Local Search Techniques for Temporal Planning in LPG

We present some techniques for planning in temporal domains specified with the recent standard languange PDDL2.1. These techniques are implemented in LPG, a fully-automated system that took part in the third International Planning Competition (Toulouse, 2002) showing excellent performance. The planner is based on a stochastic local search method and on a graph-based representation called “Temporal Action Graphs” (TA-graphs). In this paper we present some new heuristics to guide the search in LPG using this representation. An experimental analysis of the performance of LPG on a large set of test problems used in the competition shows that our techniques can be very effective, and that often our planner outperforms all other fully-automated temporal planners that took part in the contest.