Improved Airport Ground Traffic Control with Domain-Dependent Heuristics

In this paper we study the application of a domain-dependent heuristic to airport ground traffic control. We consider two variants of the problem. In the first, proposed for the International Planning Competition in 2004, the in-bound and out-bound airplanes have fixed parking and take-off positions. In the second, more realistic variant a controller can assign dynamically for each airplane the runway for take-off or the parking position, such that the total movement of planes at the airport is minimized. We are particularly interested in the second variant, which has an implicitly defined goal state where multiple states could satisfy the goal condition, and the impact of this fact on domain-independent and domain-dependent heuristics. We compare domain-independent heuristics in the Fast Downward planner on this domain to a domain-dependent heuristic.