Unlock the Future of Autonomous Driving with NAVSIM: Data-Driven Simulation and Benchmarking
Are you ready to revolutionize your autonomous vehicle research? NAVSIM, the cutting-edge data-driven simulation and benchmarking platform, is here to accelerate your progress. Discover how NAVSIM can transform your approach to autonomous driving development.
What is NAVSIM and Why Should You Care?
NAVSIM provides simulation-based metrics, like progress and time to collision, for evaluating end-to-end driving systems. Crucially, it operates under conditions where the policy has limited influence on the environment. This enables fast metric computation, better aligning with real-world, closed-loop evaluations than traditional methods.
- Realistic Simulations: Experience bird's-eye view abstractions of scenes for short-horizon simulations.
- Efficient Benchmarking: Quickly assess your autonomous vehicle's performance with precise metrics.
- Improved Alignment: Benefit from simulations that closely mirror closed-loop evaluations.
Getting Started with NAVSIM: Your Fast Track to Autonomous Vehicle Excellence
Ready to dive in? NAVSIM offers comprehensive resources to get you started quickly.
- Download and Installation: Set up NAVSIM with ease using our straightforward installation guide.
- Understanding Agents: Learn to create and customize intelligent agents for diverse simulation scenarios.
- Data Format and Classes: Master the NAVSIM data format to unlock the full potential of our datasets.
- Dataset Splits: Optimize your training and testing with carefully designed dataset splits.
- Extended PDM Score: Decipher and leverage the Potential for Driving Mileage (PDM) score for advanced evaluations.
- Traffic Simulation: Explore NAVSIM's traffic simulation capabilities, including reactive traffic agents.
- Leaderboard Submission: Showcase your results and compete with other researchers on the NAVSIM leaderboard.
NAVSIM Changelog: Stay Ahead of the Curve
Keep up-to-date with the latest NAVSIM advancements:
- v2.1.1 (2025/04/13): Updated dataset for the Hugging Face Warmup leaderboard with minor fixes. Remember to re-download the synthetic dataset to submit.
- v2.1 (2025/04/08): New dataset for the Hugging Face Warmup leaderboard, plus support for two-stage reactive traffic agents.
- v2.0 (2025/02/28): Extended PDM Score, two-stage pseudo closed-loop simulation, and reactive traffic agent policies.
- v1.1 (2024/09/03): Leaderboard on Hugging Face, baseline checkpoints, and updated docs.
- v1.0 (2024/04/21): Parallelization of metric caching, Transfuser baseline, and standardized training/test splits.
License and Citation: Building on a Foundation of Open Innovation
NAVSIM is released under the Apache 2.0 license, fostering collaboration and innovation in autonomous vehicle research. When using NAVSIM, please cite our work:
@inproceedings { Dauner2024NEURIPS,
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},
title = { NAVSIM: Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarking},
booktitle = { Advances in Neural Information Processing Systems (NeurIPS)},
year = { 2024},
}
Explore Related Resources: Expand Your Autonomous Driving Toolkit
Dive deeper into the world of autonomous driving with these invaluable resources:
- SLEDGE
- tuPlan garage
- CARLA garage
- Survey on E2EAD
- PlanT
- KING
- TransFuser
- NEAT
NAVSIM is transforming the way autonomous vehicles are developed and tested through data-driven approaches. Embrace the power of NAVSIM and drive the future of autonomous technology today. Use our non-reactive autonomous vehicle simulation tool to enhance your autonomous vehicle development.