Execution-Aligned Progressive Noise for Consistent Asynchronous Replanning in Generative Robot Policies

Di Wu; Ping Liu; Xuhua Chen; He Zheng; Lingfeng Zhang; Tao Zhang

cs.ROOct 5, 2026

Execution-Aligned Progressive Noise for Consistent Asynchronous Replanning in Generative Robot Policies

Authors: Di Wu, Ping Liu, Xuhua Chen, He Zheng, Lingfeng Zhang, Tao Zhang

Organizations: Magiclab Robotics Technology Co., Ltd., China · Southeast University, Nanjing, China

Abstract

Continuous asynchronous replanning is essential for real-time generative robot policies, but independent stochastic initialization can cause mode switching and inconsistent continuation across action chunks. We propose Execution-Aligned Progressive Noise (EAPN), which introduces structured stochasticity at both inter-chunk and intra-chunk levels. Across replanning steps, EAPN propagates a shared noise trajectory and aligns it with the actual execution displacement, establishing execution-aligned inter-chunk correlation. Within each action chunk, it models temporal correlation along action time. The aligned stochastic history is further combined with committed action context to condition subsequent generation, allowing new chunks to continue from execution-consistent generative states rather than restart from independent noise. We evaluate EAPN on D3IL, Kinetix, LIBERO, and real-world manipulation tasks. EAPN improves multimodal behavior consistency on D3IL and achieves an average success rate of 88.59% on Kinetix. On LIBERO, it remains robust and maintains strong task performance even under long inference delays. Real-robot experiments further achieve 90.0% success on Object Storage and 96.7% on bimanual Cloth Folding, demonstrating reliable continuous execution under asynchronous replanning.

Figures & tables

Fig. 1: Frame-interval inversion of a standard $\pi_{0.5}$ action policy reveals temporally local dependency in the initial noise. The policy is trained with token-wise independent Gaussian noise, and the inversion procedure contains no EAPN structure. (a) Execution-aligned inverted initial noise from 16 held-out validation episodes together with an independent Gaussian reference; the same-episode 5-NN hit rate in the raw space is 100%. (b) Mean cosine similarity is 0.774 within the same episode, 0.043 across different episodes, and 0.000 for the independent Gaussian reference. (c) As the observation interval increases from 5 to 45 frames, the within-episode similarity decreases continuously from 0.855 to 0.378. Thick curves denote the mean over 16 episodes and shaded regions indicate 95% bootstrap confidence intervals. (d) Ten real observations from frames 20–65 of Episode B are matched one-to-one with their inverted latent points.

Fig. 3: Execution-aligned latent structure in a standard $\pi_{0.5}$ policy. Latent correlation matrices between neighboring Action Chunks for actual execution displacements $D=1,5,9$ . The black dashed line indicates the execution-aligned position $j=h+D$ determined by the actual execution displacement.

Method	Success $\uparrow$	MSR $\downarrow$	Switch Gap $\downarrow$
DDPM-ACT	73.13% (351/480)	3.969%	0.001913
EAPN	78.96% (379/480)	3.167%	0.001672

TABLE I: Task success and behavioral consistency under consecutive replanning on D3IL Avoiding.

Fig. 4: Comparison under asynchronous inference on Kinetix. Left: average success rate over 12 tasks for different execution horizons $K$ with fixed $D=1$ , and for different inference delays $D$ . Right: per-task results.

Variant	Success $\uparrow$	Switch Gap $\downarrow$	Tail Gap $\downarrow$
ttRTC	46.60	1.58	1.58
+ Inter-Chunk Noise	76.89	1.55	1.51
+ Intra-Chunk Noise	78.85	1.54	1.50
EAPN	79.26	1.55	1.50

TABLE II: Component ablation of EAPN on Kinetix mjc_swimmer .

Panel of Figure 7 — Fig. 5: Real-world experimental setups. (a) Object Storage; (b) Cloth Folding

Fig. 6: Representative asynchronous execution sequences for Cloth Folding. The baseline exhibits repeated corrective motions during execution, whereas EAPN maintains more continuous task progression in this example.

