Search papers, labs, and topics across Lattice.
This paper introduces AsyncShield, a plug-and-play asynchronous control framework designed to mitigate spatiotemporal misalignment in cloud-based Vision-Language-Action (VLA) navigation caused by network latency. AsyncShield uses a temporal pose buffer and kinematic transformations to convert temporal lag into spatial pose offsets, restoring the VLA's original geometric intent. The system is formulated as a constrained Markov decision process (CMDP) and solved using the PPO-Lagrangian algorithm to balance intent tracking and LiDAR-based obstacle avoidance.
Network jitter in cloud-based robot control can be overcome by converting temporal lag into spatial pose offsets, restoring the VLA's original geometric intent without fine-tuning.
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.