the dynamics of the solving algorithm is taken into account. The paper focusses on the combined analysis of the dynamics of the system being controlled and the solving algorithm. Since a small sampling period is assumed, a continuous-time formulation is employed to characterize both the solving algorithm and the plant. The main proposition of the paper consists in updating the control law for every sampling instant with the iterate value available in the iterative algorithm at that sampling instant. A singular perturbation formulation and the notion of Lyapunov stability are used to perform stability analysis of the over-all system. Tracking control problems are studied as well as control problems requiring computation of inverse kinematics. The results may be conservative, due to Lyapunov's direct method.