{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,7,30]],"date-time":"2025-07-30T16:31:43Z","timestamp":1753893103084,"version":"3.41.2"},"reference-count":35,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,4,30]],"date-time":"2025-04-30T00:00:00Z","timestamp":1745971200000},"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>In multi-skill imitation learning for robots, expert datasets with complete motion features are crucial for enabling robots to learn and transition between different skills. However, such datasets are often difficult to obtain. As an alternative, datasets constructed using only joint positions are more accessible, but they are incomplete and lack details, making it challenging for existing methods to effectively learn and model skill transitions. To address these challenges, this study introduces the Seamless Multi-Skill Learning (SMSL) framework. Integrated within the Adversarial Motion Priors framework and incorporating self-trajectory augmentation techniques, SMSL effectively utilizes high-quality historical experiences to guide agents in learning skills and generating smooth, natural transitions between them, addressing the learning difficulties caused by incomplete expert datasets. Additionally, the research incorporates an adaptive command sampling mechanism to balance the training opportunities for skills of various difficulties and prevent catastrophic forgetting. Our experiments highlight potential issues with baseline methods when imitating incomplete expert datasets and demonstrate the superior performance of the SMSL framework. Sim-to-real experiments on real Solo8 robots further validate the effectiveness of SMSL. Overall, this study confirms the SMSL framework\u2019s capability in real robotic applications and underscores its potential for autonomous skill learning and generation from minimal data.<\/jats:p>","DOI":"10.3389\/frobt.2025.1542692","type":"journal-article","created":{"date-parts":[[2025,4,30]],"date-time":"2025-04-30T07:48:57Z","timestamp":1745999337000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Seamless multi-skill learning: learning and transitioning non-similar skills in quadruped robots with limited data"],"prefix":"10.3389","volume":"12","author":[{"given":"Jiaxin","family":"Tu","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Peng","family":"Zhai","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yueqi","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xiaoyi","family":"Wei","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhiyan","family":"Dong","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Lihua","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2025,4,30]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"227","DOI":"10.1007\/s43154-022-00087-4","article-title":"Task allocation using a team of robots","volume":"3","author":"Aziz","year":"2022","journal-title":"Curr. Robot. Rep."},{"key":"B2","first-page":"1371","volume-title":"Survey: robot programming by demonstration","author":"Billard","year":"2008"},{"key":"B3","article-title":"Limited preference aided imitation learning from imperfect demonstrations","volume-title":"Forty-first international conference on machine learning","author":"Cao","year":"2024"},{"key":"B4","doi-asserted-by":"publisher","first-page":"25","DOI":"10.1109\/iros47612.2022.9981973","article-title":"Adversarial motion priors make good substitutes for complex reward functions","author":"Escontrela","year":"2022"},{"key":"B5","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2404.19456","article-title":"Imitation learning: a survey of learning methods, environments and metrics","author":"Gavenski","year":"2024","journal-title":"arXiv Prepr"},{"key":"B6","doi-asserted-by":"publisher","first-page":"139","DOI":"10.1145\/3422622","article-title":"Generative adversarial networks","volume":"63","author":"Goodfellow","year":"2020","journal-title":"Commun. ACM"},{"key":"B7","doi-asserted-by":"publisher","first-page":"3650","DOI":"10.1109\/lra.2020.2976639","article-title":"An open torque-controlled modular robot architecture for legged locomotion research","volume":"5","author":"Grimminger","year":"2020","journal-title":"IEEE Robotics Automation Lett."},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.5555\/3157382.3157608","article-title":"Generative adversarial imitation learning","volume":"29","author":"Ho","year":"2016","journal-title":"Adv. neural Inf. Process. Syst."},{"key":"B9","doi-asserted-by":"publisher","first-page":"9758","DOI":"10.1109\/icra57147.2024.10610151","article-title":"Multi-task learning of active fault-tolerant controller for leg failures in quadruped robots","author":"Hou","year":"2024","journal-title":"arXiv Prepr. arXiv:2402.08996"},{"key":"B10","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3054912","article-title":"Imitation learning: a survey of learning methods","volume":"50","author":"Hussein","year":"2017","journal-title":"ACM Comput. Surv. (CSUR)"},{"key":"B11","doi-asserted-by":"crossref","DOI":"10.15607\/RSS.2021.XVII.011","article-title":"Rma: rapid motor adaptation for legged robots","author":"Kumar","year":"2021"},{"article-title":"Can agents run relay race with strangers? generalization of rl to out-of-distribution trajectories","year":"2023","author":"Lan","key":"B12"},{"key":"B13","doi-asserted-by":"publisher","first-page":"eabc5986","DOI":"10.1126\/scirobotics.abc5986","article-title":"Learning quadrupedal