mlsanityData to 7 Oct 2026

Streaming Diffusion Policy: Fast Policy Synthesis with Variable Noise Diffusion Models

arXiv 2406.04806 ↗First posted 7 Jun 2024cs.RO cs.AIcited by 51

On the map: Diffusion models → Diffusion and flow policies for robotics

also in 1 more
Abstract (arXiv)

Diffusion models have seen rapid adoption in robotic imitation learning, enabling autonomous execution of complex dexterous tasks. However, action synthesis is often slow, requiring many steps of iterative denoising, limiting the extent to which models can be used in tasks that require fast reactive policies. To sidestep this, recent works have explored how the distillation of the diffusion process can be used to accelerate policy synthesis. However, distillation is computationally expensive and can hurt both the accuracy and diversity of synthesized actions. We propose SDP (Streaming Diffusion Policy), an alternative method to accelerate policy synthesis, leveraging the insight that generating a partially denoised action trajectory is substantially faster than a full output action trajectory. At each observation, our approach outputs a partially denoised action trajectory with variable levels of noise corruption, where the immediate action to execute is noise-free, with subsequent actions having increasing levels of noise and uncertainty. The partially denoised action trajectory for a new observation can then be quickly generated by applying a few steps of denoising to the previously predicted noisy action trajectory (rolled over by one timestep). We illustrate the efficacy of this approach, dramatically speeding up policy synthesis while preserving performance across both simulated and real-world settings.

Key facts. Cited by 51 papers in the map (data through 7 Oct 2026). Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models. Builds on 31 of its 37 references that are in the map. First posted on arXiv 7 Jun 2024.

Builds on

31 of 37 references.

Closest ancestors · 12

Same direction, newest first

  1. 2024Consistency Policy: Accelerated Visuomotor Policies via Consistency Distillation13 May
  2. 2024Don't Start from Scratch: Behavioral Refinement via Interpolant-based Policy Diffusion25 Feb
  3. 2023PlayFusion: Skill Acquisition via Diffusion from Language-Annotated Play7 Dec
  4. 2023Adaptive Online Replanning with Diffusion Models14 Oct
  5. 2023EDMP: Ensemble-of-costs-guided Diffusion for Motion Planning20 Sep
  6. 2023Motion Planning Diffusion: Learning and Planning of Robot Motions with Diffusion Models3 Aug
  7. 2023Crossway Diffusion: Improving Diffusion-based Visuomotor Policy via Self-supervised Learning4 Jul
  8. 2023Goal-Conditioned Imitation Learning using Score-based Diffusion Policies5 Apr
  9. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusion7 Mar
  10. 2023Imitating Human Behaviour with Diffusion Models25 Jan
  11. 2023Diffusion-based Generation, Optimization, and Planning in 3D Scenes15 Jan
  12. 2022SE(3)-DiffusionFields: Learning smooth cost functions for joint grasp and motion optimization through diffusion8 Sep

Other references · 19

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2020Denoising Diffusion Implicit Modelscited by 9,655
  3. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 2,994
  4. 2023Consistency Modelscited by 1,778
  5. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  6. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,023
  7. 2022On Distillation of Guided Diffusion Modelscited by 721
  8. 2021What Matters in Learning from Offline Human Demonstrations for Robot Manipulationcited by 699
  9. 2023Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusioncited by 448
  10. 2024Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robotscited by 373
  11. 2023Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMCcited by 216
  12. 2023Scaling Up and Distilling Down: Language-Guided Robot Skill Acquisitioncited by 156
7 more
  1. 2023AR-Diffusion: Auto-Regressive Diffusion Model for Text Generationcited by 113
  2. 2024Rolling Diffusion Modelscited by 102
  3. 2023TEDi: Temporally-Entangled Diffusion for Long-Term Motion Synthesiscited by 48
  4. 2024PoCo: Policy Composition from and for Heterogeneous Robot Learningcited by 42
  5. 2023Movement Primitive Diffusion: Learning Gentle Robotic Manipulation of Deformable Objectscited by 40
  6. 2022Language Control Diffusion: Efficiently Scaling through Space, Time, and Taskscited by 20
  7. 2023Memory-Consistent Neural Networks for Imitation Learningcited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 1; 2025: 19; 2026: 31.

51 papers.

Sort:

By area:

