Michael Katz
contact information
links
Professional Associations
Professional Associations: Association for the Advancement of Artificial Intelligence (AAAI)Software
The page includes links to the open-sourced software I was partially responsible for developing.
IBM Research AI Planning Service (DockerHub Image)
K* search implementation of a Top-K planner (DockerHub Image)
Cerberus planner, post-IPC 2018 version (DockerHub Image)
Forbid-Iterative (FI) Planner suite for top-k, top-quality, and diverse computational tasks (DockerHub Image)
Diversity score computation for a set of plans (DockerHub Image)
Data
IPC: A Graph Data Set Compiled from International Planning Competitions
IPC: An IMAGE Data Set Compiled from International Planning Competitions
PDDL domain inspired by risk management problem
PDDL benchmark generated from Data Science grammar
PDDL instances of structurally restricted planning tasks
Competitions
I have participated in the deterministic part of the IPC-2018 - International Planning Competition 2018
with multiple planners. One of them, Delfi has won the sequential optimal track. The source code for Delfi can be found here.
Other planners from IPC2018 include
- Metis, IPC 2018 version
- MERWIN, IPC 2018 version (Extends Mercury with Novelty heuristic [1])
- Cerberus, IPC 2018 version (Extends Mercury with Novelty heuristic [1] and native support for conditional effects [2])
- Mercury 2014, IPC 2018 version
[1] M. Katz, N. Lipovetzky, D. Moshkovich, A. Tuisov, Adapting Novelty to Classical Planning as Heuristic Search, in Proceedings of The International Conference on Automated Planning and Scheduling (ICAPS), Pittsburgh, PA, USA, 2017.
[2] M. Katz, Red-Black Heuristics for Planning Tasks with Conditional Effects, in Proceedings of The 33rd AAAI Conference on Artificial Intelligence (AAAI), Honolulu, HI, USA, 2019.
I have participated in the deterministic part of the IPC-2014 - International Planning Competition 2014
with two planners, Mercury for the sequential satisficing track and Metis for the sequential optimal track.
Mercury (with Joerg Hoffmann) has won two awards:
- Runner-Up in the sequential satisficing track, and
- Innovative Planner Award.
The source code for the planners can be found here (or by request):
More details on Mercury:
Mercury is a classical planner - a solver for classical (deterministic) planning tasks specified in PDDL 3.1.
It is an open source planner, built on top of the Fast Downward planning system. The source code is available here.
Running the planner:
- To build the planner run ./build
- To run the planner use ./plan <DOMAIN> <PROBLEM> <PLAN_FILE>
PDDL examples can be found at https://bitbucket.org/aibasel/downward-benchmarks
Mercury has won two awards at the 2014 International Planning Competition:
- Runner-up of the Deterministic Sequential Track
- Innovative Planner Award
Papers related to Mercury:
- Mercury Planner: Pushing the Limits of Partial Delete Relaxation
M Katz, J HoffmannEighth International Planning Competition (IPC 2014)
- Red-black planning: a new systematic approach to partial delete relaxation
C Domshlak, J Hoffmann, M KatzArtificial Intelligence 221 (April 2015), 73–114
- Who said we need to relax all variables?
M Katz, J Hoffmann, C Domshlak23rd International Conference on Automated Planning and Scheduling (ICAPS 2013)
- Red-Black Relaxed Plan Heuristics
M Katz, J Hoffmann, C Domshlak27th AAAI Conference on Artificial Intelligence (AAAI 2013)
- Red-Black Relaxed Plan Heuristics Reloaded
M Katz, J HoffmannSixth Annual Symposium on Combinatorial Search (SoCS 2013)
- Pushing the Limits of Partial Delete Relaxation: Red-Black DAG Heuristics
M Katz, J HoffmannHeuristics and Search for Domain Independent Planning (HSDIP 2014)