World Action Models in Real Time: An Empirical Study of Smooth Execution via Asynchronous Deployment
2026-08-03 • Robotics
Robotics
AI summaryⓘ
The authors studied ways to improve how robots perform actions when using World Action Models, which can be slow and cause jerky or delayed movements. They tested six different methods to overlap thinking (model inference) and doing (action execution) on a bimanual robot. They found that making sure the robot’s observations, predictions, and actions match up in time is very important to avoid sudden jumps. Among the methods tested, prefix-conditioned generation worked best by learning smooth action sequences during training. Their work helps guide how to run these slow action models smoothly in real-world robots.
World Action Modelsaction inferenceasynchronous executionrobot controltrajectory smoothingdenoisingprefix-conditioned generationbimanual robottemporal alignmentreal-time robotics
Authors
Motubrain Team
Abstract
World Action Models generate fixed-horizon action chunks through iterative denoising, creating substantial inference latency that can cause pauses, stale actions, and discontinuities during robotic execution. We present an empirical study of asynchronous deployment strategies that overlap model inference with action execution to enable responsive and smooth control. We compare six strategies, including synchronous execution, pure asynchronous switching, post-hoc action blending, denoising-time blending, inference-time velocity guidance, and prefix-conditioned generation, on a 10 Hz bimanual robot. Evaluation combines offline trajectory analysis with online experiments across dynamic manipulation, precision-critical placement, and long-horizon tasks. Our results identify accurate temporal alignment between observations, predictions, and executed commands as a fundamental requirement. Alignment errors produce persistent chunk-boundary discontinuities that cannot be corrected through blending alone. With proper alignment, direct action weighting provides a simple and smooth baseline but sacrifices accuracy in precision-critical tasks. Inference-time velocity guidance fails to reliably constrain committed actions on our platform. In contrast, prefix-conditioned generation achieves the best overall balance between task performance, execution speed, and trajectory smoothness by learning consistent action continuations during training. These findings clarify the practical trade-offs among asynchronous deployment strategies and provide guidance for deploying high-latency World Action Models in real-time robotic systems.