{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,13]],"date-time":"2026-07-13T22:06:45Z","timestamp":1783980405057,"version":"3.55.0"},"reference-count":25,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,3,25]],"date-time":"2026-03-25T00:00:00Z","timestamp":1774396800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Robot. AI"],"abstract":"<jats:p>\n                    Foundation models for embodied artificial intelligence (Embodied AI) increasingly adopt diffusion modules as the action generation core of vision\u2013language\u2013action (VLA) policies, but the diffusion module\u2019s iterative denoising imposes prohibitive inference latency for real-time deployment. We address this bottleneck in isolation by rethinking the\n                    <jats:italic>diffusion action generation module<\/jats:italic>\n                    itself. We present\n                    <jats:bold>\n                      <jats:italic>Fast Robot Motion Diffusion (FRMD)<\/jats:italic>\n                    <\/jats:bold>\n                    , a fast robot motion diffusion framework that (i) operates in\n                    <jats:italic>trajectory-parameter<\/jats:italic>\n                    space by predicting movement-primitive coefficients in a low-dimensional manifold, and (ii) collapses multi-step sampling into a\n                    <jats:italic>single inference step<\/jats:italic>\n                    via\n                    <jats:italic>trajectory-level consistency distillation<\/jats:italic>\n                    over the probability-flow ordinary differential equation (ODE). Concretely, FRMD replaces stepwise action generation with a one-pass mapping from noise to full trajectories, followed by a fixed-cost basis expansion; this reduces policy latency from hundreds to tens of milliseconds without modifying upstream vision or language encoders. On standard robotic manipulation task benchmarks, FRMD attains 7 times faster than the vanilla diffusion policy and 10 times faster than the state-of-the-art MPD method, while matching the task success of multi-step diffusion policies. By targeting the diffusion component used throughout VLA systems, FRMD provides a plug-in, latency-optimized motion generator that preserves the advantages of diffusion and makes real-time embodied AI feasible.\n                  <\/jats:p>","DOI":"10.3389\/frobt.2026.1751688","type":"journal-article","created":{"date-parts":[[2026,3,25]],"date-time":"2026-03-25T05:31:33Z","timestamp":1774416693000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["FRMD: fast robot motion diffusion via trajectory-level consistency distillation"],"prefix":"10.3389","volume":"13","author":[{"given":"Xirui","family":"Shi","sequence":"first","affiliation":[{"name":"Department of Electrical and Computer Engineering, University of Alberta","place":["Edmonton, AB, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yi","family":"Hu","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, University of Alberta","place":["Edmonton, AB, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jun","family":"Jin","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, University of Alberta","place":["Edmonton, AB, Canada"]},{"name":"Alberta Machine Intelligence Institute (Amii)","place":["Edmonton, AB, Canada"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,3,25]]},"reference":[{"key":"B1","volume-title":"\u03c0","author":"Black","year":"2024"},{"key":"B2","volume-title":"Motion planning diffusion: learning and adapting robot motion planning with diffusion models","author":"Carvalho","year":"2024"},{"key":"B3","volume-title":"Pixart-{\\","author":"Chen","year":"2024"},{"key":"B4","doi-asserted-by":"publisher","DOI":"10.1177\/02783649241273668","article-title":"Diffusion policy: visuomotor policy learning via action diffusion","volume":"44","author":"Chi","year":"2023","journal-title":"Int. J. Robotics Res."