IPC 2023 Classical Tracks
This is the website for the classical (sequential, deterministic) track of the IPC 2023. This is the 10th IPC containing classical tracks making it the oldest part of IPC.
The results were presented at the 33rd International Conference on Automated Planning and Scheduling on July 11 in Prague. The presentation slides of this talk contain additional details.
folding | labyrinth | quantum. | recharg. | ricochet. | rubiks | slither. | SUM | |
---|---|---|---|---|---|---|---|---|
ragnarok | 8 | 8 | 13 | 14 | 17 | 10 | 7 | 77 |
scorpion-2023 | 8 | 5 | 14 | 14 | 17 | 10 | 6 | 74 |
odin | 8 | 5 | 13 | 14 | 17 | 10 | 6 | 73 |
dofri | 8 | 5 | 13 | 13 | 17 | 10 | 4 | 70 |
cegarplusplus | 9 | 5 | 13 | 14 | 17 | 0 | 7 | 65 |
hapori-stonesoup-opt | 7 | 2 | 13 | 14 | 9 | 11 | 6 | 62 |
fdss-2023-opt | 7 | 3 | 13 | 13 | 12 | 9 | 4 | 61 |
hapori-mip2-opt | 7 | 1 | 13 | 14 | 9 | 10 | 6 | 60 |
hapori-ibacop2-opt | 6 | 1 | 13 | 12 | 15 | 7 | 4 | 58 |
hapori-greedy-opt | 5 | 1 | 13 | 11 | 9 | 10 | 7 | 56 |
baseline-blind | 7 | 1 | 7 | 12 | 11 | 8 | 4 | 50 |
decstar-opt | 6 | 1 | 12 | 11 | 8 | 8 | 4 | 50 |
hapori-delfi-opt | 5 | 2 | 12 | 12 | 8 | 0 | 2 | 41 |
complementary | 5 | 1 | 12 | 13 | 3 | 0 | 3 | 37 |
decabstar | 2 | 1 | 12 | 10 | 7 | 0 | 5 | 37 |
symk | 3 | 1 | 9 | 13 | 4 | 0 | 7 | 37 |
fts-ms-opt | 1 | 1 | 12 | 13 | 2 | 0 | 7 | 36 |
baseline-lmcut | 2 | 1 | 12 | 8 | 5 | 0 | 6 | 34 |
hapori-epslr-opt | 2 | 1 | 9 | 6 | 4 | 10 | 2 | 34 |
SymBD-2023-opt | 2 | 1 | 9 | 13 | 1 | 0 | 6 | 32 |
dom-opt | 2 | 1 | 12 | 6 | 4 | 0 | 6 | 31 |
hapori-epsdt-opt | 1 | 0 | 9 | 6 | 4 | 7 | 4 | 31 |
dalai-opt | 2 | 1 | 11 | 7 | 4 | 0 | 4 | 29 |
fts-sbd-opt | 1 | 0 | 4 | 13 | 0 | 0 | 4 | 22 |
folding | labyrinth | quantum. | recharg. | ricochet. | rubiks | slither. | SUM | |
---|---|---|---|---|---|---|---|---|
maidu-sat | 6.80 | 0.00 | 19.60 | 13.94 | 11.36 | 14.16 | 6.00 | 71.86 |
levitron-sat | 8.66 | 0.00 | 19.60 | 13.94 | 11.44 | 14.16 | 4.00 | 71.79 |
fdss-2023-sat | 8.95 | 0.00 | 19.49 | 13.80 | 8.51 | 14.13 | 6.00 | 70.88 |
baseline-lama | 9.70 | 1.00 | 17.86 | 13.19 | 9.83 | 12.18 | 5.00 | 68.76 |
hapori-ibacop2-sat | 8.69 | 3.91 | 17.07 | 13.46 | 7.37 | 11.22 | 6.00 | 67.72 |
disco-sat | 8.58 | 0.00 | 17.78 | 13.95 | 9.14 | 11.77 | 5.00 | 66.24 |
cerberus-sat | 6.39 | 0.00 | 18.03 | 10.43 | 12.73 | 10.93 | 6.00 | 64.52 |
spock | 5.83 | 0.00 | 19.67 | 13.94 | 5.01 | 13.90 | 6.00 | 64.36 |
decstar-sat | 5.40 | 0.00 | 17.33 | 13.29 | 8.41 | 12.42 | 6.00 | 62.85 |
tftm-argmax-sat | 6.73 | 0.00 | 16.74 | 9.55 | 12.73 | 9.98 | 6.00 | 61.73 |
tftm-co1-sat | 6.67 | 0.00 | 15.84 | 10.43 | 12.65 | 10.10 | 6.00 | 61.68 |
hapori-mip2-sat | 5.64 | 3.91 | 18.86 | 13.28 | 5.04 | 10.00 | 4.00 | 60.74 |
dalai-sat | 4.84 | 4.00 | 17.29 | 13.00 | 8.82 | 5.00 | 3.00 | 55.95 |
