Stopping Criteria for Value and Strategy Iteration on Concurrent Stochastic Reachability Games

09/18/2019 ∙ by Julia Eisentraut, et al. ∙ 0

We consider concurrent stochastic games played on graphs with reachability and safety objectives. These games can be solved by value iteration as well as strategy iteration, each of them yielding a sequence of under-approximations of the reachability value and a sequence of over-approximation of the safety value, converging to it in the limit. For both approaches, we provide the first (anytime) algorithms with stopping criteria. The stopping criterion for value iteration is based on providing a convergent sequence of over-approximations, which then allows to estimate the distance to the true value. For strategy iteration, we bound the error by complementing the strategy iteration algorithm for reachability by a new strategy iteration algorithm under-approximating the safety-value.



There are no comments yet.


page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.