{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,10,22]],"date-time":"2025-10-22T22:25:28Z","timestamp":1761171928723,"version":"build-2065373602"},"reference-count":52,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,10,22]],"date-time":"2025-10-22T00:00:00Z","timestamp":1761091200000},"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                    Industrial terminal assembly tasks are often repetitive and involve handling components with tight tolerances that are susceptible to damage. Learning an effective terminal assembly policy in real-world is challenging, as collisions between parts and the environment can lead to slippage or part breakage. In this paper, we propose a safe reinforcement learning approach to develop a visuo-tactile assembly policy that is robust to variations in grasp poses. Our method minimizes collisions between the terminal head and terminal base by decomposing the assembly task into three distinct phases. In the first\n                    <jats:italic>grasp<\/jats:italic>\n                    phase,a vision-guided model is trained to pick the terminal head from an initial bin. In the second\n                    <jats:italic>align<\/jats:italic>\n                    phase, a tactile-based grasp pose estimation model is employed to align the terminal head with the terminal base. In the final\n                    <jats:italic>assembly<\/jats:italic>\n                    phase, a visuo-tactile policy is learned to precisely insert the terminal head into the terminal base. To ensure safe training, the robot leverages human demonstrations and interventions. Experimental results on PLC terminal assembly demonstrate that the proposed method achieves 100% successful insertions across 100 different initial end-effector and grasp poses, while imitation learning and online-RL policy yield only 9% and 0%.\n                  <\/jats:p>","DOI":"10.3389\/frobt.2025.1660244","type":"journal-article","created":{"date-parts":[[2025,10,22]],"date-time":"2025-10-22T04:17:58Z","timestamp":1761106678000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Visuo-tactile feedback policies for terminal assembly facilitated by reinforcement learning"],"prefix":"10.3389","volume":"12","author":[{"given":"Yuchao","family":"Li","sequence":"first","affiliation":[]},{"given":"Ziqi","family":"Jin","sequence":"additional","affiliation":[]},{"given":"Jin","family":"Liu","sequence":"additional","affiliation":[]},{"given":"Daolin","family":"Ma","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2025,10,22]]},"reference":[{"key":"B1","first-page":"1577","article-title":"Efficient online reinforcement learning with offline data","volume-title":"International conference on machine learning (PMLR)","author":"Ball","year":"2023"},{"key":"B2","article-title":"Agibot world colosseo: a large-scale manipulation platform for scalable and intelligent embodied systems","author":"Bu","year":"2025"},{"key":"B3","doi-asserted-by":"publisher","first-page":"3427","DOI":"10.1109\/LRA.2022.3146565","article-title":"Using collocated vision and tactile sensors for visual servoing and localization","volume":"7","author":"Chaudhury","year":"2022","journal-title":"IEEE Robotics Automation Lett."},{"key":"B4","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/WRCSARA60131.2023.10261820","article-title":"Fusing vision and force: a framework of reinforcement learning for elastic peg-in-hole assembly","volume-title":"2023 WRC symposium on advanced robotics and automation (WRC SARA)","author":"Dang","year":"2023"},{"key":"B5","first-page":"56","article-title":"A correct and complete algorithm for the generation of mechanical assembly sequences","volume-title":"1989 IEEE international conference on robotics and automation","author":"De Mello","year":"1989"},{"key":"B6","doi-asserted-by":"crossref","first-page":"811","DOI":"10.1109\/ICRA.2019.8793659","article-title":"A learning framework for high precision industrial assembly","volume-title":"2019 international conference on robotics and automation (ICRA)","author":"Fan","year":"2019"},{"key":"B7","first-page":"158","article-title":"Implicit