ApxNoveltyAnytime | 5.00 | 0.00 | 18.31 | 8.00 | 15.32 | 5.00 | 4.00 | 55.63 |
hapori-stonesoup-sat | 7.69 | 3.91 | 18.44 | 12.32 | 4.17 | 4.00 | 4.00 | 54.52 |
NoveltyFBAnytime | 1.00 | 0.00 | 18.31 | 8.00 | 13.80 | 5.00 | 4.00 | 50.12 |
opcount4sat-sat | 2.00 | 3.76 | 13.32 | 10.11 | 6.15 | 0.00 | 4.00 | 39.34 |
hapori-epsdt-sat | 7.40 | 14.14 | 0.00 | 5.38 | 0.62 | 6.03 | 0.00 | 33.57 |
hapori-delfi-sat | 3.81 | 0.00 | 12.52 | 8.85 | 2.35 | 4.00 | 1.00 | 32.54 |
FSM | 3.03 | 0.00 | 6.60 | 8.22 | 3.00 | 6.00 | 3.00 | 29.84 |
powerlifted-sat | 9.69 | 0.00 | 16.71 | 0.00 | 0.00 | 0.00 | 2.00 | 28.40 |
hapori-epslr-sat | 1.00 | 0.00 | 0.00 | 3.45 | 0.00 | 11.00 | 5.00 | 20.45 |
hapori-greedy-sat | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
folding | labyrinth | quantum. | recharg. | ricochet. | rubiks | slither. | SUM | |
---|---|---|---|---|---|---|---|---|
baseline-lama-first | 3.35 | 0.00 | 16.86 | 3.79 | 0.94 | 13.16 | 2.19 | 40.28 |
decstar-agl | 2.62 | 0.00 | 15.26 | 5.01 | 2.72 | 12.95 | 1.70 | 40.25 |
fdss-2023-agl | 2.96 | 1.07 | 15.31 | 4.47 | 0.56 | 11.07 | 2.39 | 37.82 |
ApxNoveltyTarski | 1.91 | 2.66 | 19.53 | 3.92 | 3.41 | 3.13 | 1.80 | 36.35 |
disco-agl | 3.39 | 0.00 | 16.82 | 6.47 | 1.08 | 6.30 | 1.91 | 35.96 |
maidu-agl | 2.52 | 0.28 | 15.57 | 4.41 | 1.57 | 9.64 | 1.93 | 35.93 |
levitron-agl | 2.49 | 0.34 | 15.64 | 4.11 | 1.00 | 9.48 | 2.55 | 35.61 |
ApxNovelty | 1.43 | 0.00 | 19.64 | 3.03 | 3.41 | 4.00 | 1.81 | 33.32 |
cerberus-agl | 2.74 | 0.00 | 13.47 | 2.38 | 1.83 | 7.79 | 2.64 | 30.84 |
dalai-agl | 3.05 | 0.34 | 13.64 | 6.57 | 3.54 | 2.67 | 1.01 | 30.81 |
hapori-stonesoup-agl | 0.87 | 0.74 | 16.70 | 3.89 | 0.81 | 3.86 | 1.62 | 28.49 |
hapori-mip2-agl | 0.37 | 0.77 | 13.60 | 2.47 | 0.82 | 7.12 | 2.67 | 27.82 |
tftm-argmax-agl | 2.73 | 0.00 | 12.15 | 2.45 | 1.74 | 6.01 | 2.64 | 27.73 |
tftm-co1-agl | 2.56 | 0.00 | 12.26 | 2.37 | 1.89 | 6.01 | 2.61 | 27.70 |
NoveltyFB | 0.14 | 0.00 | 18.88 | 0.83 | 1.18 | 4.00 | 0.48 | 25.50 |
fts-ff-agl | 0.00 | 0.00 | 9.61 | 5.19 | 0.55 | 0.00 | 1.97 | 17.31 |
hapori-epsdt-agl | 1.84 | 2.65 | 0.00 | 2.82 | 0.51 | 5.20 | 0.00 | 13.03 |
hapori-epslr-agl | 0.01 | 0.00 | 0.00 | 2.48 | 0.00 | 7.46 | 0.98 | 10.94 |
hapori-delfi-agl | 0.68 | 0.00 | 5.06 | 2.40 | 1.53 | 0.64 | 0.52 | 10.83 |
opcount4sat-agl | 0.00 | 0.04 | 3.04 | 6.28 | 0.00 | 0.00 | 0.74 | 10.10 |
FSM | 0.00 | 0.00 | 0.82 | 0.95 | 0.24 | 3.17 | 0.09 | 5.26 |
powerlifted-agl | 0.73 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.07 | 0.80 |
hapori-greedy-agl | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
Repository with all competition tasks can be found here.
List of competition domains:
