Toward Real-Time VLAs: Stage-Aware Two-Step Flow Denoising and System-Level Evaluation

2026-09-30

AI Summary

This study addresses the temporal mismatch between the low inference rate of Vision-Language-Action (VLA) models and the high execution frequency required by robots, proposing a distributed real-time VLA framework. Exploiting the phase heterogeneity inherent in Flow Matching-based trajectory denoising, we design a two-stage non-uniform sampling strategy that compresses ten iterative steps into two. This is coupled with end-to-end latency analysis and action provenance logging to enable system-level co-optimization. Experimental results demonstrate that the proposed approach reduces inference latency to 21.96 ms and significantly lowers computational overhead in garment-folding tasks with negligible performance degradation. These findings validate the effectiveness of jointly optimizing inference and control for real-time robotic applications.

Abstract

Vision-language-action (VLA) models face a timing gap between low-rate inference and high-rate robot execution. We characterize this gap through end-to-end latency measurements of model inference and the robot execution chain. Repeated Flow Matching denoising contributes substantially to inference cost, while robot-side delays mainly arise from perception acquisition, communication scheduling, and physical response. Analysis of the velocity field shows relatively stable magnitude and direction in early integration, followed by stronger directional correction near the terminal steps. Based on this stage heterogeneity, we propose two-stage non-uniform denoising, reducing the number of steps from 10 to 2 and model-inference time from 61.557 ms to 21.956 ms. We also develop a distributed real-time VLA framework with independent inference, action-publication, and robot-control rates, modular observation acquisition, and action-provenance logging. Using π0.5 as the baseline, we evaluate six real-time execution methods on a long-horizon physical garment-folding task. Legato performs best overall among training-based methods, while Temporal Smoothing leads among training-free methods; both perform strongly in task success, completion time, action continuity, and acceleration smoothness. Combining two-step denoising with representative execution methods substantially reduces inference cost with a small reduction in task performance. These results motivate joint optimization of model-inference efficiency and robot-system timing.

Problem

• Vision-Language-Action models
• Real-time inference
• Latency gap
• Flow matching denoising
• Robot execution

Innovation

• Vision-Language-Action Models
• Flow Matching Denoising
• Real-Time Inference
• Distributed Framework
• Robotic Execution