QMC methods for the solution of delay differential equations. (Q1811587)

The quasi Monte Carlo (QMC) methods for Runge-Kutta solution techniques of differential equations, which were developed by \textit{G. Stengle} [Appl. Math. Lett. 3, 25--29 (1990; Zbl 0725.65071); Numer. Math. 70, 119--128 (1995; Zbl 0817.65058)], \textit{C. Lécot} [Math. Comput. Simul. 55, 113--121 (2001; Zbl 0983.65084)], \textit{I. Coulibaly} and \textit{C. Lécot} [Math. Comput. 68, 651--659 (1999; Zbl 1043.65083)] and \textit{C. Lécot} and \textit{A. Koudiraty} [Numerical analysis of Runge-Kutta quasi-Monte Carlo methods, Math. Comput. Simul. (to appear)], are extended to delay differential equations of the form \(y^{\prime}(t)= f\big(t,y(t),y(t-\tau(t))\big)\). The retarded argument is approximated by interpolation, after which the conventional quasi Monte Carlo Runge-Kutta methods can be applied. A proof of the convergence of this method is given in a general form, that does not depend on specific quasi Monte Carlo Runge Kutta method. Finally, a numerical investigation is presented that shows, similarly to ordinary differential equations, that this quasi randomized method leads to an improvement for heavily oscillating delay differential equations compared even to high-order Runge-Kutta schemes.