arXiv cs.AI· Jonas Eschmann, Giuseppe Loianno·· 13 小时前AI 评分38
HyperDrone：每秒数百万帧的仿真到现实迁移四旋翼模拟器
HyperDrone: Simulation-to-Reality Transfer at Millions of Frames per Second
AI 导读
HyperDrone 是一款四旋翼模拟器，批量渲染速度比现有 SOTA 批量四旋翼模拟器快 20 倍以上，其 HyperLux 光线追踪引擎支持 Nvidia OptiX、Vulkan、Apple Metal 和 WebGPU 后端，在消费级 GPU 上最高保真设置下仍可达数百万 FPS。
正文 · 原文
Abstract:We present HyperDrone, an extremely fast quadrotor simulator focused on high-performance batch rendering that is more than 20x faster than existing state-of-the-art batch quadrotor simulators. HyperDrone combines highly optimized multirotor dynamics with HyperLux, a new ray-tracing engine that supports Nvidia OptiX, Vulkan, Apple Metal and WebGPU backends that takes full advantage of modern ray-tracing hardware, capable of running both components entirely on the GPU. HyperLux renders RGB, depth, semantic/instance segmentation, surface normals and optical flow observations at configurable levels of fidelity, including reflections, realistic lighting, motion blur, anti-aliasing, and Physically Based Rendering (PBR). Even in the highest-fidelity settings, HyperLux achieves millions of Frames Per Second (FPS) on consumer-grade GPUs while rendering realistic 3D scenes. We demonstrate the merits of HyperDrone by training a vision-based yaw-control policy purely in simulation and transferring it to a real small-scale quadrotor (Crazyflie) with a camera, running completely onboard.
| Comments: | Accepted at Conference on Robot Learning (CoRL 2026) |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Systems and Control (eess.SY) |
| Cite as: | arXiv:2610.28950 [cs.RO] |
| (or arXiv:2610.28950v1 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2610.28950 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jonas Eschmann [view email]
[v1]
Tue, 20 Oct 2026 22:03:11 UTC (5,542 KB)
来源：arXiv cs.AI · arxiv.org