ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Dai, Yuntao, Gu, Hang, Wang, Teng, Cheng, Qianyu, Zheng, Yifei, Qiu, Zhiyong, Gong, Lei, Lou, Wenqi, Zhou, Xuehai
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909974645440512
author Dai, Yuntao
Gu, Hang
Wang, Teng
Cheng, Qianyu
Zheng, Yifei
Qiu, Zhiyong
Gong, Lei
Lou, Wenqi
Zhou, Xuehai
author_facet Dai, Yuntao
Gu, Hang
Wang, Teng
Cheng, Qianyu
Zheng, Yifei
Qiu, Zhiyong
Gong, Lei
Lou, Wenqi
Zhou, Xuehai
contents Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. However, their deployment in dynamic, real-world environments is severely hin dered by high inference latency. While smooth robotic interaction requires control frequencies of 20 to 30 Hz, current VLA models typi cally operate at only 3-5 Hz on edge devices due to the memory bound nature of autoregressive decoding. Existing optimizations often require extensive retraining or compromise model accuracy. To bridge this gap, we introduce ActionFlow, a system-level inference framework tailored for resource-constrained edge plat forms. At the core of ActionFlow is a Cross-Request Pipelin ing strategy, a novel scheduler that redefines VLA inference as a macro-pipeline of micro-requests. The strategy intelligently batches memory-bound Decode phases with compute-bound Prefill phases across continuous time steps to maximize hardware utilization. Furthermore, to support this scheduling, we propose a Cross Request State Packed Forward operator and a Unified KV Ring Buffer, which fuse fragmented memory operations into efficient dense computations. Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dy namic manipulation on edge hardware. Our work is available at https://anonymous.4open.science/r/ActionFlow-1D47.
format Preprint
id arxiv_https___arxiv_org_abs_2512_20276
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge
Dai, Yuntao
Gu, Hang
Wang, Teng
Cheng, Qianyu
Zheng, Yifei
Qiu, Zhiyong
Gong, Lei
Lou, Wenqi
Zhou, Xuehai
Artificial Intelligence
Robotics
Vision-Language-Action (VLA) models have emerged as a unified paradigm for robotic perception and control, enabling emergent generalization and long-horizon task execution. However, their deployment in dynamic, real-world environments is severely hin dered by high inference latency. While smooth robotic interaction requires control frequencies of 20 to 30 Hz, current VLA models typi cally operate at only 3-5 Hz on edge devices due to the memory bound nature of autoregressive decoding. Existing optimizations often require extensive retraining or compromise model accuracy. To bridge this gap, we introduce ActionFlow, a system-level inference framework tailored for resource-constrained edge plat forms. At the core of ActionFlow is a Cross-Request Pipelin ing strategy, a novel scheduler that redefines VLA inference as a macro-pipeline of micro-requests. The strategy intelligently batches memory-bound Decode phases with compute-bound Prefill phases across continuous time steps to maximize hardware utilization. Furthermore, to support this scheduling, we propose a Cross Request State Packed Forward operator and a Unified KV Ring Buffer, which fuse fragmented memory operations into efficient dense computations. Experimental results demonstrate that ActionFlow achieves a 2.55x improvement in FPS on the OpenVLA-7B model without retraining, enabling real-time dy namic manipulation on edge hardware. Our work is available at https://anonymous.4open.science/r/ActionFlow-1D47.
title ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge
topic Artificial Intelligence
Robotics
url https://arxiv.org/abs/2512.20276