Action-Prior Denoising for Smooth Real-Time Chunking
ROKAE (Shandong) Robot Group Co., Ltd.(ROKAE(山东)机器人集团有限公司);School of Mathematical Sciences, University of Chinese Academy of Sciences(中国科学院大学数学科学学院);The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州))
实时分块(RTC)通过将新生成的动作块条件于前一块已提交的动作,使得分块动作策略能够在推理延迟下运行。训练时RTC在学习过程中模拟这种延迟,避免了部署时昂贵的指导,但其二元前缀掩码将所有非前缀令牌视为完全无约束。这低估了异步执行:早期重叠动作是固定的,而后期重叠动作虽然可编辑,但仍应保持接近先前的计划。我们提出Soft RTC,一种基于动作先验去噪的训练时RTC泛化方法。Soft RTC从部分去噪状态而非纯噪声中构建损坏的重叠令牌,并通过轻量级的令牌级混合规则在推理时将对齐的前一块作为相同先验注入。在12个已发布的大型Kinetix关卡上,短软窗口在整体解决率上几乎与硬训练时RTC相当(0.809 vs. 0.815),而中等窗口相对于硬RTC将高延迟动作变化和急动度分别降低了9.1%和9.6。与推理时RTC基线不同,两种变体都保持近朴素运行时间。一项小型初步真实机器人分拣研究提供了额外证据,表明训练时RTC可以提高完成率,并且Soft RTC在测试策略中给出了最低的命令动作有限差分指标。
英文摘要:
Real-time chunking (RTC) lets chunked action policies operate under inference delay by conditioning a newly generated action chunk on actions already committed by the previous chunk. Training-time RTC simulates this delay during learning and avoids expensive guidance at deployment, but its binary prefix mask treats all non-prefix tokens as fully unconstrained. This under-models asynchronous execution: early overlap actions are fixed, while later overlap actions remain editable but should still stay close to the previous plan. We propose Soft RTC, a training-time RTC generalization based on action-prior denoising. Soft RTC constructs corrupted overlap tokens from partially denoised states instead of pure noise and injects the aligned previous chunk as the same prior during inference through a lightweight token-wise blending rule. On the 12 released large Kinetix levels, a short soft window nearly matches hard training-time RTC in overall solve rate (0.809 vs. 0.815), while a medium window reduces high-delay action delta and jerk by 9.1% and 9.6% relative to hard RTC. Both variants keep near-naive runtime, unlike inference-time RTC baselines. A small preliminary real-robot sorting study provides additional evidence that training-time RTC can improve completion and that Soft RTC gives the lowest commanded-action finite-difference metrics among the tested policies.
Denoising Tells When to Replan: Denoising-Variance Adaptive Chunking for Flow-Based Robot Policies
Beijing Institute of Technology(北京理工大学);The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳));Shenzhen Loop Area Institute(深圳河套学院);Hunan University(湖南大学);Xi’an Jiaotong University(西安交通大学);Renmin University of China(中国人民大学);Harbin Institute of Technology, Shenzhen(哈尔滨工业大学(深圳))
Implicit Action Chunking for Smooth Continuous Control
Department of Data and Systems Engineering, The University of Hong Kong, Hong Kong SAR, China(香港大学数据与系统工程系);Beijing Institute of Technology, Zhuhai, China(北京理工大学珠海学院);College of Computer Science, Sichuan University, Chengdu, China(四川大学计算机学院)
Learning High-Frequency Continuous Action Chunks in Latent Space
School of Computer Science, Shanghai Jiao Tong University, Shanghai, China(上海交通大学计算机科学学院);National University of Singapore, Singapore(新加坡国立大学);Institute of Automation, Chinese Academy of Sciences, Beijing, China(中国科学院自动化研究所);Fudan University, Shanghai, China(复旦大学)
FocalPolicy: Frequency-Optimized Chunking and Locally Anchored Flow Matching for Coherent Visuomotor Policy
State Key Laboratory of Robotics and Intelligent Systems, Shenyang Institute of Automation, Chinese Academy of Sciences(机器人与智能系统国家重点实验室,沈阳自动化研究所,中国科学院);University of the Chinese Academy of Sciences(中国科学院大学);University of Trento(特伦多大学)