All participating planners can be used as Apptainer images. To build such an image, install Apptainer and clone the corresponding planner repository. You can find links to them in the list of participants. Code repositories contain two branches: ipc2023-classical
contains exactly the code version that ran in the IPC, while latest
can contain additional bug fixes published after the competition. We recommend using latest
for all experiments. Some code repositories contain multiple Apptainer files for different planners that share a code base. Use Apptainer to build the planner from one of those recipes like this:
git clone https://github.com/ipc2023-classical/planner8.git
cd planner8
sudo apptainer build maidu_sat.sif Apptainer.maidu_sat
Some planners require CPLEX to run. To use them, you have to acquire a CPLEX license (IBM offers free academic licenses) and download the installer file cplex_studio2211.linux_x86_64.bin
. Place it into a directory and use the environment variable IPC_THIRD_PARTY
to identify it.
export IPC_THIRD_PARTY=/path/to/some/directory
cp cplex_studio2211.linux_x86_64.bin $IPC_THIRD_PARTY
git clone https://github.com/ipc2023-classical/planner17.git
cd planner17
sudo apptainer build ragnarok.sif Apptainer.ragnarok
All planners are also available through planutils.
Please forward the following calls to all interested parties.
Event | Date |
---|---|
Call for domains | August 1, 2022 |
Call for participation | October, 2022 |
Domain expression of interest deadline | September 30, 2022 |
Domain submission deadline | December 9, 2022 |
Demo problems provided | December, 2022 |
Planner registration | January 13, 2023 |
Feature stop (final planner submission) | March 24, 2023 |
Planner Abstract submission deadline | April 21, 2023 |
Contest run | May - June, 2023 |
Results announced | July, 2023 |
Result analysis deadline | October 31, 2023 |
IPC 2023 will use a subset of PDDL 3.1, as done since IPC 2011. Planners must support the subset of the language involving STRIPS, action costs, negative preconditions, and conditional effects (possibly in combination with forall, as in IPC 2014 and 2018). We will also consider including domains with disjunctive preconditions and existential quantifiers, in which case we provide an automatic translation compiling these features away, and we run all planners on both variants and select the best result per domain.
Most planners in previous IPCs rely on a grounding procedure to instantiate the entire planning task prior to start solving it. In IPC 2023, we will follow in the steps of the previous IPC by including domains and problems that are hard to ground.
SymBD (planner abstract) (code)
Alvaro Torralba
Symbolic Bidirectonal Blind Search
Hapori MIPlan Optimal All Data (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of optimal IPC planners computed with the MIP formulation by Nunez, Borrajo and Linares (2015).
