On regularity properties and approximations of value functions for stochastic differential games in domains (Q465472)

A two-person zero-sum stochastic differential game with discounted cost till exit from a domain (possibly the whole space) and a state and control dependent discount factor is considered. Results ensuring its Lipschitz or local Lipschitz properties are presented, leading to the main result which shows that there exists a constant \(N\) depending only on the problem parameters such that for any \(K > 0\), the value function is uniformly approximated within \(N/K\) by the value function of a related stochastic differential game the running cost of which is modified in a \(K\)-dependent manner. For many cases of interest, the latter value function has second derivatives uniformly bounded by a constant times \(K\).