Numerical methods for forward-backward stochastic differential equations (Q2564697)

This paper presents numerical methods for approximating the solutions of forward-backward stochastic differential equations of the form \[ X_t=x+\int^t_0 b(s,X_s,Y_s,Z_s)ds+ \int^t_0\sigma(s,X_s,Y_s)dW_s, \] \[ Y_t=g(X_T)+ \int^T_t\widehat b(s,X_s,Y_s,Z_s)ds+ \int^T_t\widehat\sigma(s,X_s,Y_s,Z_s)dW_s, \] where \(t\in[0,T]\), \(\{W_t\}\) is a \(d\)-dimensional Brownian motion, \((X,Y,Z)\) takes values in \(\mathbb{R}^m\times\mathbb{R}^m\times\mathbb{R}^{m\times d}\), and \(b\), \(\widehat b\), \(\sigma\), \(\widehat\sigma\), and \(g\) are smooth. The bulk of the paper is devoted to introducing approximations needed to devise the numerical methods, proving that the resulting approximate solutions converge to the actual solution, and establishing the rate of this convergence.