A Monte Carlo method for the simulation of first passage times of diffusion processes (Q1610840)

The authors consider a diffusion process satisfying a one-dimensional time-homogeneous stochastic differential equation of Itô-type. They are interested in determining via a numerical approximation the first passage time, that is the first time the diffusion process reaches an absorbing boundary. The proposed numerical method is a nested Monte Carlo algorithm in the following sense. The outer ``loop'' consists of simulating (with a strong scheme such as the Milstein scheme) \(N\) trajectories of the process on a grid on the interval of interest. If a trajectory has passed the boundary on one of the subintervals of the grid, its first passage time is estimated to be the middle of the subinterval, and a new trajectory is started. The inner ``loop'' on each subinterval is required to account for undetected crossings that may occur inside it. A suitable tied-down process is introduced and the probability of it crossing the boundary is determined using a Monte Carlo method on the tied-down process. An analysis of the algorithm is carried out and the advantages of the new method compared with previously proposed methods are discussed. Finally, numerical examples are presented.