Planning with Temporally Extended Goals Using Heuristic Search

In the publications below, we describe methods to compile a planning problem with temporally extended goals (TEGs) into a planning problem that only has final-state goals. With this translation in hand, it is possible to exploit heuristic search techniques for doing planning with TEGs. In particular, we've shown that our approach, used with the heuristic-search planner FF outperforms, often by orders of magnitude, blind-search techniques for planning with TEGs.

Conference Publications

Jorge A. Baier and Sheila McIlraith. Planning with First-Order Temporally Extended Goals Using Heuristic Search. Proceedings of the 21st National Conference on Artificial Intelligence (AAAI-06), pp. 788-795, July 2006, Boston, MA. Test suites
Jorge A. Baier and Sheila McIlraith. Planning with Temporally Extended Goals Using Heuristic Search. Proceedings of the 16th International Conference on Automated Planning and Scheduling (ICAPS-06), June 2006, Lake District, UK

Software

Here is the translator for propositional TEGs. This translator needs SWI-Prolog. The input is a planning problem in prolog syntax (more details in the documentation in the package). The output are PDDL domains/problems.
Here is the translator for first-order TEGs. This translator needs SWI-Prolog. The input is a planning problem in prolog syntax (more details in the documentation in the package). The output are PDDL domains/problems.

We expect to release PDDL-to-PDDL translator in the near future.