{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,30]],"date-time":"2026-07-30T04:59:56Z","timestamp":1785387596121,"version":"3.55.0"},"reference-count":27,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,30]],"date-time":"2026-07-30T00:00:00Z","timestamp":1785369600000},"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>Embodied picking for e-commerce fulfillment remains vulnerable to dense clutter, reflective packaging, and background variation, which can undermine the effectiveness and robustness of visuomotor policies learned from demonstrations. A key limitation is the absence of explicit object grounding, causing policies to exploit spurious contextual cues rather than task-relevant visual evidence. To address this issue, we propose the Foveated Diffusion Policy (FDP), which integrates object-centric visual grounding into diffusion-based action generation. FDP adopts a panorama\u2013fovea dual-stream visual encoder: a low-resolution panorama stream captures global scene context, while a detection-guided fovea stream uses ROI Align to extract high-resolution object tokens. At each denoising step, these tokens are incorporated into the diffusion denoiser through object-guided cross-attention, thereby grounding action generation in target-specific visual evidence instead of undifferentiated global representations. We evaluate FDP on ten benchmark datasets spanning six simulated manipulation tasks with both proficient-human and multi-human demonstrations. FDP achieves average success rates of 81.0% and 76.8%, respectively, outperforming diffusion-based baselines and delivering particularly strong performance on object-centric tasks, including Push-T (86.7 \u00b1 2.5%) and Lift (100.0 \u00b1 0.0%). Ablation studies on a diagnostic subset show that foveated conditioning improves the average success rate by 11.3 percentage points over the non-foveated variant. FDP also maintains an 84.0% success rate under severe localization perturbations (\u03c3 = 10 pixels), demonstrating resilience to detection noise. Furthermore, accelerated DDIM sampling reduces policy-inference latency to approximately 55 ms with 10 denoising steps, supporting the feasibility of closed-loop control. Collectively, these results demonstrate that explicit object grounding is a promising and computationally practical approach to improving the robustness of visuomotor policies in warehouse-relevant simulated manipulation scenarios.<\/jats:p>","DOI":"10.3389\/frobt.2026.1883320","type":"journal-article","created":{"date-parts":[[2026,7,30]],"date-time":"2026-07-30T04:21:26Z","timestamp":1785385286000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Object-grounded embodied picking for e-commerce warehouse fulfillment: a foveated diffusion policy for operational robustness"],"prefix":"10.3389","volume":"13","author":[{"given":"Chaoqun","family":"Li","sequence":"first","affiliation":[{"name":"School of Education and Management, Bozhou Vocational and Technical College","place":["Bozhou, Anhui, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Longfei","family":"Wang","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Zhejiang Normal University","place":["Jinhua, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Huiying","family":"Xu","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Zhejiang Normal University","place":["Jinhua, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaolei","family":"Zhang","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Zhejiang Normal University","place":["Jinhua, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhenglong","family":"Wan","sequence":"additional","affiliation":[{"name":"Hefei Geek+ Robotics Co., Ltd.","place":["Hefei, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xinzhong","family":"Zhu","sequence":"additional","affiliation":[{"name":"School of Computer Science and Technology, Zhejiang Normal University","place":["Jinhua, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,30]]},"reference":[{"key":"B1","first-page":"4254","article-title":"Pay attention! robustifying a deep visuomotor policy through task-focused visual attention","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Abolghasemi","year":"2019"},{"key":"B2","doi-asserted-by":"crossref","first-page":"381","DOI":"10.1007\/978-3-031-20074-8_22","article-title":"ClearPose: large-scale transparent object dataset and benchmark","volume-title":"Computer Vision\u2013ECCV 2022","author":"Chen","year":"2022"},{"key":"B3","doi-asserted-by":"publisher","first-page":"1684","DOI":"10.1177\/02783649241273668","article-title":"Diffusion policy: visuomotor policy learning via action diffusion","volume":"44","author":"Chi","year":"2024","journal-title":"Int. J. Robotics Res."},{"key":"B4","doi-asserted-by":"publisher","first-page":"172","DOI":"10.1109\/TASE.2016.2600527","article-title":"Analysis and observations from the first amazon picking challenge","volume":"15","author":"Correll","year":"2018","journal-title":"IEEE Trans. Automation Sci. Eng."},{"key":"B5","doi-asserted-by":"publisher","first-page":"45","DOI":"10.1109\/MRA.2023.3310861","article-title":"The cluttered environment picking benchmark (CEPB) for advanced warehouse automation: evaluating the perception, planning, control, and grasping of manipulation systems","volume":"31","author":"D\u2019Avella","year":"2024","journal-title":"IEEE Robotics and Automation Mag."