ChronoFlow-Policy: Unifying Past-Current-Future Interaction Flow in Visuomotor Policy Learning

2026-06-30

AI Summary

Existing visuomotor policies struggle to jointly model the dynamic interactions spanning past, present, and future states, limiting their robustness in non-Markovian, long-horizon manipulation tasks. This work proposes ChronoFlow-Policy, which introduces, for the first time, a unified temporal representation called ChronoFlow. It captures the Past-Current-Future interaction dynamics between objects and the gripper through sparse 3D keypoints and leverages a diffusion model to learn this representation end-to-end alongside action sequences. Evaluated across 14 simulated and 5 real-world manipulation tasks, the method significantly outperforms strong baselines, demonstrating enhanced policy performance and robustness in long-horizon, non-Markovian scenarios.

Abstract

Visual signals play a crucial role in policy learning by enabling models to capture object motion and interaction dynamics. Just as humans reason about actions using both past experience and anticipated outcomes, effective policies should integrate past interactions with future predictions. However, existing visuomotor policies typically model either historical context or future dynamics in isolation, lacking a unified temporal representation of interaction dynamics. In this work, we introduce \textbf{ChronoFlow}, a temporally unified representation that captures \textbf{past, current, and future} interaction dynamics through sparse 3D keypoints of both objects and the gripper. Based on this representation, we propose \textbf{ChronoFlow-Policy}, a diffusion-based visuomotor policy that jointly learns ChronoFlow and action sequences through a co-training objective. Experiments on 14 simulated tasks and 5 real-world manipulation tasks demonstrate that ChronoFlow-Policy consistently outperforms strong diffusion-policy baselines and improves robustness in long-horizon and non-Markovian manipulation scenarios.

Problem

• visuomotor policy
• temporal representation
• interaction dynamics
• past-current-future integration
• non-Markovian manipulation

Innovation

• ChronoFlow
• temporal unification
• visuomotor policy
• diffusion policy
• 3D keypoints