behavioral cloning","volume-title":"Conference on robot learning","author":"Florence","year":"2022"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1805.11686","article-title":"Variational inverse control with events: a general framework for data-driven reward definition","volume":"31","author":"Fu","year":"2018","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"B9","first-page":"1587","article-title":"Addressing function approximation error in actor-critic methods","volume-title":"International conference on machine learning","author":"Fujimoto","year":"2018"},{"key":"B10","doi-asserted-by":"publisher","first-page":"3812","DOI":"10.1109\/tro.2024.3428430","article-title":"Evetac: an event-based optical tactile sensor for robotic manipulation","volume":"40","author":"Funk","year":"2024","journal-title":"IEEE Trans. Robotics"},{"key":"B11","doi-asserted-by":"publisher","first-page":"201","DOI":"10.1007\/bf01891840","article-title":"Orienting polygonal parts without sensors","volume":"10","author":"Goldberg","year":"1993","journal-title":"Algorithmica"},{"key":"B12","first-page":"1861","article-title":"Soft actor-critic: off-policy maximum entropy deep reinforcement learning with a stochastic actor","volume-title":"International conference on machine learning","author":"Haarnoja","year":"2018"},{"key":"B13","doi-asserted-by":"crossref","first-page":"8298","DOI":"10.1109\/ICRA46639.2022.9812019","article-title":"Visuotactile-rl: learning multimodal manipulation policies with deep reinforcement learning","volume-title":"2022 international conference on robotics and automation (ICRA)","author":"Hansen","year":"2022"},{"key":"B14","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2503.08548","article-title":"Tla: tactile-language-action model for contact-rich manipulation","author":"Hao","year":"2025","journal-title":"arXiv Prepr. arXiv:2503.08548"},{"key":"B15","doi-asserted-by":"crossref","first-page":"5375","DOI":"10.1109\/ICRA46639.2022.9811940","article-title":"Contact-rich manipulation of a flexible object based on deep predictive learning using vision and tactility","volume-title":"2022 international conference on robotics and automation (ICRA)","author":"Ichiwara","year":"2022"},{"key":"B16","doi-asserted-by":"crossref","first-page":"6023","DOI":"10.1109\/ICRA.2019.8794127","article-title":"Residual reinforcement learning for robot control","volume-title":"2019 international conference on robotics and automation (ICRA)","author":"Johannink","year":"2019"},{"key":"B17","doi-asserted-by":"publisher","first-page":"10685","DOI":"10.48550\/arXiv.2203.10685","article-title":"Tactile pose estimation and policy learning for unknown object manipulation","author":"Kelestemur","year":"2022","journal-title":"arXiv Prepr. arXiv:2203"},{"key":"B18","article-title":"Adam: a method for stochastic optimization","author":"Kingma","year":"2014"},{"key":"B19","doi-asserted-by":"crossref","first-page":"10207","DOI":"10.1109\/IROS47612.2022.9982242","article-title":"On cad informed adaptive robotic assembly","volume-title":"2022 IEEE\/RSJ international conference on intelligent robots and systems (IROS)","author":"Koga","year":"2022"},{"key":"B20","doi-asserted-by":"publisher","DOI":"10.1137\/S0363012901385691","article-title":"Actor-critic algorithms","volume":"12","author":"Konda","year":"1999","journal-title":"Adv. neural Inf. Process. Syst."},{"key":"B21","doi-asserted-by":"publisher","first-page":"3838","DOI":"10.1109\/lra.2020.2977257","article-title":"Digit: a novel design for a low-cost compact high-resolution tactile sensor with application to in-hand manipulation","volume":"5","author":"Lambeta","year":"2020","journal-title":"IEEE Robotics Automation Lett."