},{"key":"B5","first-page":"390","article-title":"Motionlcm: real-time controllable motion generation via latent consistency model","volume-title":"European conference on computer vision","author":"Dai","year":"2024"},{"key":"B6","volume-title":"The ingredients for robotic diffusion transformers","author":"Dasari","year":"2024"},{"key":"B7","first-page":"4567","article-title":"Probabilistic dynamic movement primitives for efficient robotic trajectory generation","volume-title":"Proceedings of the IEEE international conference on robotics and automation (ICRA)","author":"Doe","year":"2017"},{"key":"B8","article-title":"Palm-e: an embodied multimodal language model","author":"Driess","year":"2023"},{"key":"B9","doi-asserted-by":"publisher","first-page":"1602","DOI":"10.1198\/jasa.2011.tm11181","article-title":"Tweedie\u2019s formula and selection bias","volume":"106","author":"Efron","year":"2011","journal-title":"J. Am. Stat. Assoc."},{"key":"B10","doi-asserted-by":"publisher","first-page":"6822","DOI":"10.3390\/s20236822","article-title":"Measuring smoothness as a factor for efficient and socially accepted robot motion","volume":"20","author":"Guill\u00e9n Ruiz","year":"2020","journal-title":"Sensors"},{"key":"B11","first-page":"770","article-title":"Deep residual learning for image recognition","volume-title":"Proceedings of the IEEE conference on computer vision and pattern recognition","author":"He","year":"2016"},{"key":"B12","doi-asserted-by":"publisher","first-page":"6840","DOI":"10.48550\/arXiv.2006.11239","article-title":"Denoising diffusion probabilistic models","author":"Ho","year":"2020","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"B13","doi-asserted-by":"publisher","first-page":"328","DOI":"10.1162\/NECO_a_00393","article-title":"Dynamical movement primitives: learning attractor models for motor behaviors","volume":"25","author":"Ijspeert","year":"2013","journal-title":"Neural Computation"},{"key":"B14","article-title":"Planning with diffusion for flexible behavior synthesis","volume-title":"International conference on learning representations","author":"Janner","year":"2022"},{"key":"B15","article-title":"Latent consistency models for accelerated text-to-image generation","author":"Kim","year":"2023"},{"key":"B16","volume-title":"Decoupled weight decay regularization","author":"Loshchilov","year":"2017"},{"key":"B17","doi-asserted-by":"publisher","first-page":"5775","DOI":"10.48550\/arXiv.2206.00927","article-title":"Dpm-solver: a fast ode solver for diffusion probabilistic model sampling in around 10 steps","volume":"35","author":"Lu","year":"2022","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"B18","volume-title":"Manicm: real-time 3d diffusion policy via consistency model for robotic manipulation","author":"Lu","year":"2024"},{"key":"B19","volume-title":"Maniskill: generalizable manipulation skill benchmark with large-scale demonstrations","author":"Mu","year":"2021"},{"key":"B20","first-page":"115","article-title":"Probabilistic movement primitives","volume-title":"Proceedings of the IEEE international conference on robotics and automation (ICRA)","author":"Paraschos","year":"2013"},{"key":"B21","doi-asserted-by":"publisher","first-page":"1133","DOI":"10.1177\/02783649231201196","article-title":"Dynamic movement primitives in robotics: a tutorial survey","volume":"42","author":"Saveriano","year":"2023","journal-title":"Int. J. Robotics Res."},{"key":"B22","doi-asserted-by":"publisher","first-page":"5338","DOI":"10.1109\/LRA.2024.3382529","article-title":"Movement primitive diffusion: learning gentle robotic manipulation of deformable objects","volume":"9","author":"Scheikl","year":"2024","journal-title":"IEEE Robotics Automation Lett."},{"key":"B23","volume-title":"Denoising diffusion implicit models","author":"Song","year":"2020"},{"key":"B24","volume-title":"Consistency models","author":"Song","year":"2023"},{"key":"B25","first-page":"1094","article-title":"Meta-world: a benchmark and evaluation for multi-task and meta reinforcement learning","volume-title":"Proceedings of Machine Learning Research (PMLR) Conference on robot learning","author":"Yu","year":"2020"}],"container-title":["Frontiers in Robotics and AI"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2026.1751688\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,25]],"date-time":"2026-03-25T05:31:52Z","timestamp":1774416712000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2026.1751688\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,25]]},"references-count":25,"alternative-id":["10.3389\/frobt.2026.1751688"],"URL":"https:\/\/doi.org\/10.3389\/frobt.2026.1751688","relation":{},"ISSN":["2296-9144"],"issn-type":[{"value":"2296-9144","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,25]]},"article-number":"1751688"}}