mlsanityData to 14 Oct 2026

Efficient Vision-Language-Action Management and Serving for Robot Factories

arXiv 2609.12075 ↗First posted 10 Sep 2026cs.DC cs.AR cs.LGcited by 0

On the map: Multimodal LLMs / vision-language models → Embodied and 3D vision-language models

also in 2 more
Abstract (arXiv)

Vision-Language-Action (VLA) models show high robotic manipulation capabilities via a two-stage design: a Vision-Language Model (VLM) stage followed by an Action Diffusion Transformer (ADiT) stage. Since robots must meet strict Service-Level Objectives (SLOs) for safety, VLA inference is inherently latency-critical. Meeting these SLOs requires high-end GPUs, yet weight, cost, and power constraints preclude integrating such GPUs on-robot. Prior works offload VLA inference to edge servers that serve many robots on VLA models. However, current VLA systems lack support for multi-request, multi-model execution on a multi-GPU server under SLOs, while existing serving systems for multi-stage models are optimized for throughput and stage disaggregation across separate GPUs, which are ill-suited for the millisecond-scale stages of VLA models. We design Robion, the first VLA serving and management system for multi-robot, multi-model requests on multi-GPU edge servers that meets SLOs. Our serving engine disaggregates the VLM and ADiT stages within a GPU via two streams, dynamically restricting the SMs on VLM stream so ADiT always finds SMs to run alongside it, and co-locates multiple models by sharing these streams across them, prioritizing requests by least remaining SLO time. Our management engine enables flexible model placements on multi-GPU servers, and integrates an intelligent traffic controller that maximizes per-model batching under the chosen placement while bounding each GPU's load to meet SLOs. For individual models, Robion serves on average 6.7× and 1.5× higher robot load within 98% SLO attainment over vLLM-Omni, the most widely used multi-stage serving system, and Monolithic, which runs VLM and ADiT as a single pipeline, respectively. In a large-scale experiment of serving 8 different models on a 4-GPU server, Robion can serve up to 64 robots within 98% SLO attainment.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 48 of its 76 references that are in the map. First posted on arXiv 10 Sep 2026.

Builds on

48 of 76 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Vision-Language-Action Models for Robotics: A Review Towards Real-World Applications8 Oct

Other references · 47

  1. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  2. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  3. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,761
  4. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,568
  5. 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,141
  6. 2024DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Servingcited by 617
  7. 2023Splitwise: Efficient generative LLM inference using phase splittingcited by 558
  8. 2024Taming Throughput-Latency Tradeoff in LLM Inference with Sarathi-Servecited by 526
  9. 2026Fast-WAM: Do World Action Models Need Test-time Future Imagination?cited by 349
  10. 2024A Survey on Vision-Language-Action Models for Embodied AIcited by 279
  11. 2025FlashInfer: Efficient and Customizable Attention Engine for LLM Inference Servingcited by 258
  12. 2024Mooncake: A KVCache-centric Disaggregated Architecture for LLM Servingcited by 196
35 more
  1. 2024Llumnix: Dynamic Scheduling for Large Language Model Servingcited by 174
  2. 2025Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tailcited by 146
  3. 2020Serving DNNs like Clockwork: Performance Predictability from the Bottom Upcited by 120
  4. 2025PD-VLA: Accelerating Vision-Language-Action Model Integrated with Action Chunking via Parallel Decodingcited by 88
  5. 2025VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Cachingcited by 86
  6. 2023Fairness in Serving Large Language Modelscited by 78
  7. 2025Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Surveycited by 69
  8. 2024xDiT: an Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelismcited by 56
  9. 2024MuxServe: Flexible Spatial-Temporal Multiplexing for Multiple LLM Servingcited by 52
  10. 2024ExeGPT: Constraint-Aware Resource Scheduling for LLM Inferencecited by 45
  11. 2025Running VLAs at Real-time Speedcited by 43
  12. 2025A Survey on Efficient Vision-Language-Action Modelscited by 34
  13. 2026Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectoriescited by 29
  14. 2026How Fast Can I Run My VLA? Demystifying VLA Inference Performance with VLA-Perfcited by 19
  15. 2025Seesaw: High-throughput LLM Inference via Model Re-shardingcited by 15
  16. 2022iGniter: Interference-Aware GPU Resource Provisioning for Predictable DNN Inference in the Cloudcited by 14
  17. 2025Saliency-Aware Quantized Imitation Learning for Efficient Robotic Controlcited by 12
  18. 2026TIC-VLA: A Think-in-Control Vision-Language-Action Model for Robot Navigation in Dynamic Environmentscited by 10
  19. 2026AsyncVLA: An Asynchronous VLA for Fast and Robust Navigation on the Edgecited by 10
  20. 2025LithOS: An Operating System for Efficient Machine Learning on GPUscited by 9
  21. 2024Tally: Non-Intrusive Performance Isolation for Concurrent Deep Learning Workloadscited by 7
  22. 2026Execution-State Capsules: Graph-Bound Execution-State Checkpoint and Restore for Low-Latency, Small-Batch, On-Device Physical-AI Servingcited by 7
  23. 2025MoE-Inference-Bench: Performance Evaluation of Mixture of Expert Large Language and Vision Modelscited by 6
  24. 2025TetriServe: Efficiently Serving Mixed DiT Workloadscited by 4
  25. 2026OxyGen: Unified KV Cache Management for VLA Inference under Multi-Task Parallelismcited by 4
  26. 2026GENSERVE: Efficient Co-Serving of Heterogeneous Diffusion Model Workloadscited by 4
  27. 2026M*: A Modular, Extensible, Serving System for Multimodal Modelscited by 3
  28. 2026Offload or Overload: A Platform Measurement Study of Mobile Robotic Manipulation Workloadscited by 2
  29. 2026Action ControlNet: A Lightweight Delay-Aware Adapter for Smooth Asynchronous Control in Vision-Language-Action Modelscited by 2
  30. 2026Action Chunk Scheduling for Batched Robot Policy Servingcited by 2
  31. 2026Agile-VLA: Few-Shot Industrial Pose Rectification via Implicit Affordance Anchoringcited by 1
  32. 2026Kairos: A Scalable Serving System for Physical AIcited by 1
  33. 2026GF-DiT: Scheduling Parallelism for Diffusion Transformer Servingcited by 1
  34. 2026RT-VLA: Real-Time Vision-Language-Action Models via Knowledge Distillationcited by 1
  35. 2026FlashRT: Agent Harness for Guiding Agents to Deploy Real-Time Multimodal Applicationscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Characterizing Vision-Language-Action Models across XPUs: Constraints and Acceleration for On-Robot Deployment
  2. 2026KerColle: Unlocking Fine-Grained GPU Concurrency in Vision-Language-Action Models
  3. 2026Beyond LLM Serving: Characterizing Vision-Language-Action Workloads for Embodied AI System Design
  4. 2025Cross-Platform Scaling of Vision-Language-Action Models from Edge to Cloud GPUs
  5. 2026vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models
  6. 2026VLAgents: A Policy Server for Efficient VLA Inference
  7. 2026RhinoVLA Technical Report
  8. 2025ActionFlow: A Pipelined Action Acceleration for Vision Language Models on Edge