This is the official implementation of Learning to Drive in New Cities Without Human Demonstrations.
DOCKER_BUILDKIT=1 docker build --build-arg USE_CUDA=true --tag gpudrive:latest --progress=plain .
# Run Docker container
docker run --gpus all -it --rm --shm-size=20G -v ${PWD}:/workspace gpudrive:latest /bin/bash
# Build gpudrive
mkdir build
cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_POLICY_VERSION_MINIMUM=3.5 && find external -type f -name "*.tar" -delete
make -j
NuPlan scenarios (Pittsburgh, Singapore, Boston) are available as the NOMAD dataset on Hugging Face.
python baselines/bc/collect_multi_discrete_demo.py
python baselines/bc/bc.py
python baselines/goal/generate_scene_heuristic_modular.py
python baselines/bcsp/ppo_finetuning.py
python baselines/eval/wosac_evaluation.py
If you find this repo helpful, don't forget to give us a star! If this repo helps your research, please cite this work:
@article{wang2026learning,
title={Learning to Drive in New Cities Without Human Demonstrations},
author={Wang, Zilin and Rahmani, Saeed and Cornelisse, Daphne and Sarkar, Bidipta and Goldie, Alexander David and Foerster, Jakob Nicolaus and Whiteson, Shimon},
journal={arXiv preprint arXiv:2602.15891},
year={2026}
}
@inproceedings{kazemkhani2025gpudrive,
title={GPUDrive: Data-driven, multi-agent driving simulation at 1 million FPS},
author={Saman Kazemkhani and Aarav Pandya and Daphne Cornelisse and Brennan Shacklett and Eugene Vinitsky},
booktitle={Proceedings of the International Conference on Learning Representations (ICLR)},
year={2025},
url={https://arxiv.org/abs/2408.01584},
eprint={2408.01584},
archivePrefix={arXiv},
primaryClass={cs.AI},
}