  1. 2025Real-Time Execution of Action Chunking Flow Policiescited by 199
  2. 2025Stable Video Infinity: Infinite-Length Video Generation with Error Recyclingcited by 68
  3. 2025Leave No Observation Behind: Real-time Correction for VLA Action Chunkscited by 45
  4. 2025Streaming Flow Policy: Simplifying diffusion/flow-matching policies by treating action trajectories as flow trajectoriescited by 34
  5. 2025Diffuse-CLoC: Guided Diffusion for Physics-based Character Look-ahead Controlcited by 27
  6. 2026FASTER: Rethinking Real-Time Flow VLAscited by 20
  7. 2026Learning Native Continuation for Action Chunking Flow Policiescited by 18
  8. 2025SAIL: Faster-than-Demonstration Execution of Imitation Learning Policiescited by 16
  9. 2025Falcon: Fast Visuomotor Policies via Partial Denoisingcited by 14
  10. 2025IMLE Policy: Fast and Sample Efficient Visuomotor Policy Learning via Implicit Maximum Likelihood Estimationcited by 13
  11. 2025Responsive Noise-Relaying Diffusion Policy: Responsive and Efficient Visuomotor Controlcited by 13
  12. 2025FreqPolicy: Efficient Flow-based Visuomotor Policy via Frequency Consistencycited by 13
38 more
  1. 2024Score and Distribution Matching Policy: Advanced Accelerated Visuomotor Policies via Matched Distillationcited by 12
  2. 2025D3P: Dynamic Denoising Diffusion Policy via Reinforcement Learningcited by 10
  3. 2025Real-Time Iteration Scheme for Diffusion Policycited by 8
  4. 2025Compose Your Policies! Improving Diffusion-based or Flow-based Robot Policies via Test-time Distribution-level Compositioncited by 8
  5. 2025Vid-CamEdit: Video Camera Trajectory Editing with Generative Rendering from Estimated Geometrycited by 7
  6. 2026STEP: Warm-Started Visuomotor Policies with Spatiotemporal Consistency Predictioncited by 7
  7. 2025RA-DP: Rapid Adaptive Diffusion Policy for Training-Free High-frequency Robotics Replanningcited by 6
  8. 2026Efficient Video Diffusion Models: Advancements and Challengescited by 6
  9. 2026Sparse ActionGen: Accelerating Diffusion Policy with Real-time Pruningcited by 5
  10. 2026TIDAL: Temporally Interleaved Diffusion and Action Loop for High-Frequency VLA Controlcited by 5
  11. 2026One-Step Flow Policy: Self-Distillation for Fast Visuomotor Policiescited by 5
  12. 2025Delay-Aware Diffusion Policy: Bridging the Observation-Execution Gap in Dynamic Taskscited by 4
  13. 2026πR2: Reactive Real-time Flow Policiescited by 4
  14. 2025Real-Time Diffusion Policies for Games: Enhancing Consistency Policies with Q-Ensemblescited by 3
  15. 2026Smoother Action Chunking Flow Policy via Prior-Corrected Orthogonal Trust-Region Guidancecited by 2
  16. 2026ActionCache: Training-Free Acceleration for Vision-Language-Action Models with Action Caching and Refinementcited by 2
  17. 20253D-CovDiffusion: 3D-Aware Diffusion Policy for Coverage Path Planningcited by 1
  18. 2026Accelerated Sequential Flow Matching: A Bayesian Filtering Perspectivecited by 1
  19. 2026Implicit Drifting Policy: One-Step Action Generation via Conditional Expert Geometrycited by 1
  20. 2026TacForcing: Streaming Action Generation with Execution-Time Tactile Feedbackcited by 1
  21. 2026Rolling-WAM: World Action Models with Rolling Imaginationcited by 1
  22. 2025SeFA-Policy: Fast and Accurate Visuomotor Policy Learning with Selective Flow Alignmentcited by 0
  23. 2026Liquid Networks with Mixture Density Heads for Efficient Imitation Learningcited by 0
  24. 2026EgoForce: Robust Online Egocentric Motion Reconstruction via Diffusion Forcingcited by 0
  25. 2026Test-time Sparsity for Extreme Fast Action Diffusioncited by 0
  26. 2026From Noise to Control: Parameterized Diffusion Policiescited by 0
  27. 2026Real-Time Execution with Autoregressive Policiescited by 0
  28. 2026LAGO Policy: Latency-Aware Asynchronous Diffusion Policies with Goal-Directed Collision-Free Planning for Smooth Manipulationcited by 0
  29. 2026BayesFP: Posterior Estimation for Flow-Based Policies via Feynman-Kac Samplingcited by 0
  30. 2026Spatial Attention: Adapting Execution Horizons for Diffusion Policies via Observation Sensitivitycited by 0
  31. 2026A Single Diffusion-Policy Controller for Multi-Task Block Pushing with Zero-Shot Sim-to-Real Transfercited by 0
  32. 2026PredActor: Predictive Action Diffusion for Steerable Onboard Humanoid Controlcited by 0
  33. 2026Imperfection for Precision: Upcycling Imperfect Data for High-Precision Robotic Manipulationcited by 0
  34. 2026RAVEL: Asynchronous Rolling Inference for Flow-Based Vision-Language-Action Modelscited by 0
  35. 2026Revision, Not Restart: Revisable Visual Plans for Closed-Loop World-Action Modelscited by 0
  36. 2026Staircase Policy: Streaming Inference for World-Action Models with Large Action Chunkscited by 0
  37. 2026Urgent Actions Go First: Urgency-Aware Denoising for Real-Time VLA Controlcited by 0
  38. 2026World Motion Models: Flexible Sequence Modeling of SE(3) Trajectoriescited by 0

50 most cited of 51 All 51 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2024One-Step Diffusion Policy: Fast Visuomotor Policies via Diffusion Distillation
  2. 2025Efficient Task-specific Conditional Diffusion Policies: Shortcut Model Acceleration and SO(3) Optimization
  3. 2025Two-Steps Diffusion Policy for Robotic Manipulation via Genetic Denoising
  4. 2025Time-Unified Diffusion Policy with Action Discrimination for Robotic Manipulation
  5. 2024Unpacking the Individual Components of Diffusion Policy
  6. 2026Immiscible Diffusion Policy: Preserving Multimodal Robot Actions through Label-Free Noise Assignment
  7. 2026Action-to-Action Flow Matching
  8. 2024WARPD: World model Assisted Reactive Policy Diffusion