论文arXiv 2611.04096今天

实时HJ可达性GPU求解器实现机载安全验证，RTX4000达112Hz

提出面向机器人控制环的CUDA版HJ可达性求解器，在RTX 4000 Ada上3D BRT计算达112Hz，较JAX基准快5.9-17.5倍；Jetson AGX Thor端侧运行66Hz，为黑盒策略提供确定性安全层。

AI 整理

端侧部署
评测与安全

本文由 AI 根据公开内容整理，可能有误，请以原文为准：Online Hamilton-Jacobi Reachability Onboard the Robot Control Loop: A Real-Time GPU Solver for Safety-Critical Autonomy（arXiv 2611.04096）

核心内容

针对具身智能黑盒策略缺乏形式化安全保障的痛点，研究团队发布基于WENO5+TVD-RK3+Lax-Friedrichs格式的Hamilton-Jacobi可达性CUDA求解器。该方案将原本离线的BRT计算迁移至机载控制环内，在NVIDIA RTX 4000 Ada工作站上以FP32精度实现3D Air3D BRT 112Hz、4D追逃博弈61Hz及2D风场漂移74-147Hz的实时解算，相比同精度JAX参考实现提速5.9-17.5倍，确立了在线安全验证的工程可行性。

该系统在嵌入式端与多智能体场景同样验证了高效性：NVIDIA Jetson AGX Thor以FP32精度运行Air3D达66Hz（N=41），较hj_reachability快3.8-8.9倍；FP64精度下与基准误差仅0.04%且提速2.0-4.6倍。通过内核级标签传播扩展，单次HJ求解即可恢复K智能体捕获域Voronoi剖分，在6架四旋翼野火响应闭环仿真中，时变非均匀风场下每帧耗时仅4.0ms，且计算开销独立于集群规模。

关键看点

工作站性能：RTX 4000 Ada FP32下3D BRT达112Hz，较JAX基准提速5.9-17.5倍
端侧部署：Jetson AGX Thor运行Air3D达66Hz，较hj_reachability快3.8-8.9倍
数值精度：FP64模式复现Mitchell Air3D基准误差仅0.04%，速度提升2.0-4.6倍
多机扩展：内核级标签传播实现K智能体Voronoi剖分，6机仿真单帧仅4.0ms

工程与产业观察

该工作突破了HJ可达性分析因维度灾难无法在线应用的瓶颈，证明了消费级与嵌入式GPU足以支撑3-4维动力学系统的实时安全护盾。对于依赖VLA模型的人形机器人与无人机集群，该求解器提供了不依赖概率置信度的确定性安全兜底，是构建可认证具身智能系统的关键底层算法组件。

英文摘要 (Abstract)

Hamilton-Jacobi (HJ) reachability provides certified backward reachable tubes (BRTs) for safety-critical robotic systems, but the level-set computation scales exponentially in the state dimension and has historically been performed offline. As autonomy stacks (ADAS/AV, humanoid robotics, and vision-language-action models) increasingly rely on black-box, probabilistic policies, a deterministic safety layer with formal guarantees is an essential complement. We present a CUDA implementation of the canonical WENO5 + TVD-RK3 + Lax-Friedrichs level-set Hamilton-Jacobi-Isaacs solver for 2D, 3D, and 4D dynamics that is fast enough to run inside a robot's control loop. In 32-bit floating point (FP32) on a single workstation GPU (NVIDIA RTX 4000 Ada), it computes the 3D Air3D BRT at 112 Hz, a 4D pursuit-evasion BRT at 61 Hz, and a 2D wind-affected drifter BRT at 74-147 Hz, 5.9-17.5x faster than the JAX reference hj_reachability at matched precision and 1.7-16x faster than a torch.compile reimplementation of the same scheme. In 64-bit floating point (FP64) it agrees with hj_reachability on eight benchmark cases, reproduces Mitchell's Air3D BRT to within 0.04% of the reference cell count, and runs 2.0-4.6x faster. On an embedded NVIDIA Jetson AGX Thor it computes Air3D at 66 Hz (N=41) and the 2D drifter at 54 Hz, 3.8-8.9x faster than hj_reachability on the same device. As a multi-agent application, an in-kernel label-propagation extension recovers the K-agent capture-basin Voronoi tessellation from a single combined HJ solve, independent of fleet size; it runs at 4.0 ms per tick in a closed-loop wildfire-response simulation of six quadcopters under time-varying, non-uniform wind.

作者：Kevin Weekly, Alexandre M. Bayen