Linear programming has been successfully used to compute admissible heuristics for cost-optimal classical planning. Although one of the strengths of linear programming is the ability to express and reason about numeric variables and constraints, their use in numeric planning is limited. In this work, we extend linear programming-based heuristics for classical planning to support numeric state variables. In particular, we propose a model for the interval relaxation, coupled with landmarks and state equation constraints. We consider both linear programming models and their harder-to-solve, yet more informative, integer programming versions. Our experimental analysis shows that considering an NP-Hard heuristic often pays off and that A* search using our integer programming heuristics establishes a new state of the art in cost-optimal numeric planning.

In particular, modeling context dependent eects, concurrent execution of actions with dierent duration, and continuous resources are all awkward, or impossible, within the STRIPS language. To overcome the rst of these limitations, Pednault (1989) dened the (nowadays widely accepted) ADL language, which amongst other things allows for conditional eects (eects that only occur when their condition holds true in the state of execution). To overcome (one or both of) the latter two limitations, various proposals have been made (e.g., Ghallab & Laruelle, 1994; Koehler, 1998; Smith & Weld, 1999). The most recent eort in this direction is the PDDL2.1 language dened by Fox and Long (2002) as the input language for the 3rd International Planning Competition (IPC-3). The IPC series is a biennial challenge for the planning community, inviting planning systems to participate in a large scale publicly accessible evaluation. IPC-3 was hosted at AIPS-2002, and stressed planning beyond the STRIPS formalism, featuring tracks for temporal and numeric planners. This article describes the approach behind one of the planners that participated in IPC-3, Metric-FF. Metric-FF is an extension of the FF system (that can handle ADL) to numeric constructs.

Planning with numeric state variables has been a challenge for many years, and was a part of the 3rd International Planning Competition (IPC-3). Currently one of the most popular and successful algorithmic techniques in STRIPS planning is to guide search by a heuristic function, where the heuristic is based on relaxing the planning task by ignoring the delete lists of the available actions. We present a natural extension of ``ignoring delete lists'' to numeric state variables, preserving the relevant theoretical properties of the STRIPS relaxation under the condition that the numeric task at hand is ``monotonic''. We then identify a subset of the numeric IPC-3 competition language, ``linear tasks'', where monotonicity can be achieved by pre-processing. Based on that, we extend the algorithms used in the heuristic planning system FF to linear tasks. The resulting system Metric-FF is, according to the IPC-3 results which we discuss, one of the two currently most efficient numeric planners.