OPTIC

OPTIC, for Optimizing Preferences and TIme-dependent Costs, extends POPF to plan quality beyond the length of a schedule. Of the planners we compare, it is the only one that handles PDDL3 preferences.

Authors: J. Benton, Amanda Coles, Andrew Coles
Institutions: At publication: Arizona State University, King's College London
Kind of planning: Temporal and numeric, with PDDL3 preferences and constraints
Source: OPTIC on SourceForge

How it works

OPTIC extends POPF's scheduler, heuristic and search. It models preferences with PDDL3 and costs that vary with time with PDDL+ constructs, and it schedules plans with a mixed integer program so as to minimise preference cost. Relaxed plan length guides the search, and a second, admissible heuristic prunes it.

What it handles

What POPF handles, plus PDDL3 state-trajectory constraints and preferences. Autoplan does not yet handle preferences. See the feature comparison.

Results in our benchmarks

Track	Problems	OPTIC	Autoplan
IPC 1998–2018 classical	2,742	824	2,450
IPC 2023 classical	140	0	109
IPC 2023 numeric	400	72	383
IPC 2026 numeric	320	43	270
IPC 2002 temporal	436	141	338
IPC 2004 temporal	976	361	715
IPC 2006 temporal	450	141	308
IPC 2008 temporal	390	201	348
IPC 2011 temporal	240	89	218
IPC 2014 temporal	200	26	181
IPC 2018 temporal	90	39	86

OPTIC gcc8-2019-01-08: 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

J. Benton, Amanda Coles, Andrew Coles. Temporal Planning with Preferences and Time-Dependent Continuous Costs. Proceedings of ICAPS 2012.