Ragnarok (planner abstract) (code)
Dominik Drexler, Daniel Gnad, Paul Höft, Jendrik Seipp, David Speck, Simon Ståhlberg
Sequential portfolio of optimal planners developed at Linköping University
Hapori Stone Soup Optimal (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of optimal IPC planners computed with the Stone Soup algorithm
FTSPlan (planner abstract) (code)
Álvaro Torralba, Silvan Sievers, Rasmus G. Tollund, Kristian Ø. Nielsen
Optimal planner based on reformulation and search under the merge-and-shrink representation search. After reformulation uses symbolic bidirectional search or A* with the merge-and-shink heuristic, depending on the configuration.
CEGAR++ (planner abstract) (code)
Raphael Kreft, Silvan Sievers
Pattern Databases for interesting patterns up to size 2, patterns computed with hill climbing and CEGAR, combined with domain abstractions and Cartesian abstractions computed with CEGAR in a maximum heuristic over SCP heuristics, generated through greedy computation of hybrid-optimized orders.
QDom-Lmcut (planner abstract) (code)
Alvaro Torralba
Dominance Pruning
Hapori Explainable Planner Selection - Single Decision Tree Optimal (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Explainable Planner Selection for Classical Planning (AAAI 2022) using Fawcetts et al (ICAPS 2014) PDDL features and linear regression.
Hapori Delfi Optimal (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Deep Learning for Cost-Optimal Planning: Task-Dependent Planner Selection (Sievers AAAI 2019)
Symk (planner abstract) (code)
David Speck
Symbolic bidirectional blind search
Hapori Greedy Optimal (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of optimal IPC planners computed with the greedy offline approximation algorithm by Streeter and Smith (UAI 2008).
DecAbStar (planner abstract) (code)
Daniel Gnad, Silvan Sievers, Alvaro Torralba
Optimal planner based on decoupled state-space search and abstraction heuristics.
Fast Downward Stone Soup 2023 Optimal (planner abstract) (code)
Clemens Büchner, Remo Christen, Augusto Blaas Corrêa, Salomé Eriksson, Patrick Ferber, Jendrik Seipp, Silvan Sievers
Portfolio of configurations from the main Fast Downward branch, scheduled with the Stone Soup algorithm.
DALAI 2023 Optimal (planner abstract) (code)
Clemens Büchner, Remo Christen, Salomé Eriksson, Thomas Keller
Disjunctive Action Landmarks All In -- Path-dependent landmark heuristic search tailored to find optimal solutions.
DecStar-2023 (planner abstract) (code)
Daniel Gnad, Alvaro Torralba, Alexander Shleyfman
Optimal planner based on decoupled state-space search.
Odin (planner abstract) (code)
Dominik Drexler, Jendrik Seipp, David Speck
Transition cost partitioning over abstraction heuristics
Hapori Explainable Planner Selection - Linear Regression Optimal (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Explainable Planner Selection for Classical Planning (AAAI 2022) using Fawcetts et al (ICAPS 2014) PDDL features and linear regression.
Dofri (planner abstract) (code)
Paul Höft, David Speck, Jendrik Seipp
forward search with A* and the Saturated Post-Hoc optimization heuristic that selectively reuses previous SPhO LP solutions.
Hapori IBaCop2 Sat (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of optimal IPC planners computed on an instance-based fashion. A set of 3 planners are pre-selected based on PDDL task and heuristic features.
Scorpion 2023 (planner abstract) (code)
Jendrik Seipp
Saturated cost partitioning over abstraction heuristics.
ComplementaryPDB (planner abstract) (code)
Santiago Franco, Stefan Edelkamp, Ionut Moraru
Modified version of complementary heuristic, where we are using completely new bin packing algorithms(paper pending), in situ learning of all the algorithm parameters critical to the pattern selection performance (previously only which pattern generation algorithm we use). Also we have added a new pattern generator inspired on how Gamer chooses a single PDB which it keeps improving. Also the selection algorithm is based on size of search space (previously selection criteria was time). Some features from previous complementary heuristic as in the iJCAI 18 paper are yet to be adapted to this version, e.g. mutation for local search of succesful selection, stratified sampling.
Hapori Explainable Planner Selection - Single Decision Tree Satisficing (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Explainable Planner Selection for Classical Planning (AAAI 2022) using Fawcetts et al (ICAPS 2014) PDDL features and linear regression.
