We introduce a new discounting semantics for priced timed automata. Discounting provides a way to model optimal-cost problems for infinite traces and has applications in optimal scheduling and other areas.
In the discounting semantics, prices decrease exponentially, so that the contribution of a certain part of the behaviour to the overall cost depends on how far into the future this part takes place. We consider the optimal infinite run problem under this semantics: Given a priced timed automaton, find an infinite path with minimal discounted price. We show that this problem is computable, by a reduction to a similar problem on finite weighted graphs.
The proof relies on a new theorem on minimization of monotonous functions defined on infinite-dimensional zones, which is of interest in itself.
