[论文] World Action Models Enable Continual Imitation Learning with Recurrent Generative Replays

论文概要

研究领域: 计算机视觉、机器人学
作者: Manish Kumar Govind, Dominick Reilly, Smit Patel
发布时间: 2026-06-27
arXiv: 2606.27374

中文摘要

超越预测机器人动作，世界动作模型（WAMs）还能生成未来的视觉观察。我们基于这种生成能力，提出递归生成回放（REGEN），一种持续模仿学习框架，合成伪回放轨迹，使机器人策略能够排练先前学习的任务，而无需存储其原始人类演示。在持续适应过程中，REGEN递归查询WAM，合成仅以先前任务指令和当前任务观察为条件的伪回放轨迹。在模拟和真实世界操作设置中的实验表明，相对于顺序微调，REGEN将灾难性遗忘降低了高达50%，同时接近需要访问真实回放数据的特权经验回放方法的性能。我们分析了限制生成回放的因素，识别出长程视觉退化和动作-观察不一致性是主要瓶颈。我们的结果确立了WAMs作为无需存储演示的持续机器人学习的有前景基础。

原文摘要

Going beyond predicting robot actions, World Action Models (WAMs) can also generate future visual observations. We build on this generative capability to propose Recurrent Generative Replay (REGEN), a continual imitation learning framework that synthesizes pseudo-replay trajectories, enabling a robot policy to rehearse previously learned tasks without storing their original human demonstrations. During continual adaptation, REGEN recursively queries the WAM to synthesize pseudo-replay trajectories conditioned only on prior task instructions and current-task observations. Experiments in both simulation and real-world manipulation settings show that REGEN reduces catastrophic forgetting by up to 50% relative to sequential fine-tuning, while approaching the performance of privileged...

自动采集于 2026-06-27

#论文 #arXiv #cs.RO #cs.CV #小凯