TFTM2 (planner abstract) (code)
Alex Tuisov, Michael Katz
Planner based on the IJCAI 2021 paper The Fewer the Merrier: Pruning Preferred Operators with Novelty.
Spock (planner abstract) (code)
Jendrik Seipp, Mauricio Salerno, Raquel Fuentetaja
FDSS 2018 with action elimination. Removes redundant actions from each found plan to -potentially- rapidly find a cheaper plan.
Powerlifted Satisficing (planner abstract) (code)
Augusto B. Corrêa, Guillem Francès, Markus Hecher, Davide Mario Longo, Jendrik Seipp
Powerlifted planner using a sequential portfolio configuration
DiSCO (planner abstract) (code)
Maximilian Fickert, Daniel Gnad
Satisficing planner based on Decoupled Search and COnjunctions.
OpCount4Sat - Satistificng (planner abstract) (code)
Daniel Matheus Doebber, André Grahl Pereira, Augusto B. Corrêa
Operator Counting Heuristics for Satisficing Planning
Hapori Delfi Satisficing (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Deep Learning for Cost-Optimal Planning: Task-Dependent Planner Selection (Sievers AAAI 2019)
Hapori MIPlan All Data Sat (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of optimal and satisficing IPC planners computed with the MIP formulation by Nunez, Borrajo and Linares (2015).
Hapori Greedy Satisficing (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of satisficing IPC planners computed with the greedy offline approximation algorithm by Streeter and Smith (UAI 2008).
DALAI 2023 Satisficing (planner abstract) (code)
Clemens Büchner, Remo Christen, Salomé Eriksson, Thomas Keller
Disjunctive Action Landmarks All In -- Path-dependent landmark heuristic search tailored to find some solution fast and improve improve it iteratively.
Scorpion Maidu Satisficing (planner abstract) (code)
Augusto B. Corrêa, Guillem Francès, Markus Hecher, Davide Mario Longo, Jendrik Seipp
Version of the Scorpion planner using width-search algorithms
Cerberus (planner abstract) (code)
Michael Katz
Planner Cerberus from IPC 2018
Approximate Novelty Anytime (planner abstract) (code)
Anubhav Singh, Nir Lipovetzky, Miquel Ramirez, Javier Segovia-Aguas
The anytime configuration of the planner. It uses ApxNovelty search to find the first satisficing plan. After which, it switches to lazy wastar with a upper bound on the plan cost, and continues to optimize the solutions until timeout.
Hapori Stone Soup Satisficing (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of satisficing IPC planners computed with the Stone Soup algorithm
DecStar-2023 (planner abstract) (code)
Daniel Gnad, Alvaro Torralba, Alexander Shleyfman
Satisficing planner based on decoupled state-space search.
FSM (planner abstract) (code)
Gustavo Prolla Lacroix, Rafael Vales Bettker, André Grahl Pereira
A learning-based planner for short-time sampling, training, and testing.
Forward Backward Anytime Novelty Search (planner abstract) (code)
Anubhav Singh, Chao Lie, Nir Lipovetzky, Miquel Ramirez, Javier Segovia-Aguas
The anytime configuration of the planner. It uses Forward Backward Novelty Search search to find the first satisficing plan. After which, it switches to lazy wastar with a upper bound on the plan cost, and continues to optimize the solutions until timeout.
Fast Downward Stone Soup 2023 Satisficing (planner abstract) (code)
Clemens Büchner, Remo Christen, Augusto Blaas Corrêa, Salomé Eriksson, Patrick Ferber, Jendrik Seipp, Silvan Sievers
Portfolio of configurations from the main Fast Downward branch, scheduled with the Stone Soup algorithm.
TFTM1 (planner abstract) (code)
Alex Tuisov, Michael Katz
Planner based on the IJCAI 2021 paper The Fewer the Merrier: Pruning Preferred Operators with Novelty.
Hapori Explainable Planner Selection - Linear Regression Satisficing (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Explainable Planner Selection for Classical Planning (AAAI 2022) using Fawcetts et al (ICAPS 2014) PDDL features and linear regression.
