Temporal Fast Downward

This is a slightly modified version of TFD v0.4 (from IPC 2014).

What is TFD? TFD is a temporal planning system that successfully participated in the temporal satisficing track of the 6th International Planning Competition 2008. It is released under the terms of the GPL. A snapshot of the source code is available in the download section.

The algorithms used in TFD are described in the ICAPS 2009 paper [1]. TFD is based on the Fast Downward planning system and uses an adaptation of the context-enhanced additive heuristic [2] to guide the search in the temporal state space induced by the given planning problem.

TFD is mainly developed by Gabriele Röger, Patrick Eyerich, Christian Dornhege, and Robert Mattmüller.

Original Links

Installation

git clone https://github.com/neighthan/tfd
cd tfd
./build

Usage

downward/tfd domain_path problem_path solution_path [planner_config]

You can find a several benchmark domain and problem files in the benchmarks directory. To test your installation, try running

woodworking-numeric takes 0.01 seconds, plan takes 50 seconds

downward/tfd benchmarks/openstacks-numericadl/domain.pddl benchmarks/openstacks-numericadl/p01.pddl plan.txt

This should take around 5 minutes to generate an optimal plan. The duration of the optimal plan should be about 82 seconds.
The planner will periodically output improved plans, numbered sequentially as plan.txt.1, plan.txt.2, etc. until it times out or finds the optimal plan.

You can add a symlink to tfd/downward/tfd from somewhere on your path (e.g. ~/bin) to run this command from anywhere.

Modifications

convert from python 2 to 3 + python formatting changes
rename plan.py to tfd, make it robust to running from different directories, delete the intermediate files by default
don't treat errors as warnings during compilation

Citation

If you use TFD in your research, please cite

@inproceedings{Eyerich2009UsingTC,
  title={Using the Context-enhanced Additive Heuristic for Temporal and Numeric Planning},
  author={Patrick Eyerich and Robert Mattm{\"u}ller and Gabriele R{\"o}ger},
  booktitle={ICAPS},
  year={2009}
}