Research map / 3D vision / Face and shape
AeroAgent: A Vision-Physics-Decision Framework for Aerodynamic Vehicle Design
Ye Liu, Shouyi Liu, Huiyu Yang, Jianghang Gu, Wenhao Fan, Zhongxin Yang, Ding Wang, Simeng Chen, Zirun Jiang, Yuanwei Bin, Shiyi Chen, Yuntian Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use AeroFormer surrogate instead of next-best baseline | Aerodynamic surrogate · Vehicle aero surrogate, pressure field | relative L2: 0.1797 -> 0.1072 | Improved |
| use AeroFormer surrogate instead of next-best baseline | Aerodynamic surrogate · Vehicle aero surrogate, velocity field | relative L2: 0.0521 -> 0.0279 | Improved |
| use AeroFormer surrogate instead of 3D-GeoCA | Aerodynamic surrogate · Vehicle aero surrogate, Cd | R2: 0.9310 -> 0.9484 | Improved |
| run five propose-evaluate-refine iterations with the surrogate on a SUV | AeroAgent · Five-step refinement SUV | Cd: 0.401 -> 0.351 | Improved |
| run five propose-evaluate-refine iterations on a sports car | AeroAgent · Five-step refinement sports car | Cd: 0.339 -> 0.261 | Improved |
| run five iterations with surrogate-guided edits on production-style vehicles | AeroAgent · Production-style vehicles, five classes | drag: 2-12% lower drag | Improved |
| use surrogate in inner loop and CFD only for final top-K | AeroAgent · Production-style vehicles | high-fidelity CFD calls: 50-80% fewer | Improved |
| use linear-attention mesh-free AeroFormer instead of Transolver | Aerodynamic surrogate · Training at 5e6 points | GPU memory (GB): 70.2 -> 54.3 | Improved |
| use linear-attention AeroFormer instead of Transolver | Aerodynamic surrogate · Inference at 5e6 points | GPU memory (GB): 9.3 -> 7.5 | Improved |
| apply planner edits progressively on a single coupe | AeroAgent · Single coupe, five shapes | Cd | Improved |
| AeroAgent edits with GPT-5 aesthetic predicate | AeroAgent · Production-style vehicles | aesthetic satisfaction E(x;U) | No clear effect |
Closest papers by meaning
- 3DID: Direct 3D Inverse Design for Aerodynamics with Physics-Aware Optimization (NeurIPS 2025 · similarity 0.88)
- DrivAerStar: An Industrial-Grade CFD Dataset for Vehicle Aerodynamic Optimization (NeurIPS 2025 · similarity 0.88)
- AeroGTO: An Efficient Graph-Transformer Operator for Learning Large-Scale Aerodynamics of 3D Vehicle Geometries (AAAI 2025 · similarity 0.87)
- DrivAerNet++: A Large-Scale Multimodal Car Dataset with Computational Fluid Dynamics Simulations and Deep Learning Benchmarks (NeurIPS 2024 · similarity 0.87)
- DragSolver: A Multi-Scale Transformer for Real-World Automotive Drag Coefficient Estimation (ICML 2025 · similarity 0.87)
- AutoHood3D: A Multi‑Modal Benchmark for Automotive Hood Design and Fluid–Structure Interaction (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