Hapori IBaCop2 Sat (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of satisficing IPC planners computed on an instance-based fashion. A set of 5 planners are pre-selected based on PDDL task and heuristic features.
Levitron Satisficing (planner abstract) (code)
Augusto B. Corrêa, Guillem Francès, Markus Hecher, Davide Mario Longo, Jendrik Seipp
Hybrid planner using both the grounded planner Scorpion Maidu and the lifted planner Powerlifted
OpCount4Sat - Agile (planner abstract) (code)
Daniel Matheus Doebber, André Grahl Pereira, Augusto B. Corrêa
Operator Counting Heuristics for Satisficing Planning
Hapori Explainable Planner Selection - Single Decision Tree Agile (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Explainable Planner Selection for Classical Planning (AAAI 2022) using Fawcetts et al (ICAPS 2014) PDDL features and linear regression.
Hapori Stone Soup Agile (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of satisficing IPC planners computed with the Stone Soup algorithm
Forward Backward Novelty Search (planner abstract) (code)
Anubhav Singh, Chao Lie, Nir Lipovetzky, Miquel Ramirez, Javier Segovia-Aguas
Forward Backward Approximate BFWS with novelty approximation and goal count heuristics, f=<#w,#g>, see ICAPS 2021 papers on Approximate Novelty Search and Width-based backward search
DiSCO (planner abstract) (code)
Maximilian Fickert, Daniel Gnad
Agile planner based on Decoupled Search and COnjunctions.
Hapori Delfi Agile (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Deep Learning for Cost-Optimal Planning: Task-Dependent Planner Selection (Sievers AAAI 2019)
Levitron Agile (planner abstract) (code)
Augusto B. Corrêa, Guillem Francès, Markus Hecher, Davide Mario Longo, Jendrik Seipp
Hybrid planner using both the grounded planner Scorpion Maidu and the lifted planner Powerlifted
DALAI 2023 Agile (planner abstract) (code)
Clemens Büchner, Remo Christen, Salomé Eriksson, Thomas Keller
Disjunctive Action Landmarks All In -- Path-dependent landmark heuristic search tailored to find solutions fast.
Grounding Schematic Representation with GRINGO for Width-based Search (planner abstract) (code)
Anubhav Singh, Nir Lipovetzky, Miquel Ramirez, Javier Segovia-Aguas, Guillem Francès
Approximate Best First Width Search with novelty approximation and goal count heuristics, f=<#w,#g>, the planner leverages Tarski to ground the schematic representation of the planning problem, refer https://tarski.readthedocs.io/en/latest/notebooks/grounding-reachability-analysis.html
Hapori Explainable Planner Selection - Linear Regression Agile (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Planner selection portfolio as described in Explainable Planner Selection for Classical Planning (AAAI 2022) using Fawcetts et al (ICAPS 2014) PDDL features and linear regression.
Scorpion Maidu Agile (planner abstract) (code)
Augusto B. Corrêa, Guillem Francès, Markus Hecher, Davide Mario Longo, Jendrik Seipp
Version of the Scorpion planner using width-search algorithms
TFTM1 (planner abstract) (code)
Alex Tuisov, Michael Katz
Planner based on the IJCAI 2021 paper The Fewer the Merrier: Pruning Preferred Operators with Novelty.
Hapori Greedy Agile (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of satisficing IPC planners computed with the greedy offline approximation algorithm by Streeter and Smith (UAI 2008).
FTSPlan (planner abstract) (code)
Álvaro Torralba, Silvan Sievers, Rasmus G. Tollund, Kristian Ø. Nielsen
Optimal planner based on reformulation and search under the merge-and-shrink representation search. After reformulation uses Greedy Best First Search with the FF heuristic.
Approximate Novelty (planner abstract) (code)
Anubhav Singh, Nir Lipovetzky, Miquel Ramirez, Javier Segovia-Aguas
Approximate Best First Width Search with novelty approximation and goal count heuristics, f=<#w,#g>, leverages Fast Downward to ground the schematic representation of the planning problem
Hapori MIPlan All Data Random Order Agl (planner abstract) (code)
Patrick Ferber, Michael Katz, Jendrik Seipp, Silvan Sievers, Daniel Borrajo, Isabel Cenamor, Tomas de la Rosa, Fernando Fernandez-Rebollo, Carlos Linares, Sergio Nunez, Alberto Pozanco, Horst Samulowitz, Shirin Sohrabi
Sequential portfolio of satisficing and agile IPC planners computed with the MIP formulation by Nunez, Borrajo and Linares (2015).
