Second order Runge-Kutta methods for Stratonovich stochastic differential equations (Q2458222)

A new class of stochastic Runge-Kutta (SRK) methods is developed for weakly approximating the solution of the system of \(d\)-dimensional Stratonovich stochastic differential equations whose integral equation form is \[ X_t=X_{t_0}+ \int^t_{t_0}a(s,X_s)\,ds+ \sum^m_{j=1} \int^t_{t_0}b^j(s,X_s)\circ dW_s^j \] where \(W\) is an \(m\)-dimensional Wiener process. It is proved that the SRK methods have order 2 if certain conditions (found using colored rooted tree analysis) are satisfied. Numerical results for three examples are presented to illustrate the small error incurred in approximating first and second order moments of the solution. A major advantage of these new SRK methods over previously known SRK methods and Euler-Maruyama methods is found to be the much smaller number of functional evaluations required. This makes the new methods feasible in practice for systems with higher values of \(d\) and \(m\).