STDArm: Transferring Visuomotor Policies From Static Data Training to Dynamic Robot Manipulation
🤖 AI Summary
To address the zero-shot transfer challenge of deploying statically trained visuomotor policies on dynamic mobile platforms (e.g., quadrupedal robots, drones)—exacerbated by platform motion disturbances, sensor-actuator latency, scarcity of high-quality dynamic training data, and constrained edge-computing resources—this paper proposes the first real-time action correction framework. Our method integrates high-frequency action management, a lightweight temporal disturbance compensation network, and an online latency estimation module. Crucially, it enables end-to-end zero-shot transfer from static to dynamic deployment without re-collecting dynamic data or modifying the original policy. We validate the framework across two robotic arms, four mobile platforms, and three representative manipulation tasks. Results demonstrate centimeter-level operational accuracy, real-time suppression of motion-induced disturbances, full preservation of the original policy’s manipulation capabilities, and significantly enhanced robustness and practicality in dynamic environments.
Application Category
📝 Abstract
Recent advances in mobile robotic platforms like quadruped robots and drones have spurred a demand for deploying visuomotor policies in increasingly dynamic environments. However, the collection of high-quality training data, the impact of platform motion and processing delays, and limited onboard computing resources pose significant barriers to existing solutions. In this work, we present STDArm, a system that directly transfers policies trained under static conditions to dynamic platforms without extensive modifications. The core of STDArm is a real-time action correction framework consisting of: (1) an action manager to boost control frequency and maintain temporal consistency, (2) a stabilizer with a lightweight prediction network to compensate for motion disturbances, and (3) an online latency estimation module for calibrating system parameters. In this way, STDArm achieves centimeter-level precision in mobile manipulation tasks. We conduct comprehensive evaluations of the proposed STDArm on two types of robotic arms, four types of mobile platforms, and three tasks. Experimental results indicate that the STDArm enables real-time compensation for platform motion disturbances while preserving the original policy's manipulation capabilities, achieving centimeter-level operational precision during robot motion.
Problem
Research questions and friction points this paper is trying to address.
Transferring static-trained policies to dynamic robot platforms
Compensating for motion disturbances in real-time manipulation
Achieving centimeter-level precision during robot motion
Innovation
Methods, ideas, or system contributions that make the work stand out.
Real-time action correction framework for dynamic transfer
Lightweight prediction network compensates motion disturbances
Online latency estimation calibrates system parameters
🔎 Similar Papers
No similar papers found.
