We will open source the complete code after the paper is accepted ！
- [2024/10.08]: 🔥 We release the HE-Drive paper on arXiv !
HE-Drive is a groundbreaking end-to-end autonomous driving system that prioritizes human-like driving characteristics, ensuring both temporal consistency and comfort in generated trajectories. By leveraging sparse perception for key 3D spatial representations, a DDPM-based motion planner for generating multi-modal trajectories, and a VLM-guided trajectory scorer for selecting the most comfortable option, HE-Drive sets a new standard in autonomous driving performance and efficiency. This innovative approach not only significantly reduces collision rates and improves computational speed compared to existing solutions but also delivers the most comfortable driving experience based on real-world data.
@article{wang2024he,
title={HE-Drive: Human-Like End-to-End Driving with Vision Language Models},
author={Wang, Junming and Zhang, Xingyu and Xing, Zebin and Gu, Songen and Guo, Xiaoyang and Hu, Yang and Song, Ziying and Zhang, Qian and Long, Xiaoxiao and Yin, Wei},
journal={arXiv preprint arXiv:2410.05051},
year={2024}
}
Please kindly star ⭐️ this project if it helps you. We take great efforts to develop and maintain it 😁.
- nuScenes
- Real-World Data
- OpenScene/NAVSIM
Many thanks to these excellent open source projects: