A new adaptive Runge–Kutta method for stochastic differential equations

In this paper, we will present a new adaptive time stepping algorithm for strong approximation of stochastic ordinary differential equations. We will employ two different error estimation criteria for drift and diffusion terms of the equation, both of them based on forward and backward moves along the same time step. We will use step size selection mechanisms suitable for each of the two main regimes in the solution behavior, which correspond to domination of the drift-based local error estimator or diffusion-based one. Numerical experiments will show the effectiveness of this approach in the pathwise approximation of several standard test problems.