AsyncShield: A Plug-and-Play Edge Adapter for Asynchronous Cloud-based VLA Navigation

2026-04-27

AI Summary

This work addresses the spatiotemporal misalignment caused by network latency in cloud-deployed vision-language-action (VLA) models for mobile navigation, which often leads to collisions. The authors propose AsyncShield, a plug-and-play asynchronous control framework that converts time delays into spatial pose offsets via a deterministic physics-based white-box mapping. Instead of relying on black-box temporal prediction, AsyncShield introduces a kinematics-aware pose buffering mechanism. By integrating constrained Markov decision processes (CMDPs), PPO-Lagrangian reinforcement learning, and a collision-radius inflation strategy, AsyncShield dynamically balances semantic intent following with high-frequency LiDAR-based obstacle avoidance—without requiring fine-tuning of the underlying VLA model. Experiments demonstrate that the method significantly improves navigation success rates and safety in both simulation and real-world environments, exhibiting strong zero-shot generalization capabilities.

Technology Category

• Computer Vision: Large Vision Models
• Planning, Routing, and Scheduling: Planning with Language Models
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Agentic search

Abstract

While Vision-Language-Action (VLA) models have been demonstrated possessing strong zero-shot generalization for robot control, their massive parameter sizes typically necessitate cloud-based deployment. However, cloud deployment introduces network jitter and inference latency, which can induce severe spatiotemporal misalignment in mobile navigation under continuous displacement, so that the stale intents expressed in past ego frames may become spatially incorrect in the current frame and lead to collisions. To address this issue, we propose AsyncShield, a plug-and-play asynchronous control framework. AsyncShield discards traditional black-box time-series prediction in favor of a deterministic physical white-box spatial mapping. By maintaining a temporal pose buffer and utilizing kinematic transformations, the system accurately converts temporal lag into spatial pose offsets to restore the VLA's original geometric intent. To balance intent restoration fidelity and physical safety, the edge adaptation is formulated as a constrained Markov decision process (CMDP). Solved via the PPO-Lagrangian algorithm, a reinforcement learning adapter dynamically trades off between tracking the VLA intent and responding to high-frequency LiDAR obstacle avoidance hard constraints. Furthermore, benefiting from a standardized universal sub-goal interface, domain randomization, and perception-level adaptation via Collision Radius Inflation, AsyncShield operates as a lightweight, plug-and-play module. Simulation and real-world experiments demonstrate that, without fine-tuning any cloud-based foundation models, the framework exhibits zero-shot and robust generalization capabilities, effectively improving the success rate and physical safety of asynchronous navigation.

Problem

• Vision-Language-Action
• asynchronous navigation
• spatiotemporal misalignment
• cloud-based deployment
• robot control

Innovation

• AsyncShield
• Vision-Language-Action (VLA)
• spatiotemporal misalignment
• constrained MDP
• edge adaptation