Pathwise approximation of random ordinary differential equations. (English) Zbl 0998.65010
This paper presents an averaging procedure for improving the Euler and Heun methods of numerical approximation of the solution of the vector random ordinary differential equation
\[
\frac{dx}{dt}=G(t,\omega) +g(t,\omega) H(x,\omega),
\]
where \(G\), \(g\) are Hölder continuous in \(t\), and \(H\) is once (for Euler) or twice (for Heun) continuously differentiable in \(x\). It is shown that by averaging \(G\) and \(g\) over \(N\) equally spaced values of \(t\) in each discretization subinterval, order of convergence 1 for the Euler method and 2 for the Heun method can be attained, whereas the standard Euler and Heun methods may have only fractional order. Two examples are given which demonstrate that the averaging procedure approximations can show a substantial reduction in error over the approximations generated by the standard Euler and Heun methods.
Reviewer: Melvin D.Lax (Long Beach)
MSC:
65C30 | Numerical solutions to stochastic differential and integral equations |
65L06 | Multistep, Runge-Kutta and extrapolation methods for ordinary differential equations |
34F05 | Ordinary differential equations and systems with randomness |
65L20 | Stability and convergence of numerical methods for ordinary differential equations |
60H10 | Stochastic ordinary differential equations (aspects of stochastic analysis) |
60H35 | Computational methods for stochastic equations (aspects of stochastic analysis) |
65L70 | Error bounds for numerical methods for ordinary differential equations |