DecStar-2023 (planner abstract) (code)
Daniel Gnad, Alvaro Torralba, Alexander Shleyfman
Agile planner based on decoupled state-space search.
Fast Downward Stone Soup 2023 Agile (planner abstract) (code)
Clemens Büchner, Remo Christen, Augusto Blaas Corrêa, Salomé Eriksson, Patrick Ferber, Jendrik Seipp, Silvan Sievers
Portfolio of configurations from the main Fast Downward branch, scheduled with the Stone Soup algorithm.
TFTM2 (planner abstract) (code)
Alex Tuisov, Michael Katz
Planner based on the IJCAI 2021 paper The Fewer the Merrier: Pruning Preferred Operators with Novelty.
FSM (planner abstract) (code)
Gustavo Prolla Lacroix, Rafael Vales Bettker, André Grahl Pereira
A learning-based planner for short-time sampling, training, and testing.
Cerberus (planner abstract) (code)
Michael Katz
Planner Cerberus from IPC 2018
Powerlifted Agile (planner abstract) (code)
Augusto B. Corrêa, Guillem Francès, Markus Hecher, Davide Mario Longo, Jendrik Seipp
Powerlifted planner using a sequential portfolio configuration
To register a team, the participants need to send an e-mail with a subject containing "[Registration]" to ipc2023-classical@googlegroups.com. The e-mail must contain:
Based on that, we will create private repositories under the ipc2023-classical organization and add all participants as users with with write access and participants can commit to the repository as they wish until the "feature stop" deadline (March 10, 2023).
As in previous editions, the competitors must submit the source code of their planners that will be run by the organizers on the actual competition domains/problems, unknown to the competitors until this time. This way no fine-tuning of the planners will be possible.
As in the previous IPC 2018, we will use the container technology Apptainer (formerly known as Singularity) to promote reproducibility and help with compilation issues that have caused problems in the past. In contrast to the previous IPC, we will host repositories of planners ourselves. The repositories will be hosted on Github under the ipc2023-classical organization, and they will be kept private until the end of the competition when we make them public, i.e., after the competition is concluded, we plan to make all planners, domains, and all related data accessible from one place.
This year, we will also allow to submit multiple planners to multiple tracks
from a single repository. In each repository, we only consider the branch
ipc2023-classical
. Feel free to use other branches for development as you wish, but we will
ignore them. Any file called Apptainer.<shortname>
in the root directory of
this branch defines one entry. For the <shortname>
, please use the name and
variant of your planner as a short identifier (a single word, up to 16
characters long, starting with a letter, using only letters, digits, and
underscores). If you build different versions of your planner from the same
repository, use a different <shortname>
per version. A single entry can
participate in multiple tracks, see "Apptainer Images" for details.
We got many registrations and in particular many registrations with multiple submissions in a team. To keep the computation time manageable, we ask that you limit your submissions to one variant per planner. If two of your submissions are substantially different ideas, it is of course fine to submit both. As a guideline, if two submissions would be described with two planner abstracts, then they are two submissions. If you would describe them in one planner abstract with a short paragraph describing the differences between them, they are variants of the same planner.
After some discussions, we decided to allow multiple variants of the same planner if you think they are likely end up on very different spots in scoring (i.e., not next to each other). In that case, you should still write two planner abstracts, but one may discuss the difference to the other submission. In that case, we ask you to write a paragraph into the second planner abstract that discusses your reasons of why you think this should be a separate submission, i.e., why you think these two submissions will not end up next to each other in the scoring. In the post-IPC analysis, we then ask you to evaluate this reasoning: if the planners indeed ended up in very different spots, what made the difference? If they ended up next to each other, why did this happen? Moreover, if the leaderboard ends up with multiple variants of the same planner in the first places, we reserve the right to merge these entries. New variants can be submitted until April 16 via pull requests.