},{"key":"B6","first-page":"49","article-title":"Guided cost learning: deep inverse optimal control via policy optimization","volume-title":"Proceedings of the 33rd International Conference on Machine Learning","author":"Finn","year":"2016"},{"key":"B7","first-page":"158","article-title":"Implicit behavioral cloning","volume-title":"Proceedings of the 5th Conference on Robot Learning","author":"Florence","year":"2022"},{"key":"B8","doi-asserted-by":"publisher","first-page":"665","DOI":"10.1038\/s42256-020-00257-z","article-title":"Shortcut learning in deep neural networks","volume":"2","author":"Geirhos","year":"2020","journal-title":"Nat. Mach. Intell."},{"key":"B9","doi-asserted-by":"publisher","first-page":"188","DOI":"10.1016\/j.tics.2005.02.009","article-title":"Eye movements in natural behavior","volume":"9","author":"Hayhoe","year":"2005","journal-title":"Trends Cognitive Sci."},{"key":"B10","first-page":"6840","article-title":"Denoising diffusion probabilistic models","volume-title":"Advances in Neural Information Processing Systems","author":"Ho","year":"2020"},{"key":"B11","first-page":"19","article-title":"BOP: benchmark for 6d object pose estimation","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Hodan","year":"2018"},{"key":"B12","first-page":"9902","article-title":"Planning with diffusion for flexible behavior synthesis","volume-title":"Proceedings of the 39th International Conference on Machine Learning","author":"Janner","year":"2022"},{"key":"B13","first-page":"1","article-title":"End-to-end training of deep visuomotor policies","volume":"17","author":"Levine","year":"2016","journal-title":"J. Mach. Learn. Res."},{"key":"B14","first-page":"38","article-title":"Grounding DINO: marrying DINO with grounded pre-training for open-set object detection","volume-title":"Lecture Notes in Computer Science","author":"Liu","year":"2024"},{"key":"B15","first-page":"11525","article-title":"Object-centric learning with slot attention","volume-title":"Advances in Neural Information Processing Systems","author":"Locatello","year":"2020"},{"key":"B16","article-title":"VIP: towards universal visual reward and representation via value-implicit pre-training","volume-title":"Proceedings of the Eleventh International Conference on Learning Representations","author":"Ma","year":"2023"},{"key":"B17","first-page":"1678","article-title":"What matters in learning from offline human demonstrations for robot manipulation","volume-title":"Proceedings of the 5th Conference on Robot Learning","author":"Mandlekar","year":"2022"},{"key":"B18","first-page":"22955","article-title":"Behavior transformers: cloning k modes with one stone","volume-title":"Advances in Neural Information Processing Systems","author":"Muhammad","year":"2022"},{"key":"B19","first-page":"892","article-title":"R3M: a universal visual representation for robot manipulation","volume-title":"Proceedings of the 6th Conference on Robot Learning","author":"Nair","year":"2023"},{"key":"B20","first-page":"627","article-title":"A reduction of imitation learning and structured prediction to no-regret online learning","volume-title":"Proceedings of the 14th International Conference on Artificial Intelligence and Statistics","author":"Ross","year":"2011"},{"key":"B21","first-page":"894","article-title":"CLIPort: what and where pathways for robotic Manipulation","volume-title":"Proceedings of the 5th Conference on Robot Learning","author":"Shridhar","year":"2022"},{"key":"B22","article-title":"Denoising diffusion implicit models","volume-title":"Proceedings of the International Conference on Learning Representations","author":"Song","year":"2021"},{"key":"B23","first-page":"3343","article-title":"DenseFusion: 6d object pose estimation by iterative dense fusion","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Wang","year":"2019"},{"key":"B24","volume-title":"PoseCNN: A Convolutional Neural Network for 6d Object Pose Estimation in Cluttered Scenes","author":"Xiang","year":"2017"},{"key":"B25","volume-title":"ROBI: A multi-view Dataset for Reflective Objects in Robotic bin-picking","author":"Yang","year":"2021"},{"key":"B26","first-page":"726","article-title":"Transporter networks: rearranging the visual world for robotic manipulation","volume-title":"Proceedings of the 2020 Conference on Robot Learning","author":"Zeng","year":"2021"},{"key":"B27","doi-asserted-by":"crossref","DOI":"10.15607\/RSS.2023.XIX.016","volume-title":"Learning fine-grained Bimanual Manipulation with low-cost Hardware","author":"Zhao","year":"2023"}],"container-title":["Frontiers in Robotics and AI"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2026.1883320\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,30]],"date-time":"2026-07-30T04:21:31Z","timestamp":1785385291000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2026.1883320\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,30]]},"references-count":27,"alternative-id":["10.3389\/frobt.2026.1883320"],"URL":"https:\/\/doi.org\/10.3389\/frobt.2026.1883320","relation":{},"ISSN":["2296-9144"],"issn-type":[{"value":"2296-9144","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,30]]},"article-number":"1883320"}}