Asynchronous Robot Policies

Sombit Dey; Luc Van Gool; Danda Pani Paudel

cs.RONov 3, 2026

Asynchronous Robot Policies

Authors: Sombit Dey, Luc Van Gool, Danda Pani Paudel

Organizations: INSAIT, Sofia Univeristy St. Kliment Ohridski

Abstract

Scaling vision-language foundation models has significantly advanced generalist robot manipulation, but the increased model scale introduces computational latency that degrades real-time physical reactivity. Existing Vision-Language-Action (VLA) architectures execute high-level semantic reasoning and action ex- ecution sequentially within the primary control loop, creating a fundamental tension between semantic-reasoning capability and robot agility. To address this, we propose Asynchronous Robot Policy (ARP), an architectural framework that decouples semantic reasoning from low-level action generation using asynchronous processes operating at independent frequencies. In ARP, a high-capacity Vision-Language Model (VLM) asynchronously updates a semantic memory at low fre- quency, while a lightweight Vision-Action Expert, conditioned on this semantic memory, generates robot actions at a much higher control frequency. Such a formulation, enables having a fast perception-action loop without compromising on the semantic capability. To overcome asynchronous memory staleness, ARP trains the vision-action expert on the semantic staleness and execution delay, with temporally mismatched observation across processes. On Franka manipulation tasks, ARP achieves an 18% higher success rate and 20% less time need for execution in quasi-static setups, while 3-5 times baseline performance in dynamic simulation environments. Furthermore, ARP enables operating the VLM at 10 times lower than its nominal frequency with no significant performance loss making it much more efficient than the baseline. These results highlight the architectural advantages of modeling robot policies as asynchronous processes, enabling real-world reactive deployment without sacrificing the scale of foundation models.

Figures & tables

Panel of Figure 1 — Figure 1: a) Architectural and formulation differences between Asynchronous Robot Policies (ARP) and the standard Vision-Language-Action (VLA) paradigm. b) Performance comparison between ARP and standard VLAs across two key axes (both, higher the better): success rate and reactivity .

Figure 2: Architecture design behind ARP. The semantic expert populates the memory with the latest available observation, whereas the vision-action expert cross attends to the stale-memory. The adaptive normalization is influenced by both the flow step

\tau

as well as staleness.

	Spatial	Object	Goal	Long	Average
	SR (%)	SR (%)	SR (%)	SR (%)	SR (%)
Diffusion Policy ( Chi et al., 2023 )	78.3	92.5	68.3	50.5	72.4
$\pi_{0}$ ( Black et al., 2024 )	96.8	98.8	95.8	85.2	94.2
$\pi_{0}$ -1b-replicated	83.2	95.2	90.6	51.8	80.2
$\pi_{0}$ -8b-replicated	85.4	95.0	90.0	60.5	82.7
ARP-1B	89.4	97.6	89.8	62.4	84.8

Table 1: LIBERO simulation benchmark results. Success rates (SR) across four LIBERO benchmark task suites ( Liu et al., 2023 ) . Row shading indicates execution settings: grey denotes the model are pretrained on large robot dataset.

Figure 3: Franka Evaluation Results on Real-world experiments and RoboLab simulation.

Table 5

Figure 4: Qualitative example of Dynamic-Bump task. After bumping the ball, ARP is able to react in time to grasp it whereas the $\pi_{0}$ model fails.

Figure 5: A qualitative example from the LIBERO_Perturb benchmark, target displaced by 5cm.

Figure 6: Robustness to VLM slowdown. Success rates on LIBERO are reported as VLM inference is limited by a compute budget. Our asynchronous 1B policy remains robust across varying VLM compute budget compared to a

\pi_{0}

-style baseline, indicating bigger backbones can be deployed in the proposed way.

Training staleness	Sync	5 Hz	2.5 Hz	1 Hz	0.5 Hz
$d_{\mathrm{t}}\equiv 0$	81.8	63.0	23.5	4.8	2.0
$\mathcal{U}\{1,\ldots,8\}$	79.5	81.5	56.8	11.8	5.0
$\mathcal{U}\{1,\ldots,40\}$ (default)	85.2	83.2	84.8	82.0	81.8
$\mathcal{U}\{1,\ldots,80\}$	81.0	84.5	87.5	87.5	86.5

Table 4: Effect of cache-age distribution. Success rate (%) reported for the 1B model across 500 episodes per cell. Broader staleness bands substantially improve policy robustness under slow VLM refresh rates.

Policy	SR%	Time (s)	EE ISJ	EE SPARC
ARP-DROID	31.9	19.5	456	-5.11
ARP-DROID-RTC	37.5	17.1	296	-3.90
$\Delta$	+5.6	-15.3%	-39.4%	+24.9%

Table 5: Asynchronous vs. synchronous inference for the ARP-3B checkpoint on DROID RoboLab tasks. ISJ is integrated squared jerk (lower is smoother); SPARC is spectral arc length (closer to zero is smoother). Metrics from RoboLab ( Yang et al., 2026 ) , EE indicated end-effector.

Appendix figures & tables2 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 7: Task success rate of ARP-1B compared to the

\pi_{0}

-1B baseline across varying VLM inference frequencies. ARP-1B maintains consistent performance despite lower update rates, outperforming the baseline by up to +18.2% at 0.5 Hz.