We prepared a demo submission that showcases how to set up the repository and Apptainer scripts.
Your Apptainer recipe files have to specify the following labels:
Name
: name of the planner
Description
: a short description of your planner
Authors
: a list of authors, including contact email addresses (Firstname Lastname <firstname.lastname@email.example>
)
License
: the license under which you publish this code. It has to be permissive enough to allow us to publish the code after the competition.
Tracks
: a comma-separated list of tracks this image should participate in. Use only the following terms to identify tracks: optimal
, satisficing
, agile
. For example, to run your planner in the optimal and agile track use Tracks optimal, agile
.
Your Apptainer recipe must contain all of the following labels describing
supported PDDL features. Each label must be set to either yes
or no
, or if
the feature is supported only partially, then set it to partially,
followed
by the description of what is and isn't supported:
SupportsDerivedPredicates
: specify whether your planner supports (:derived ...)
constructSupportsUniversallyQuantifiedPreconditions
: (forall ...)
in actions' preconditionsSupportsExistentiallyQuantifiedPreconditions
: (exists ...)
in actions' preconditionsSupportsUniversallyQuantifiedEffects
: (forall ...)
in actions' effectsSupportsNegativePreconditions
: (not (...))
in actions' preconditionsSupportsEqualityPreconditions
: (= ?x obj1)
, for action parameter ?x
and constant obj1
, in actions' preconditionsSupportsInequalityPreconditions
: (not (= ?x ?y))
or (not (= ?x obj1))
, for action parameters ?x
and ?y
, and constant obj1
, in actions' preconditionsSupportsConditionalEffects
: (when ...)
in actions' effectsSupportsImplyPreconditions
: (imply ...)
in actions' preconditionsTo improve reproducibility, we require Apptainer images to be self-contained and licensed appropriately.
The recipe should copy the content of the repository into the container at the start of the build. In particular, do not clone repositories in the recipe.
If you use third-party libraries, either install them through standard package managers (apt, yum, pip), or copy them into the repository. Please do not use git submodules to include dependencies.
If possible, use explicit versions. For example, do not use ubuntu:latest
as
your base image but pick a specific version. If you install packages through pip,
pick specific versions of those packages.
If your build depends on closed-source libraries that require a license, please contact us.
If you use a portfolio of existing planners, it is up to you to get permission from the authors of the portfolio components and give appropriate credit and licensing. We recommend contacting the planner authors.
In addition to reproducibility and licensing issues, we ask that you make your image as small as possible using the following tricks:
Use a multi-stage build where one stage is used for compiling the planner and one for running the planner. Copy the compiled planner from the first stage to the second, and only copy/install the files that are required at runtime. The size of the compilation stage then does not matter and the second stage can be limited to contain only essential files.
Strip binaries after compilation
Use small packages. For example, use python-minimal instead of python if possible.
When a competition team registers (see above), we create a private repository (or multiple repositories if needed) and add competitors as users with write access. After the “feature stop” deadline (March 10, 2023), we allow competitors to send only a pull request with bug fixes. We will review every pull request with its accompanying description of the bug fix to make sure that no big changes or parameter tuning is committed to the repository.
To help us with the debugging process, in contrast to previous years, planner authors will be responsible for detecting if the run of their planner and our analysis of the results was successful. After the feature stop deadline, we will run all planners on all tasks and give the participants access to the results of their planners. For each run, the data will contain the log files of the planner, measured time and memory consumption, exit code, and our conclusion about what this means in terms of solving the instance. We ask participants to check their results for any errors. If an error was caused by a bug in the planner, please send a pull request on Github with a detailed description of the bug and the fix. If the error was on our side (e.g., malformed PDDL) let us know as soon as possible. We will do at least two rounds of this starting after the "feature stop" deadline (exact timeline TBD).
All competitors must submit an abstract (max. 300 words) and an up to 8-page paper describing their planners. After the competition we ask the participants to analyze the results of their planner and submit an extended version of their paper. An important requirement for IPC 2023 competitors is to give the organizers the right to post their paper and the source code of their planners on the official IPC 2023 web site, and the source code of submitted planners must be released under a license allowing free non-commercial use.
Contact us: ipc2023-classical@googlegroups.com