LPRPG-P is a planner for use with problems containing propositional and numeric preferences. It uses a combination of a Linear Programme (LP) heuristic to give search guidance in the presence of numeric preferences, with a Relaxed Planning Graph heuristic for propositional preferences. Click here to download the sources of the IPC version of LPRPG-P.

To run LPRPG-P you will need a CPLEX licence - this prevents us from distributing a binary of the planner. An academic licence can be obtained free-of-charge from IBM.

To learn more about LPRPG-P read the publications on this page.

Domains

This page contains a number of domains with numeric preferences. Some are modified versions of standard IPC benchmarks, whilst others are new encodings of other problems. We thank the IPC 5 organising committee for permission to release modified versions of their domains.

Pathways Complex Numeric Preferences

Domain and Problem Files

This is a slightly modified variant of the pathways complex preferences domain from IPC 5, originally written by Yannis Dimopoulos, Alfonso Gerevini and Alessandro Saetti. In order to allow non-temporal planners to reason with the numeric preferences in this domain this is a non-temporal representation, compressed as per the TGP semantics, with all temporal preferences removed. We also corrected a small bug in the original encoding that required different amounts of the same substance for self-association.

TPP Complex Numeric Preferences

PDDL Domain Description and Problem Files
This domain again comes from IPC 5 and has been modified to make produce a version with temporal preferences removed thus only considering propositional and numeric preferences.

Transformer Numeric Preferences

Domain and Problem Files

This is a preferences variant of the Transformer domain, developed as part of the VOLTS project. There preferences are inherently numeric: whilst the hard constraints (embedded in the problem structure) require that the voltage is kept within a certain range (0.95 to 1.05) the soft constraints indicate the preference that in order to remain single-fault tolerant (without having to switch in emergency reactive systems) we prefer the voltage to remain in the range 0.97 to 1.03. As this is a discrete temporal model, and we wish to penalise the planner for going outside the range multiple times, we use a separate pair of preferences for each time point, specifying the voltage be above/below the min/max threshold.

Publications

""LPRPG-P: Relaxed Plan Heuristics for Planning with Preferences." " A. J. Coles and A. I. Coles. Proceedings of ICAPS 2011. July 2011. Download PDF (BibTeX)