locomotion over challenging terrain","volume":"5","author":"Lee","year":"2020","journal-title":"Sci. robotics"},{"volume-title":"Locomotion control of a lightweight quadruped robot","year":"2022","author":"L\u00e9ziart","key":"B14"},{"key":"B15","doi-asserted-by":"crossref","first-page":"2944","DOI":"10.1109\/ICRA48891.2023.10160421","article-title":"Versatile skill control via self-supervised adversarial imitation of unlabeled mixed motions","volume-title":"2023 IEEE international conference on robotics and automation (ICRA)","author":"Li","year":"2023"},{"article-title":"Learning agile skills via adversarial imitation of rough partial demonstrations","year":"2022","author":"Li","key":"B16"},{"key":"B17","doi-asserted-by":"publisher","first-page":"293","DOI":"10.1007\/bf00992699","article-title":"Self-improving reactive agents based on reinforcement learning, planning and teaching","volume":"8","author":"Lin","year":"1992","journal-title":"Mach. Learn."},{"article-title":"Isaac gym: high performance gpu-based physics simulation for robot learning","year":"2021","author":"Makoviychuk","key":"B18"},{"key":"B19","doi-asserted-by":"publisher","first-page":"2866","DOI":"10.1109\/icra57147.2024.10610528","article-title":"Contrastive initial state buffer for reinforcement learning","author":"Messikommer","year":"2024"},{"key":"B20","doi-asserted-by":"publisher","first-page":"eabk2822","DOI":"10.1126\/scirobotics.abk2822","article-title":"Learning robust perceptive locomotion for quadrupedal robots in the wild","volume":"7","author":"Miki","year":"2022","journal-title":"Sci. robotics"},{"key":"B21","doi-asserted-by":"publisher","first-page":"529","DOI":"10.1038\/nature14236","article-title":"Human-level control through deep reinforcement learning","volume":"518","author":"Mnih","year":"2015","journal-title":"nature"},{"key":"B22","doi-asserted-by":"crossref","first-page":"69","DOI":"10.1007\/978-3-540-74048-3_4","article-title":"Dynamic time warping","volume-title":"Information retrieval for music and motion","author":"M\u00fcller","year":"2007"},{"article-title":"Self-imitation learning","year":"2018","author":"Oh","key":"B23"},{"key":"B24","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3272127.3275014","article-title":"Sfv: reinforcement learning of physical skills from videos","volume":"37","author":"Peng","year":"2018","journal-title":"ACM Trans. Graph. (TOG)"},{"key":"B25","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3450626.3459670","article-title":"Amp: adversarial motion priors for stylized physics-based character control","volume":"40","author":"Peng","year":"2021","journal-title":"ACM Trans. Graph. (ToG)"},{"key":"B26","doi-asserted-by":"publisher","first-page":"2914","DOI":"10.5555\/3495724.3495969","article-title":"Toward the fundamental limits of imitation learning","volume":"33","author":"Rajaraman","year":"2020","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"B27","doi-asserted-by":"publisher","first-page":"2450001","DOI":"10.1142\/s0219843624500014","article-title":"Task-oriented self-imitation learning for robotic autonomous skill acquisition","volume":"21","author":"Ran","year":"2024","journal-title":"Int. J. Humanoid Robotics"},{"article-title":"Prioritized experience replay","year":"2015","author":"Schaul","key":"B28"},{"article-title":"Proximal policy optimization algorithms","year":"2017","author":"Schulman","key":"B29"},{"volume-title":"Reinforcement learning: an introduction","year":"2018","author":"Sutton","key":"B30"},{"key":"B31","doi-asserted-by":"publisher","first-page":"6325","DOI":"10.24963\/ijcai.2019\/882","article-title":"Recent advances in imitation learning from observation","author":"Torabi","year":"2019"},{"key":"B32","doi-asserted-by":"crossref","first-page":"5120","DOI":"10.1109\/ICRA48891.2023.10160751","article-title":"Advanced skills through multiple adversarial motion priors in reinforcement learning","volume-title":"2023 IEEE international conference on robotics and automation (ICRA)","author":"Vollenweider","year":"2023"},{"key":"B33","doi-asserted-by":"publisher","first-page":"4975","DOI":"10.1109\/lra.2023.3290509","article-title":"Learning robust and agile legged locomotion using adversarial motion priors","volume":"8","author":"Wu","year":"2023","journal-title":"IEEE Robotics Automation Lett."},{"key":"B34","article-title":"Sample-efficient multiagent reinforcement learning with reset replay","volume-title":"Forty-first international conference on machine learning","author":"Yang","year":"2024"},{"key":"B35","doi-asserted-by":"publisher","first-page":"12562","DOI":"10.1109\/lra.2022.3220531","article-title":"Robust adaptive ensemble adversary reinforcement learning","volume":"7","author":"Zhai","year":"2022","journal-title":"IEEE Robotics Automation Lett."}],"container-title":["Frontiers in Robotics and AI"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2025.1542692\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,4,30]],"date-time":"2025-04-30T07:49:02Z","timestamp":1745999342000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frobt.2025.1542692\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,4,30]]},"references-count":35,"alternative-id":["10.3389\/frobt.2025.1542692"],"URL":"https:\/\/doi.org\/10.3389\/frobt.2025.1542692","relation":{},"ISSN":["2296-9144"],"issn-type":[{"type":"electronic","value":"2296-9144"}],"subject":[],"published":{"date-parts":[[2025,4,30]]},"article-number":"1542692"}}