Introduction

This is the implementation of the planning domain repair approach described in the paper Towards Automated Modeling Assistance: An Efficient Approach for Repairing Flawed Planning Domains accepted by AAAI 2023. The program takes as input a (potentially flawed) PDDL domain file, a list of PDDL task files, and a list of plan files containing plans each of which is supposed to be a solution to the respective given planning task (but it might not be the case due to the flaws in the given domain) and outputs a repaired domain in which every given plan is guaranteed to be a solution.

The implementation relies on the PDDL parser (in the directory fd) that is copied from the project LAPKT.

Run the Domain Repairer

Requirement

The program is written in python. Thus, you must have Python installed in your device in order to run the code.

Command

You can use the following command to run the domain repairer, assuming that you are in the directory containing this README file.

python diagnoser.py --domain domain.pddl --tasks task-1.pddl ... task-n.pddl --plans planfile-1 ... planfile-n

Note that the plan contained in each file must be of the format that is identical to the output format of the Fast-downward planner, shown as follows.

(action-name-1 para-1 para-2 ... para-n)
                    .
                    .
                    .
(action-name-k para-1 para-2 ... para-n)

The repairer will output a repaired domain in a PDDL file if you provide a path to the argument --out_domain. E.g.,

python diagnoser.py --domain domain.pddl --tasks task-1.pddl ... task-n.pddl --plans planfile-1 ... planfile-n --out_domain output_path

The output PDDL domain file will be on the path: output_path/domain-repaired.pddl (Please do not give the PDDL file name as a part of the parameter!). Similarly, the repairer will write found repairs to a file called diagnosis if a path is given to the argument --out_diagnosis on which the file will be produced. You can use the command

python diagnoser.py --help

to check all arguments accepted by the program.

Cite

@InProceedings{Lin2023RepairingClassicalModels,
  author     = {Songtuan Lin and Alban Grastien and Pascal Bercher},
  booktitle  = {Proceedings of the 37th AAAI Conference on Artificial Intelligence (AAAI 2023)},
  title      = {Towards Automated Modeling Assistance: An Efficient Approach for Repairing Flawed Planning Domains},
  year       = {2023},
  publisher  = {AAAI Press},
}