arXiv：cs.AI· Daniele Affinita, Mingda Xu, Beno\^it Valentin Gherardi, Pascal Fua·· 5 小时前
神经代理模型协同设计气动外形与控制
Neural-Surrogate Co-Design of Aerodynamic Shape and Control
AI 导读
研究者提出一种基于梯度的双层公式，用可微神经代理模型替换原有动力学模型，并通过结构利用方法在轨迹优化器中求导，框架在固定翼滑翔机上以 NeuralFoil 作为气动代理验证。在 2D 翼型和 3D 机翼的最大距离滑翔、最短距离着陆与栖停任务上，该方法比最先进的进化基线快约五倍，并优于仅优化气动代理指标的流程。
正文
Abstract:Control co-design jointly optimizes a robot's physical design and its controller. Existing techniques are limited in the number of model parameters and the complexity of the physical models they can handle. In this paper, we propose a gradient-based bilevel formulation that replaces the dynamics models of earlier techniques with a differentiable neural surrogate and differentiates through the trajectory optimizer using a structure-exploiting approach. We validate the framework on fixed-wing gliders, using NeuralFoil as the aerodynamic surrogate. The tasks are maximum-distance gliding, minimum-distance landing, and perching. Experiments on 2D airfoils and 3D wings show that our method runs about five times faster than a state-of-the-art evolutionary baseline. It also outperforms pipelines that optimize aerodynamic proxy metrics.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2603.06760 [cs.RO] |
| (or arXiv:2603.06760v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2603.06760 arXiv-issued DOI via DataCite |
Submission history
From: Daniele Affinita [view email]
[v1]
Fri, 6 Mar 2026 14:18:18 UTC (3,130 KB)
[v2]
Tue, 13 Oct 2026 11:52:36 UTC (1,447 KB)
来源：arXiv：cs.AI · arxiv.org