arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.02271cs.PFcs.AIcs.ARcs.RO
收藏

表征VLA模型:识别边缘AI架构中的动作生成瓶颈

Characterizing VLA Models: Identifying the Action Generation Bottleneck for Edge AI Architectures

Google, Mountain View, CA(谷歌,山景城,加利福尼亚);Purdue University, West Lafayette, IN(普渡大学,沃斯堡,印第安纳)

Manoj Vishwanathan, Suvinay Subramanian, Anand Raghunathan

AI总结:

本文研究了VLA模型在边缘AI中的动作生成瓶颈,通过分析发现动作生成阶段占端到端延迟的75%,并探讨了扩展至100B参数模型的硬件需求及未来技术路径。

摘要:

视觉-语言-动作(VLA)模型是一类新兴的工作负载,对机器人和具身AI在边缘端至关重要。随着这些模型的扩展,它们展示了显著的能力提升,但必须在本地部署以满足实时应用的严格延迟要求。本文对两个代的边缘硬件进行了VLA性能分析,即Nvidia Jetson Orin和Thor平台。使用MolmoAct-7B这一最先进的VLA模型,我们识别出一个主要的执行瓶颈:端到端延迟的75%消耗在内存受限的动作生成阶段。通过分析建模和模拟,我们预测了扩展到100B参数模型所需的硬件需求。我们还探讨了高带宽内存技术和处理内存(PIM)在具身AI边缘系统中的潜在影响。

英文摘要:

Vision-Language-Action (VLA) models are an emerging class of workloads critical for robotics and embodied AI at the edge. As these models scale, they demonstrate significant capability gains, yet they must be deployed locally to meet the strict latency requirements of real-time applications. This paper characterizes VLA performance on two generations of edge hardware, viz. the Nvidia Jetson Orin and Thor platforms. Using MolmoAct-7B, a state-of-the-art VLA model, we identify a primary execution bottleneck: up to 75% of end-to-end latency is consumed by the memory-bound action-generation phase. Through analytical modeling and simulations, we project the hardware requirements for scaling to 100B parameter models. We also explore the impact of high-bandwidth memory technologies and processing-in-memory (PIM) as promising future pathways in edge systems for embodied AI.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