},{"key":"B22","first-page":"3988","article-title":"Localization and manipulation of small parts using gelsight tactile sensing","volume-title":"2014 IEEE\/RSJ international conference on intelligent robots and systems (IEEE)","author":"Li","year":"2014"},{"key":"B23","first-page":"1046","article-title":"Benchmarking off-the-shelf solutions to robotic assembly tasks","author":"Lian","year":"2021"},{"key":"B24","doi-asserted-by":"crossref","first-page":"9227","DOI":"10.1109\/ICRA57147.2024.10610567","article-title":"Generalize by touching: tactile ensemble skill transfer for robotic furniture assembly","volume-title":"2024 IEEE international conference on robotics and automation (ICRA)","author":"Lin","year":"2024"},{"key":"B25","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/tim.2024.3398136","article-title":"Real-time reconstruction of 3d tactile motion field via multi-task learning","volume":"73","author":"Liu","year":"2024","journal-title":"IEEE Trans. Instrum. Meas."},{"key":"B26","article-title":"Motion planning and the design of orienting devices for vibratory part feeders","volume-title":"IEEE journal of robotics and automation","author":"Lozano-P\u00e9rez","year":"1986"},{"key":"B27","doi-asserted-by":"publisher","first-page":"3","DOI":"10.1177\/027836498400300101","article-title":"Automatic synthesis of fine-motion strategies for robots","volume":"3","author":"Lozano-Perez","year":"1984","journal-title":"Int. J. Robotics Res."},{"key":"B28","doi-asserted-by":"publisher","first-page":"333","DOI":"10.1109\/tmech.2024.3384432","article-title":"Dexitac: soft dexterous tactile gripping","volume":"30","author":"Lu","year":"2024","journal-title":"IEEE\/ASME Trans. Mechatronics"},{"key":"B29","doi-asserted-by":"crossref","DOI":"10.15607\/RSS.2021.XVII.088","article-title":"Robust multi-modal policies for industrial assembly via reinforcement learning and demonstrations: a large-scale study","author":"Luo","year":"2021"},{"key":"B30","doi-asserted-by":"crossref","first-page":"16961","DOI":"10.1109\/ICRA57147.2024.10610040","article-title":"Serl: a software suite for sample-efficient robotic reinforcement learning","volume-title":"2024 IEEE international conference on robotics and automation (ICRA)","author":"Luo","year":"2024"},{"key":"B31","first-page":"195","article-title":"Automatic assembly by G. Boothroyd, C. poli and L.E. murch, marcel dekker, New York, 378 pp., 1982 ($45.00)","volume-title":"Robotica","author":"McKee","year":"1985"},{"key":"B32","doi-asserted-by":"publisher","first-page":"09359","DOI":"10.48550\/arXiv.2006.09359","article-title":"Accelerating online reinforcement learning with offline datasets. arxiv 2020","author":"Nair","year":"2020","journal-title":"arXiv Prepr. arXiv:2006"},{"key":"B33","doi-asserted-by":"crossref","DOI":"10.15607\/RSS.2022.XVIII.035","article-title":"Factory: fast contact for robotic assembly","author":"Narang","year":"2022"},{"key":"B34","doi-asserted-by":"publisher","first-page":"98","DOI":"10.1177\/027836498900800607","article-title":"Some paradigms for the automated design of parts feeders","volume":"8","author":"Natarajan","year":"1989","journal-title":"Int. J. Robotics Res."},{"key":"B35","first-page":"4625","article-title":"Tactile-sensitive newtonianvae for high-accuracy industrial connector insertion","author":"Okumura","year":"2022"},{"key":"B36","doi-asserted-by":"publisher","first-page":"51840","DOI":"10.1109\/access.2024.3385426","article-title":"Advancements in deep reinforcement learning and inverse reinforcement learning for robotic manipulation: toward trustworthy, interpretable, and explainable artificial intelligence","volume":"12","author":"Ozalp","year":"2024","journal-title":"IEEE Access"},{"key":"B37","doi-asserted-by":"publisher","first-page":"429","DOI":"10.1243\/pime_proc_1995_209_173_02","article-title":"Fine motion strategies for robotic peg-hole insertion","volume":"209","author":"Qiao","year":"1995","journal-title":"Proc. Institution Mech. Eng. Part C J. Mech. Eng. Sci."