Higher order numerical schemes for affinely controlled nonlinear systems (Q5952128)

Numerical schemes for affinely controlled systems received a special interest because complex nonlinear control systems do not allow, in general, an analytic solution, thus requiring numerical treatment for both analyses and controller design. The present paper develops a systematic method for the derivation of high-order schemes for affinely controlled nonlinear systems. The proposed stochastic Taylor expansion for control systems is essentially the same as the Fliess expansion that is well-known in control theory, with the main difference adapted from stochastic calculus, which allows, in particular, a transparent representation of the remainder term and a straightforward derivation of approximations of an arbitrary desired order. The paper shows how derivative-free schemes can be obtained from numerical Taylor schemes of arbitrary order. This provides an efficient device for the construction of the right kind of Runge-Kutta schemes for the affinely controlled nonlinear systems. Several simplifications to the Taylor schemes based on a special additive or commutative control structure of the system are also discussed. Furthermore, the approximation of the multiple control integrals appearing in the proposed schemes is analyzed; in particular, the approximation by averaging for a single control function, and the approximation of the set of multiple control integrals for all measurable control functions. A numerical example testing the Euler and Heun schemes for a 2-dimensional system with a single control illustrates the proposed method.