DOROTHIE

Spoken Dialogue for Handling Unexpected Situations in Interactive Autonomous Driving Agents

Research Paper | Github | Huggingface

Authored by Ziqiao Ma, Ben VanDerPloeg, Cristian-Paul Bara, Yidong Huang, Eui-In Kim, Felix Gervits, Matthew Marge, Joyce Chai

DOROTHIE (Dialogue On the ROad To Handle Irregular Events) is an innovative interactive simulation platform designed to create unexpected scenarios on the fly. This tool facilitates empirical studies on situated communication with autonomous driving agents.

Quickstart

The DOROTHIE platform is developed on the basis of the Carla 0.9.11 release.

download carla and import additional assets

$ apt install carla-simulator=0.9.11
$ cd /opt/carla-simulator/
$ gdown 1upkC0ts5_i_YSyCx7lGdQbgcsezfo5Z9
$ cp  CustomPropsPackage_0.9.11-dirty.tar.gz Import/
$ ./ImportAssets.sh

clone repo

$ git clone https://github.com/h6kplus/DOROTHIE.git DOROTHIE
$ export DOROTHIE_ROOT=$(pwd)/DOROTHIE

Install requirements:

$ virtualenv --python=python3.7 --system-site-packages dorothie_env # or whichever package manager you prefer
$ source dorothie_env/bin/activate

$ cd $DOROTHIE_ROOT
$ pip install --upgrade pip
$ pip install -r requirements.txt
$ export PYTHONPATH=$PYTHONPATH:/opt/carla-simulator/PythonAPI/carla:/opt/carla-simulator/PythonAPI/carla/dist/carla-0.9.11-py3.7-linux-x86_64.egg

Start Carla server

$ /opt/carla-simulator/CarlaUE4.sh

Run the co_wizard server in another terminal:

$ cd co_wizard
$ python3 co_wizard.py --tpl ../sb_templates/tpl0.json --metaconfig ../sb_templates/metaconfig_town03.json

Run ad_wizard in another terminal:

$ source dorothie_env/bin/activate
$ export PYTHONPATH=$PYTHONPATH:/opt/carla-simulator/PythonAPI/carla:/opt/carla-simulator/PythonAPI/carla/dist/
$ cd $DOROTHIE_ROOT/ad_wizard
$ python3 ad_wizard.py

Run the human participant interface in another terminal:

$ source dorothie_env/bin/activate
$ export PYTHONPATH=$PYTHONPATH:/opt/carla-simulator/PythonAPI/carla:/opt/carla-simulator/PythonAPI/carla/dist/
$ cd $DOROTHIE_ROOT/dorothy
$ python3 dorothy.py

Do human study on separate machines

If the server address is carlaserver.some, you can run ad_wizard and the human participant interface on different machines (after installing Carla):

Run ad_wizard in another machine:

$ source dorothie_env/bin/activate
$ export PYTHONPATH=$PYTHONPATH:/opt/carla-simulator/PythonAPI/carla:/opt/carla-simulator/PythonAPI/carla/dist/
$ cd $DOROTHIE_ROOT/ad_wizard
$ python3 ad_wizard.py -wh carlaserver.some -ch carlaserver.some

Run the human participant interface in another machine:

$ source dorothie_env/bin/activate
$ export PYTHONPATH=$PYTHONPATH:/opt/carla-simulator/PythonAPI/carla:/opt/carla-simulator/PythonAPI/carla/dist/
$ cd $DOROTHIE_ROOT/dorothy
$ python3 dorothy.py -wh carlaserver.some -ch carlaserver.some

Experiment with Different Metaconfig and Storyboard Template Settings

We have developed metaconfigs for towns 1, 2, 3, and 5 in Carla. Because some unique assets belong to specific towns, some templates can only be used with a subset of metaconfigs. Feel free to experiment with different combinations of metaconfig and template settings!

The SDN Dataset

• The textual transcripts and annotations of SDN dialogue sessions are available on HuggingFace.
• The public SDN dataset (without speech data) is available on Google Drive and DropBox.
• The speech data in SDN will be available upon request for ethical considerations, please fill in the Google Form to apply for access.

Citation

If you find this repository useful, please cite our work:

@inproceedings{ma-etal-2022-dorothie,
    title = "{DOROTHIE}: Spoken Dialogue for Handling Unexpected Situations in Interactive Autonomous Driving Agents",
    author = "Ma, Ziqiao  and
      VanDerPloeg, Benjamin  and
      Bara, Cristian-Paul  and
      Huang, Yidong  and
      Kim, Eui-In  and
      Gervits, Felix  and
      Marge, Matthew  and
      Chai, Joyce",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2022",
    month = dec,
    year = "2022",
    address = "Abu Dhabi, United Arab Emirates",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.findings-emnlp.354",
    pages = "4800--4822",
}