This repository contains code for the paper:

• Let's Collaborate: Regret-based Reactive Synthesis for Robotic Manipulation (ICRA 22) - (branch)
• Efficient Symbolic Approaches for Quantitative Reactive Synthesis with Finite Tasks (IROS 23) - (branch)

Primarily, we use this code to

1. Parse the PDDL File, construct a two-player turn-based game abstraction from a PDDL file. The syntax of the game is same as FOND.
2. Synthesize a strategy using this toolbox, and
3. Simulate the strategy using Pybullet.

Note: the Pybullet sim is old will be phased out soon with probably a better simulator. Further, the keyword :non-deterministic and the additional operator oneof is not yet supported. These are standard ways of specifying non-deterministic outcomes under an action using the PDDL Semantics (Refer PRP Paper, PRP Code). In my file, I add non-determinism as human-move action (See my code).

• Karan Muvvala <[email protected]>

• clone this repo with:

git clone --recurse-submodules https://github.com/aria-systems-group/PDDLtoSim.git .

The --recurse-submodule will automatically initialize and update each submodule in the repository.

Make sure you have Docker installed. If not, follow the instructions here.

1. cd into the root of the project

2. Build the image from the Dockerfile

docker build -t <image_name> .

Note: the dot looks for a Dockerfile in the current repository. Then spin an instance of the container by using the following command

docker run -it --name <docker_container_name> <docker image name>

For volume binding

docker run -v <HOST-PATH>:<Container-path>

For example, to volume bind your local directory to the pddltosim folder inside the Docker, use the following command

docker run -it -v $PWD:/root/pddltosim --name <dokcer_container_name> <image_name>

Here <docker_container_name> is any name of your choice and <image_name> is the docker image name from above. -it and -v are flags to run an interactive terminal and volume bind respectively.

Additionally, if you are more used to GUI and would like to edit or attach a container instance to VSCode (Link) then follow the instructions below:

1. Make sure you have the right VS code extensions installed
   • install docker extension
   • install python extension
   • install remote container extension
   • Now click on the Remote Explore tab on the left and attach VScode to a container.
2. This will launch a new vs code attached to the container and prompt you to a location to attach to. The default is root, and you can just press enter. Congrats, you have attached your container to VSCode.

• install anaconda or miniconda

• install spot if you are going to construct a DFA using an LTL formula.

• change into this repo's directory:

cd PDDLtoSim

• create the conda environment for this library:

cd conda && conda env create -f environment.yml

• activate the conda environment:

conda activate regret_syn_env

cd into the root directory, activate the conda env and run the following command

python3 main.py

All the tests related scripts are available in the tests/ directory. I use python unittest for testing individual components of my source code. Here are some commands to run the tests:

To run a specific test package:

python3 -m unittest discover -s tests.<directory-name> -bv

To run a specific test script:

python3 -m tests.<directory-name>.<module-nane> -b

To run all tests:

python3 -m unittest -bv

For more details see the tests/README.md. Note, all commands must be run from <root/of/project>.

Here are some glimpses of the simulated strategy using this toolbox. In our simulation world we consider two region of interest. A human region and a robot region. We say that the human (not shown) can reach and manipulate the boxes placed on the right side (human region) but not the ones placed on the left (robot region).

Robot building an arch with black boxes as supports and white box on top in either of the regions. The human intervenes twice and the robot is using a regret-minimizing strategy.

Robot placing objects in the specific pattern. The black box should be placed at the bottom location, the grey box in the middle and the white box should in the top location in either of the regions.

You can build spot from source, official git repo or Debian package. If you do source installation, then run the following command to verify your installation

ltl2tgba --version

If your shell reports that ltl2tgba is not found, add $prefix/bin to you $PATH environment variable by using the following command

export PATH=$PATH:/place/with/the/file

Spot installs five types of files, in different locations. $prefix refers to the directory that was selected using the --prefix option of configure (the default is /usr/local/).

1. command-line tools go into $prefix/bin/
2. shared or static libraries (depending on configure options) are installed into $prefix/lib/
3. Python bindings (if not disabled with --disable-python) typically go into a directory like $prefix/lib/pythonX.Y/site-packages/ where X.Y is the version of Python found during configure.
4. man pages go into $prefix/man
5. header files go into $prefix/include

Please refer to the README file in the tar ball or on their GitHub page for more details on trouble shooting and installation.

If the code is useful in your research, and you would like to acknowledge it, please cite one of the following paper

• Let's Collaborate: Regret-based Reactive Synthesis for Robotic Manipulation (Explicit Approach) (paper):

@INPROCEEDINGS{muvvala2022regret,
  author={Muvvala, Karan and Amorese, Peter and Lahijanian, Morteza},
  booktitle={2022 International Conference on Robotics and Automation (ICRA)}, 
  title={Let's Collaborate: Regret-based Reactive Synthesis for Robotic Manipulation}, 
  year={2022},
  pages={4340-4346},
  doi={10.1109/ICRA46639.2022.9812298}}

• Efficient Symbolic Approaches for Quantitative Reactive Synthesis with Finite Tasks (Symbolic Approach) (paper)

@article{muvvala2023efficient,
  title={Efficient Symbolic Approaches for Quantitative Reactive Synthesis with Finite Tasks},
  author={Muvvala, Karan and Lahijanian, Morteza},
  booktitle={2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  year={2023},
  pages={8666-8672},
  doi={10.1109/IROS55552.2023.10342496}}
}

Please contact me if you have questions at: [email protected]