Motivation
Given the safety concerns and high costs associated with real-world autonomous driving testing, high-fidelity simulation techniques have become crucial for advancing the capabilities of autonomous systems.
At the RealADSim Workshop @ ICCV 2025, we aim to explore two key questions:
- How well can we render? While NVS methods have made significant progress in generating photorealistic urban scenes, their performance still lags in extrapolated viewpoints when only a limited viewpoint is provided during training. However, extrapolated viewpoints are essential for closed-loop simulation. Improving the accuracy and consistency of NVS across diverse viewing angles is critical for ensuring that these simulators provide reliable environments for driving evaluation.
- How well can we drive? Despite challenges in extrapolated viewpoint rendering, existing methods enable photorealistic simulators with reasonable performance when trained on dense views. These NVS-based simulators allow autonomous driving models to be tested in a fully closed-loop manner, bridging the gap between real-world data and interactive evaluation. This shift allows for benchmarking autonomous driving algorithms under realistic conditions, overcoming the limitations of static datasets.
To seek answers to these questions, we are hosting two separate challenges at the workshop. This challenge focuses on the second question — evaluating the performance of autonomous driving algorithms in a closed-loop simulation setting.
If you're interested in the first question, please refer to the extrapolated novel view synthesis competition.
Task Description
The challenge focuses on evaluating autonomous driving algorithms using HUGSIM, a photorealistic, closed-loop driving simulator. HUGSIM provides a diverse set of challenging, photo-realistic urban scenarios, including oncoming traffic and cut-in behaviors.
Unlike typical challenges that require submissions as static files, this competition requires participants to submit models and code. This is because the closed-loop evaluation demands real-time interaction between the driving algorithm and the simulator at every timestep. Please note that your submissions are securely protected and will only be used for closed-loop evaluation. For more information on how we handle your data and ensure privacy, please refer to the Privacy section.
Evaluation Metric
The primary metric of this challenge is HD-Score (HUGSIM Driving Score) defined as:
The HD-Score at a single timestamp consists of two categories of components:
- Driving policy items, including no collisions (NC) and drivable area compliance (DAC), which are critical for ensuring driving safety.
- Contributory items, including time-to-collision (TTC) and comfort (COM), which may not directly lead to failure cases when low, but still contribute to overall driving quality. The TTC and COM terms are weighted by 5 and 2, respectively.
The HD-Score is computed as a weighted average across all simulation timestamps and then multiplied by the global route completion score Rc.
Privacy Assurance
Your model checkpoints should be stored in a private Huggingface model hub.
The tested algorithms are hosted on a Huggingface instance server, which will be destroyed once the evaluation is completed. We do not have access to this server at any point.
To ensure transparency and prevent any form of cheating, the server’s behavior is fully defined in a Dockerfile, which is shared with all participants.
Timeline
- Challenge Release: June 30, 2025
- Challenge Submission Due: August 31, 2025
- Release Results & Submit Technical Report: September 05, 2025
- Technical Report Due: September 20, 2025
To be eligible for awards, teams are required to submit a technical report of no more than 4 pages. Please note that these reports will not be included in the official ICCV proceedings.
Awards
Winners will be announced at the RealADSim Workshop @ ICCV 2025.
- Innovation Award: $9,000
- Outstanding Champion: $9,000
- Honorable Runner-up: $3,000
How to Participate
To participate in the competition, both automatic registration and manual verification are required:
- Click the "Login with Huggingface" button.
- Click the "Register" button and complete the registration form. After this automatic registration step, the "Submission Information" page will become accessible. It provides detailed instructions on how to run local tests and submit your proposal.
- Access to "My Submissions" and "New Submission" will be granted after we manually review your registration and authorize your account. This process is typically completed within 24 hours.
Citation
If you find our work useful, please kindly cite us via:
@article{zhou2024hugsim,
title={HUGSIM: A Real-Time, Photo-Realistic and Closed-Loop Simulator for Autonomous Driving},
author={Zhou, Hongyu and Lin, Longzhong and Wang, Jiabao and Lu, Yichong and Bai, Dongfeng and Liu, Bingbing and Wang, Yue and Geiger, Andreas and Liao, Yiyi},
journal={arXiv preprint arXiv:2412.01718},
year={2024}
}