Improved derivative-dependent control of stochastic systems via delayed feedback implementation (Q2207240): Difference between revisions

The authors study derivative-dependent control of \(n\)th-order stochastic systems where derivatives are not available for measurements. Two methods are presented for continuous-time delayed static output-feedback and its sampled-data implementation. The first is a stochastic extension of the deterministic method by \textit{A. Selivanov} and \textit{E. Fridman} [Automatica 98, 269--276 (2018; Zbl 1406.93265)], the problem being that in the latter the Lyapunov functionals depend on the \(n\)th order derivative and can therefore not be used in the stochastic case, as a solution of a stochastic system does not have a derivative. So, new Lyapunov functionals are introduced which depend on the deterministic and stochastic parts of the system. The second method is based on the neutral type model transformation with derivatives of order \(i=1,\dots,n- 1\) presented in the form of the finite difference, the remainder in the form of the time-derivative of the integral term. The latter depends on the same \(i\)th derivative and not the \((i+1)\)th and is therefore well-defined in the stochastic case. Now either augmented or simple, i.e. non-augmented, Lyapunov functionals are used, as proposed for second-order systems by \textit{E. Fridman} and \textit{L. Shaikhet} [Syst. Control Lett. 124, 83--91 (2019; Zbl 1408.93136)]. It turns out that for larger stochastic perturbations the second method with simple Lyapunov functionals leads to less conservative results and fewer decision variables in LMIs than the first method. With augmented functionals the second method leads to better results than with simple ones and to better results than the first method, the price being higher computational complexity. In the sampled-data case, the second method leads to more conservative results if the stochastic perturbations are small, but it allows larger stochastic perturbations even with simple Lyapunov functionals. Several numerical examples are given.