Method	$D$	Spatial	Object	Goal	Long	Avg. SR
VLASH	1	94.2	98.6	96.4	94.8	96.0
	2	93.0	99.2	94.6	94.0	95.2
	3	92.2	99.4	94.8	92.6	94.8
	4	86.0	97.4	92.6	90.8	91.7
ttRTC	1	94.6	99.0	94.8	95.4	96.0
	2	92.4	98.4	95.0	93.2	94.8

TABLE III: Success rate (%) on LIBERO under low inference delay.

Method	$D$	Spatial	Object	Goal	Long	Avg. SR
VLASH	5	69.0	71.0	79.0	67.0	71.5
	10	68.0	69.0	80.0	64.0	70.3
	15	70.0	72.0	78.0	53.0	68.3
ttRTC	5	87.0	96.0	90.0	86.0	89.8
	10	76.0	84.0	91.0	70.0	80.3
	15	70.0	80.0	86.0	56.0	73.0

TABLE IV: Success rate (%) on LIBERO under high inference delay.

Method	Success Rate $\uparrow$
	Object Storage	Cloth Folding
Naive Async	65.0% (13/20)	63.3% (19/30)
RTC	80.0% (16/20)	60.0% (18/30)
ttRTC	80.0% (16/20)	63.3% (19/30)
VLASH	85.0% (17/20)	93.3% (28/30)
EAPN	90.0% (18/20)	96.7% (29/30)

TABLE V: Comparison on real-world experiments.

Fig. 7: Motion-profile comparison on real-robot execution segments. Top: velocity. Bottom: acceleration. Curves show RTC, ttRTC, VLASH, and EAPN over 0–5 s.

Explore similar work

Start Right, Arrive Right: Asynchronous Execution via Initial Noise Selection
Jun 18, 2026Trong-Bao Ho, Quang-Tan Nguyen, Thien-Loc Ha +7Action Chunks Robot Policies
VinRobotics · VinUniversity · DFKI +2
PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking
May 30, 2026Junnan Nie, Jiayi Li, Chenghao Liu +5Action Chunks Robot Policies
Peking University · Peking University. · JD Explore Academy. +1
Dynamic Execution Horizon Prediction for Chunk-based Robot Policies
Jun 9, 2026Yuchi Zhao, Miroslav Bogdanovic, Arjun Sohal +5Action Chunks Long-Horizon Manipulation
Department of Computer Science, University of Toronto, 40 St George St., Toronto, ON M5S 2E4, Canada · Vector Institute for Artificial Intelligence, W1140-108 College St., Schwartz Reisman Innovation Campus, Toronto, ON M5G 0C6, Canada · Acceleration Consortium, 700 University Ave., Toronto, ON M7A 2S4, Canada +7

Jun 18, 2026, cs.ROJ/K move · Enter open · S save

Start Right, Arrive Right: Asynchronous Execution via Initial Noise Selection

Trong-Bao Ho, Quang-Tan Nguyen, Thien-Loc Ha, Gia-Binh Nguyen+6

VinRobotics · VinUniversity · DFKI+2

Action Chunks Robot Policies Action Generation Action Sequences Asynchronous Execution Flow Policies+2

Open paper AskPDF arXiv

Action chunking enables robot policies to produce temporally coherent behavior, but generating multi-step action sequences with flow-based policies incurs latency that is incompatible with real-time control. Under asynchronous execution, the robot continues executing the current chunk while the next one is generated, causing even minor delays to create inconsistencies at chunk boundaries. Existing methods address this problem by steering generation toward the already executed action prefix. We instead show that prefix consistency can be achieved by selecting an appropriate initial noise before generation begins, allowing the unmodified flow ODE to produce a coherent next chunk. This reframes asynchronous inference as a noise selection problem rather than a trajectory steering problem. We introduce \textbf{PAINT}, a training-free method that finds this noise via backward Euler inversion and constructs the final chunk through a repainting rule. In summary, \texttt{PAINT} requires no gradients, retraining, or policy modification; yet it improves execution consistency and task performance across \textit{12 simulated benchmarks} and \textit{6 real-world manipulation tasks} spanning single-arm, bimanual, and humanoid embodiments. Website: ~\href{https://paint-action-chunking.github.io}{\texttt{https://paint-action-chunking.github.io}}.

Abstract

Explore similar work

Start Right, Arrive Right: Asynchronous Execution via Initial Noise Selection

PACE: Phase-Aware Chunk Execution for Robot Policies with Action Chunking

Dynamic Execution Horizon Prediction for Chunk-based Robot Policies