UPMurphi

UPMurphi, the Universal Planner Murphi, solves PDDL+ problems by explicit model checking, including problems with non-linear dynamics. It produces universal plans, which say what to do from many states, not only from the initial one.

Authors
Giuseppe Della Penna, Daniele Magazzeni, Fabio Mercorio, Benedetto Intrigila
Institutions
At publication: University of L'Aquila, University of Roma "Tor Vergata"
Kind of planning
PDDL+, linear and non-linear
Source
UPMurphi homepage, github.com/gdellapenna/UPMurphi (archived; no longer maintained)

How it works

UPMurphi is built on the Murphi model checker. It discretises the continuous parts of the state, builds a transition graph and generates a universal plan from it, then checks the result with the VAL validator. Its authors call this Discretise and Validate. It searches without heuristics, and the only metric it takes is minimising total time.

What it handles

Numeric fluents, some forms of durative action, continuous effects, PDDL+ processes and events, and non-linear continuous change, which Autoplan does not yet handle. Domains must be typed. See the feature comparison.

Results in our benchmarks

TrackProblemsUPMurphiAutoplan
IPC 2023 numeric40010383
IPC 2026 numeric32011270
IPC 2002 temporal4360338
IPC 2004 temporal9760715
IPC 2006 temporal4500308
IPC 2008 temporal3900348
IPC 2011 temporal2400218
IPC 2014 temporal2000181
IPC 2018 temporal90086

UPMurphi abdd3fc: 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.

UPMurphi searches without heuristics and was built for hybrid systems, not for speed on IPC benchmarks. Low numbers here say little about the problems it was made for.

Paper

  1. Giuseppe Della Penna, Daniele Magazzeni, Fabio Mercorio, Benedetto Intrigila. UPMurphi: A Tool for Universal Planning on PDDL+ Problems. ICAPS 2009.