},{"key":"B38","first-page":"10428","article-title":"Designing network design spaces","volume-title":"Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition","author":"Radosavovic","year":"2020"},{"key":"B39","doi-asserted-by":"publisher","first-page":"10087","DOI":"10.15607\/RSS.2018.XIV.049","article-title":"Learning complex dexterous manipulation with deep reinforcement learning and demonstrations","author":"Rajeswaran","year":"2017","journal-title":"arXiv Prepr. arXiv:1709"},{"key":"B40","first-page":"5548","article-title":"Deep reinforcement learning for industrial insertion tasks with visual inputs and natural rewards","volume-title":"2020 IEEE\/RSJ international conference on intelligent robots and systems (IROS)","author":"Schoettler","year":""},{"key":"B41","first-page":"9728","article-title":"Meta-reinforcement learning for robotic industrial insertion tasks","volume-title":"2020 IEEE\/RSJ international conference on intelligent robots and systems (IROS)","author":"Schoettler","year":""},{"key":"B42","doi-asserted-by":"publisher","first-page":"5509","DOI":"10.1109\/lra.2021.3076971","article-title":"Insertionnet-a scalable solution for insertion","volume":"6","author":"Spector","year":"2021","journal-title":"IEEE Robotics Automation Lett."},{"key":"B43","doi-asserted-by":"crossref","first-page":"6330","DOI":"10.1109\/ICRA46639.2022.9811798","article-title":"Insertionnet 2.0: minimal contact multi-step insertion using multimodal multiview sensory input","volume-title":"2022 international conference on robotics and automation (ICRA)","author":"Spector","year":"2022"},{"key":"B44","doi-asserted-by":"crossref","first-page":"10781","DOI":"10.1109\/ICRA46639.2022.9811832","article-title":"Gelslim 3.0: high-Resolution measurement of shape, force and slip in a compact tactile-sensing finger","volume-title":"2022 international conference on robotics and automation (ICRA)","author":"Taylor","year":"2022"},{"key":"B45","doi-asserted-by":"crossref","first-page":"2690","DOI":"10.1145\/3474085.3475414","article-title":"Elastic tactile simulation towards tactile-visual perception","volume-title":"Proceedings of the 29th ACM international conference on multimedia","author":"Wang","year":"2021"},{"key":"B46","doi-asserted-by":"publisher","first-page":"3930","DOI":"10.1109\/lra.2022.3146945","article-title":"Tacto: a fast, flexible, and open-source simulator for high-resolution vision-based tactile sensors","volume":"7","author":"Wang","year":"2022","journal-title":"IEEE Robotics Automation Lett."},{"key":"B47","doi-asserted-by":"crossref","DOI":"10.15607\/RSS.2022.XVIII.044","article-title":"You only demonstrate once: category-level manipulation from single visual demonstration","author":"Wen","year":"2022"},{"key":"B48","first-page":"11831","article-title":"Tacdiffusion: force-domain diffusion policy for precise tactile manipulation","author":"Wu","year":"2025"},{"key":"B49","doi-asserted-by":"publisher","first-page":"2762","DOI":"10.3390\/s17122762","article-title":"Gelsight: high-resolution robot tactile sensors for estimating geometry and force","volume":"17","author":"Yuan","year":"2017","journal-title":"Sensors"},{"key":"B50","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2505.09577","article-title":"Vtla: vision-tactile-language-action model with preference learning for insertion manipulation","author":"Zhang","year":"2025","journal-title":"arXiv Prepr. arXiv:2505.09577"},{"key":"B51","doi-asserted-by":"crossref","first-page":"6386","DOI":"10.1109\/ICRA46639.2022.9812312","article-title":"Offline meta-reinforcement learning for industrial insertion","volume-title":"2022 international conference on robotics and automation (ICRA) (IEEE)","author":"Zhao","year":"2022"},{"key":"B52","doi-asserted-by":"publisher","first-page":"538","DOI":"10.1109\/tro.2024.3508134","article-title":"Tac-man: tactile-informed prior-free manipulation of articulated objects","volume":"41","author":"Zhao","year":"2024","journal-title":"IEEE Trans. Robotics"}],"container-title":["Frontiers in Robotics and AI"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2025.1660244\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,22]],"date-time":"2025-10-22T04:18:07Z","timestamp":1761106687000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2025.1660244\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,22]]},"references-count":52,"alternative-id":["10.3389\/frobt.2025.1660244"],"URL":"https:\/\/doi.org\/10.3389\/frobt.2025.1660244","relation":{},"ISSN":["2296-9144"],"issn-type":[{"value":"2296-9144","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,22]]},"article-number":"1660244"}}