Title: Extending Search for Hybrid Systems: Action Timing

Abstract:
Artificial Intelligence tree search algorithms are designed for
discrete systems.  In order to extend tree search algorithms for
hybrid systems, there are two new decisions the search algorithm must
make: choice of actions from a continuum, and choice of action
timing.  In this paper, we focus on principled ways to choose when to
branch the search tree.

We introduce anytime, depth-first, iterative-refinement algorithms for
satisficing and real-time decision-theoretic goals.  Further, we
discuss iterative adjustment techniques.  Finally, we briefly discuss
the use of information-based optimization for action timing choices.