Skip to content
drawtonomy drawtonomy

drawtonomy and Scenic

Not a comparison — different layers

Section titled “Not a comparison — different layers”

drawtonomy and Scenic are not competing tools. They sit at different layers:

  • Scenic — a probabilistic programming language for specifying scenarios as constraints over space and behavior, and sampling many concrete scenes from a single abstract spec. Developed at UC Berkeley.
  • drawtonomy — a 2D browser whiteboard for sketching a single concrete scene visually.

Scenic answers “what is the space of scenarios I want to test?” drawtonomy answers “what does one specific scene look like?”

What Scenic covers

Section titled “What Scenic covers”

Per the official documentation and GitHub repository, and the original PLDI 2019 paper:

  • A domain-specific language for specifying scenarios as probabilistic constraints.
  • Scenario sampling — generate many concrete scenes from one abstract specification.
  • Interfaces to simulators, with documented CARLA integration.
  • Used in academic benchmarks for evaluating autonomous driving systems.
  • Open-source under a BSD license.

For testing scenarios at scale — generating fleets of corner cases from a small spec — Scenic is one of the established tools.

How drawtonomy fits

Section titled “How drawtonomy fits”

drawtonomy has no DSL, no sampling, no probabilistic generation. It is meant for one concrete scene at a time, drawn by hand.

The two tools can sit next to each other in different roles:

  • Scenic generates the spectrum of scenarios to test.
  • drawtonomy produces a visual figure of one representative concrete scene for a paper, slide, or test plan — alongside Scenic-generated data.

If you’re using Scenic to generate scenarios, drawtonomy doesn’t replace it. If you’re sketching a single scene by hand and don’t need probabilistic sampling, Scenic isn’t the right tool either.

Part of the same ecosystem

Section titled “Part of the same ecosystem”

Scenic is one of the most active research contributions to scenario specification, and it has shaped how the academic AV testing community thinks about probabilistic scenarios. drawtonomy works in a much narrower role — visual single-scene sketching — but emits the same OpenSCENARIO format that Scenic can convert to, and targets the same downstream players. Both tools, alongside CARLA + ScenarioRunner, esmini, scenariogeneration, and others, contribute to the same open scenario-testing community from different angles.

See also

Section titled “See also”