A random walk on rectangles algorithm (Q2433258)

The authors introduce a new algorithm that is designed to simulate the first exit time and first exit position from a rectangle (or a parallelepiped or polygonal domain) for a Brownian motion that starts at any point inside. This method provides an approximative solution to some nonrandom Dirichlet problems for linear second order PDEs in any dimension. It represents an analogous method to the method of the random walk on spheres (WOS) and can be adapted to treat Neumann boundary conditions or Brownian motion with a constant drift. Some numerical tests and discussions are presented as well.