Policy	Pretrain	Simple			Moderate		Complex		Avg
	Dataset	SR%	Score	Time (s)	SR%	Score	SR%	Score	SR%
$\pi_{0.5}$	Proprietary	42.4%	36.6	16.7	37.5%	44.0	22.1%	31.5	37.9%
LAP-3B	OpenX	20.9%	20.2	20.0	18.6%	26.4	8.8%	20.7	18.4%
ARP-3B	–	25.8%	24.5	14.7	16.3%	23.1	6.6%	16.1	20.0%

Appendix

Table 6: Robolab Simulation results on 120 tasks and total 960 episodes per model.

VLMs for Robotics

Vision-language models for robotics connect visual and linguistic understanding to grounded navigation, manipulation, and robot control.

See all 683 papers

Jun 13, 2026

Acting While Understanding: Asynchronous Semantic-Action Decoupling for Real-Time Vision-Language-Action Models

Vision-Language-Action models (VLAs) have demonstrated strong task understanding and generalization in robotic manipulation, yet the high computational cost of full-model inference limits their deployment in low-latency, high-frequency closed-loop control. We propose an asynchronous semantic-action decoupling framework that separates semantic understanding from action generation along the internal semantic-action interface of existing VLAs, without redesigning the vision-language backbone or introducing an external planner. A low-frequency understanding module asynchronously updates reusable semantic conditions, while a high-frequency action module continuously outputs control actions without repeatedly invoking the full model. To mitigate the temporal mismatch between stale semantics and the current execution state, we further introduce historical action conditioning and time-misalignment training, which provide short-horizon execution context and improve feedback control robustness under stale semantic conditions. Experiments on LIBERO with

π_{0.5}

and UniVLA, together with real-robot deployment using UniVLA, show that the proposed framework achieves up to 35.6 Hz server-side action-module inference throughput and offers a low-intrusion path to high-frequency closed-loop control without running full VLA inference at control rate.

Shenhao Yan, Ge Wang, Qi Liu +7

Northeastern University · Ising AI · CUHK-Shenzhen

Efficient VLA Models Efficient VLM Inference Efficient VLA Model Inference VLMs for Robotics Vision-Language-Action Models Vision-Based Robot Control+1

Ask PDF

Apr 27, 2026

Libra-VLA: Achieving Learning Equilibrium via Asynchronous Coarse-to-Fine Dual-System

Vision-Language-Action (VLA) models are a promising paradigm for generalist robotic manipulation by grounding high-level semantic instructions into executable physical actions. However, prevailing approaches typically adopt a monolithic generation paradigm, directly mapping visual-linguistic features to high-frequency motor commands in a flat, non-hierarchical fashion. This strategy overlooks the inherent hierarchy of robotic manipulation, where complex actions can be naturally modeled in a Hybrid Action Space, decomposing into discrete macro-directional reaching and continuous micro-pose alignment, severely widening the semantic-actuation gap and imposing a heavy representational burden on grounding high-level semantics to continuous actions. To address this, we introduce Libra-VLA, a novel Coarse-to-Fine Dual-System VLA architecture. We explicitly decouple the learning complexity into a coarse-to-fine hierarchy to strike a training equilibrium, while simultaneously leveraging this structural modularity to implement an asynchronous execution strategy. The Semantic Planner predicts discrete action tokens capturing macro-directional intent, while the Action Refiner conditions on coarse intent to generate high-frequency continuous actions for precise alignment. Crucially, our empirical analysis reveals that performance follows an inverted-U curve relative to action decomposition granularity, peaking exactly when the learning difficulty is balanced between the two sub-systems. With the asynchronous design, our approach offers a scalable, robust, and responsive solution for open-world manipulation.

Yifei Wei, Linqing Zhong, Yi Liu +4

1Beihang University · 2AgiBot

Robotic Control Language-Conditioned Robot Manipulation Robotic Manipulation Vision-Language-Action Models

Ask PDF

May 4, 2026

Understanding Asynchronous Inference Methods for Vision-Language-Action Models

Vision-Language-Action (VLA) models offer a promising path to generalist robot control, but their inference latency causes observation staleness when generated actions are executed asynchronously. Several methods have been proposed concurrently to mitigate this problem: inference-time inpainting (IT-RTC), training-time delay simulation (TT-RTC), future-state-aware conditioning (VLASH), and lightweight residual correction (A2C2). Each takes a fundamentally different approach, but they have so far been evaluated independently with different codebases, base policies, and protocols. We present a systematic comparison of these four methods under controlled conditions. We develop two unified codebases that integrate all methods with harmonized library and dataset versions, and we benchmark them on the Kinetix suite with MLPMixer policies and on the LIBERO manipulation benchmark with SmolVLA, sweeping inference delays up to

d=20

control steps. A2C2's per-step residual correction is the most effective method on Kinetix, holding above 90% solve rate up to

d=8

, and also leads on LIBERO from

d=4

onwards. IT-RTC is competitive at low delays but degrades sharply under long chunks (

H=30

) and high delays. TT-RTC is the most robust training-based method: stable across

d_\max

choices, generalizes beyond its training delay distribution, and adds zero inference overhead. VLASH exhibits a clear low-delay vs. high-delay trade-off governed by the fine-tuning delay range

[0,d_\max]

. Code is available at https://github.com/TheAyos/async-vla-inference

Ayoub Agouzoul

École polytechnique · Systems Group, ETH Zürich

Robotic Control Efficient VLM Inference Efficient VLA Model Inference Robotic Manipulation Vision-Language-Action Models

Ask PDF

Abstract

VLMs for Robotics

Explore similar work

Acting While Understanding: Asynchronous Semantic-Action Decoupling for Real-Time Vision-Language-Action Models

Libra-VLA: Achieving Learning Equilibrium via Asynchronous Coarse-to-Fine Dual-System

Understanding Asynchronous Inference Methods for Vision-Language-Action Models