Open Martin36 opened 9 months ago
@Martin36, the initial PDDL state is computed by this function. You can probably spawn scenes from the JSON file (like in augment_trajectories.py) and use this function to dump the PDDL descriptions.
In the readme, it says that the PDDL is included in the full dataset,
Yeah sorry, I no longer have access to the uncompressed files to sort out the PDDL descriptions :(
I downloaded the full dataset and extracted it. The PDDL problem files seem to be there. However, I can't find any PDDL domain file. Do you know if there is supposed to be a domain file in the data folder? I was not able to extract the full data because my hard drive became full.
Looking at the training data included in the repo, in the folder
json_2.1.0
, it seems that this contains the PDDL plans for reaching a specific goal. The objects are represented by their positions and rotations in the "scene" dictionary. However, I was wondering if the PDDL domain and symbolically represented initial state is available somewhere.In the readme, it says that the PDDL is included in the full dataset, but since that is over 100GB it is a bit of a hassle to download. Is it possible to get just a sample from the full dataset? And how is the PDDL in the full dataset different from the one in the data in the repo?