Cross-Dataset Reimplementation of NAVSIM

Paper | Video | Documentation

"Because good things come in 123D."

🏗️ Under construction. Expect things to change, break, and improve as we build it out. Comments, suggestions, and contributions are very welcome!

nav123d reimplements NAVSIM — data-driven, non-reactive autonomous-vehicle simulation and benchmarking — on top of the 123D unified data framework, with similar agents, metrics, and training/evaluation pipeline run across datasets. It ships baseline agents (constant-velocity, ego-status MLP, PDM, TransFuser) together with training and evaluation entry points.

Installation

Editable pip install (Python 3.9 - 3.12):

pip install -e .

... or in a fresh conda environment (Python 3.9 - 3.12)

conda create -n nav123d python=3.12 -y
conda activate nav123d
pip install uv
uv pip install -e .[dev]

See the installation docs for dataset setup and environment variables.

(back to top)

Changelog

v0.1.0 (2026-05-28)

• Initial release of the package and documentation.
• Baseline agents (constant-velocity, ego-status MLP, PDM, TransFuser) with training and evaluation entry points.

(back to top)

Citation

All assets and code in this repository are under the Apache 2.0 license unless specified otherwise.

@inproceedings{Dauner2024NEURIPS,
	title = {NAVSIM: Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarking},
	author = {Daniel Dauner and Marcel Hallgarten and Tianyu Li and Xinshuo Weng and Zhiyu Huang and Zetong Yang and Hongyang Li and Igor Gilitschenski and Boris Ivanovic and Marco Pavone and Andreas Geiger and Kashyap Chitta},
	booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
	year = {2024},
}

@inproceedings{Dauner2026ARXIV,
  title={123D: Unifying Multi-Modal Autonomous Driving Data at Scale},
  author={Dauner, Daniel and Charraut, Valentin and Berle, Bastian and Li, Tianyu and Nguyen, Long and Wang, Jiabao and Jing, Changhui and Igl, Maximilian and Caesar, Holger and Ivanovic, Boris and Geiger, Andreas and Chitta, Kashyap},
  journal={arXiv preprint arXiv:2605.08084},
  year={2026}
}

(back to top)