GeoScenario: An Open DSL for Autonomous Driving Scenario Representation


Last Update: 05-May-2021


GeoScenario is a Domain Specific Language (DSL) for scenario representation to support ADS scenario-based testing in simulation. The language covers a variety of elements that compose typical traffic situations that need to be formally declared and executed on self-driving vehicle testing. The language is XML based, simple, extensible, and built on top of Open Street Map (OSM) standard.

This documentation website provides a detailed specification of the language and all its components, including examples and usage instructions. Tools and XML files for all scenarios mentioned in this document are available in the project's repository at https://git.uwaterloo.ca/wise-lab/geoscenario.

There are multiple components that are part of GeoScenario framework:

  • GeoScenario: An open language for scenario definition.
  • SDV: a driver-vehicle model to simulate human-traffic behavior with multiple vehicles in scenario-based testing.
  • GeoScenario Server (GSServer): A scenario simulation server that can parse and run GeoScenario files and multiple SDV model instances. The server can be integrated to any simulation tool in co-simulation (using shared memory or sockets).
  • GeoScenario Unreal Client (GSClient): a sample Unreal Engine based client that connects to GSServer to render the simulation with high fiedlity visuals. Can be used as an example of GSServer-Client integration to run scenarios in co-simulation. An experimental integration with Carla is also available as part of the tool-set.
  • Scenario Editor: a modified JOSM that can be used as GUI to facilitate scenario design.

Both the language and models are implementation independent. This documentation will focus on the logic behind the language, the models, how scenarios and behaviors can be defined and how it must be interpreted. GSServer and GSClient are reference implementations and will be used throughout this documentation to exemplify the language and model usage in practice. Additional details are available in both papers below.

Articles

IV’19 paper:

@INPROCEEDINGS{queiroz19,
    title = {{GeoScenario}: An Open DSL for Autonomous Driving Scenario Representation},
    author = {Rodrigo Queiroz and Thorsten Berger and Krzysztof Czarnecki},
    booktitle={2019 IEEE Intelligent Vehicles Symposium (IV)}, 
    year = {2019}
}

And the GeoScenario SDV Model paper:

@ARTICLE{queiroz22,
    title = {A Driver-Vehicle Model for ADS Scenario-based Testing},
    author = {Queiroz, Rodrigo and Sharma, Divit and Caldas, Ricardo and Czarnecki, Krzysztof and García, Sergio and Berger, Thorsten and Pelliccione, Patrizio},
    publisher = {arXiv},
    year = {2022},
    doi = {10.48550/ARXIV.2205.02911},
    url = {https://arxiv.org/abs/2205.02911},
}

```