Dojo

A differentiable physics engine for robotics

• arXiv preprint: https://arxiv.org/abs/2203.00806
• Python interface: https://github.com/dojo-sim/dojopy
• site: https://sites.google.com/view/dojo-sim
• video presentation: https://youtu.be/TRtOESXJxJQ

Update April 2023

• We are no longer actively developing Dojo, but pull requests are always welcome.
• We have updated or removed examples to account for changes since the initial version of Dojo.
• Additional developments on differentiable simulation:
  • Differentiable collision detection (Kevin Tracy): capsules, convex primitives
  • Single-level contact dynamics + collision detection (Simon Le Cleac'h): Silico

Examples

Simulation

Learning and Control

System Identification

Interfacing Other Packages

ReinforcementLearning.jl: DQN	ControlSystems.jl: LQR

Installation

Dojo can be added via the Julia package manager (type ]):

pkg> add Dojo

For convenience mechanisms and environments, add DojoEnvironments additionally:

pkg> add DojoEnvironments

Citing

@article{howelllecleach2022,
	title={Dojo: A Differentiable Simulator for Robotics},
	author={Howell, Taylor and Le Cleac'h, Simon and Bruedigam, Jan and Kolter, Zico and Schwager, Mac and Manchester, Zachary},
	journal={arXiv preprint arXiv:2203.00806},
	url={https://arxiv.org/abs/2203.00806},
	year={2022}
}

How To Contribute

Please submit a pull request or open an issue. See the docs for contribution ideas.