FutureRTC: Real-Time Robot Execution with Anticipatory-Conditioned Action Chunking
2026-07-27 • Robotics
Robotics
AI summaryⓘ
The authors address a problem in robot control where there is a delay between deciding an action and actually doing it, which causes jerky movements. They propose FutureRTC, a method that predicts what the robot will see and feel in the future to better align actions with real execution, without changing the robot's original control policy. Their approach uses predicted robot states and visual information to smooth out control and improve performance. Experiments show that this method leads to smoother motions, faster task completion, and higher success rates in both simulated and real robots.
Vision-Language-Action (VLA)asynchronous executionrobot controlproprioceptive statevisual observation predictionmotion-aware feature transportpolicy consistency lossinference delaytrajectory smoothingrobot motion prediction
Authors
Hai Jiang, Yixian Zou, Binbin Liang, Boqian Liu, Fanman Meng, Shuaicheng Liu
Abstract
Real-time deployment of Vision-Language-Action (VLA) policies necessitates asynchronous execution, wherein subsequent action chunks are computed concurrently with the execution of the current chunk, leading to prediction-execution misalignment and manifesting as inter-chunk discontinuities. Existing methods either superficially smooth chunk boundaries, require costly policy optimization, or exclusively forward-predict proprioceptive states yet neglect critical visual observations. In this paper, we propose \textbf{FutureRTC}, a plug-and-play adaptation framework that predicts execution-time observations and states for asynchronous VLA control without modifying the underlying policy. Specifically, FutureRTC features a state correction module to compensate for the discrepancy between rolled-forward and actual execution-time proprioceptive states and an observation prediction module that forecasts execution-time visual representations by leveraging robot motion as an explicit physical prior through motion-aware feature transport and reconstruction. Furthermore, we introduce a policy consistency loss to align the action chunks generated from predicted contexts with those produced under the expected execution-time inputs of the VLA policy. Extensive experiments across simulated and real-world environments demonstrate that FutureRTC achieves superior robustness to inference delays, resulting in smoother trajectories, faster execution, and consistently higher task success rates.