Metric-FF

Metric-FF extends the FF planner to numeric state variables. It carries FF's central idea, ignoring delete lists, over to numbers, and it was among the best numeric planners at IPC 2002.

Author

Jörg Hoffmann

Institution

At publication: University of Freiburg

Kind of planning

Classical and numeric

Source

Metric-FF homepage (GPL)

How it works

Metric-FF is a forward search planner. Its heuristic estimates the distance to the goal as the length of a relaxed plan, one found while ignoring what actions undo. The paper shows how to extend that relaxation to numeric variables for "monotonic" tasks, and how linear tasks can be brought into that form.

The search runs enforced hill-climbing first and falls back to a complete best-first search if that fails. Version 2.0 adds A*-epsilon for minimising additive costs, and version 2.1 an optional bound on plan cost.

What it handles

Classical PDDL with ADL features, action costs, numeric conditions and effects, and plan metrics. It does not handle derived predicates or durative actions. See the feature comparison.

In the competitions

Metric-FF was a top performer in the numeric track of IPC 2002. Its paper reports that it and LPG "were the best performing numeric planners in the competition". The organisers' report describes its numeric performance as exceptional.

Results in our benchmarks

Metric-FF 2.1: problems solved with a plan the VAL validator accepts, within 60 seconds and 4 GB of memory on one core. These are our own runs, not competition results, and a planner written for other goals or for a different kind of problem can score low here while being the right tool for its purpose. How we measured.

Paper

1. Jörg Hoffmann. The Metric-FF Planning System: Translating "Ignoring Delete Lists" to Numeric State Variables. Journal of Artificial Intelligence Research 20, 291-341, 2003.