[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84846-en":3,"doc-seo-84846-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84846,8796095462418,"Noah","https://ap-avatar.wpscdn.com/avatar/80000253c1241d02b47?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778826106357471780",8,"Research & Report","Efficient Transfer Learning of Robot Dynamic Models Using Morphological Similarity","A neural network–based transfer learning framework models the dynamics of soft, fin-actuated underwater robots by exploiting morphological similarity across platforms with different scales and hydrodynamic properties. A dynamics model trained on a larger source robot is adapted to a smaller target robot using limited labeled data. An autoencoder-based domain adaptation method learns a shared latent representation that aligns the dynamics of both robots. Experiments on two real robots enable accurate body-frame velocity state estimation on the target platform without labeled target data, supporting efficient cross-robot dynamics transfer.","Efficient Transfer Learning of Robot Dynamic Models Using  \nMorphological Similarity  \nPavlo Kupyn 1 ,2 , Yuya Hamamatsu 1 , Roza Gkliva 1 , Asko Ristolainen 1 , Maarja Kruusmaa 1  \narXiv :2607 .05665v 1 [ cs .RO] 6 Jul 2026  \nAbstract—This study proposes a neural network–based transfer learning framework for modeling the dynamics of soft, fin-actuated underwater robots. We focus on morphologically similar robots that differ in scale and hydrodynamic properties. A model trained on data from a larger robot (source domain) is adapted to a smaller one (target domain) with limited labeled data. To enable label-efficient transfer, we develop an autoencoder-based domain adaptation approach that learns a shared latent representation aligning the dynamics of both robots. Experiments on two real underwater robots show that the proposed method enables accurate state estimation of the body-frame velocities on a target platform without labeled data, highlighting its potential for efficient cross-robot dynamics transfer among morphologically similar platforms.  \nI. INTRODUCTION  \nThe morphologies of the underwater robots vary by application [1] . The limitations of conventional propeller-driven systems [2] motivate bio-inspired robots to use soft fin-based actuators that offer quieter and efficient propulsion. However, accurately modeling the dynamics of such soft actuated systems for state estimation and control remains challenging due to complex nonlinear fluid–structure interactions [3] . Due to this difficulty, most underwater vehicle systems currently rely on state estimation via expensive sensor feedback [4] . To address these challenges, Neural Network (NN) -based modeling and state estimation have shown promise in capturing these complexities by predicting robot motion directly from sensor and control input. However, they often require large labeled datasets for supervised training [5] . Acquiring such data for underwater platforms is costly and difficult due to the requirements of special equipment to obtain ground truth labels.  \nTo address these problems, transfer learning can leverage knowledge from one robot to improve learning on another [6] . In particular, this work explores dynamics transfer between two morphologically similar, fin-actuated underwater robots. Although both share the same actuation principles and kinematics, differences in geometrical scale introduce a domain gap. To bridge this gap, we systematically evaluate several dynamic modeling transfer strategies. First, we evaluated supervised baseline approaches, including neural networks trained independently on each robot’s dataset, as well as a joint model trained on combined dataset. Then, we  \n1The authors are with the Department of Computer Systems, Tallinn University of Technology, Tallinn, Estonia (Pavlo .Kupyn, Yuya .Hamamatsu, Roza .Gkliva, Asko.Ristolainen, Maarja.Kruusmaa)@[taltech.ee](taltech.ee), 2 University of Bonn, Institute of Computer Science, Bonn, Germany  \n© 2026 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses. Accepted at CoDIT 2026 .  \npropose a self-supervised autoencoder framework that learns a shared latent representation aligning both robot domains. An overview of the compared approaches is shown in Fig. 1. By encoding the joint input feature space, the model captures each robot’s dynamics and enables an accurate prediction for the smaller robot even without labeled target data. Accurate dynamic forward modeling serves as a foundation for modelbased control. Improving the transferability of learned dynamics directly contributes to enhancing control in soft robot control applications. We validate the proposed framework on two morphologically similar fin-actuated underwater robots: U-CAT, referred to as the source domain, and Micro-CAT, serving as the target domain.  \nII. RELATED WORKS  \nSoft-actuated underwater robots are difficult to model due to nonlinear fluid–structure intera","cbCais9wPBmXWBVm","https://ap.wps.com/l/cbCais9wPBmXWBVm","pdf",7982841,2,1,6,"English","en",105,"# Introduction\n# Related Works\n# Supervised Baseline Models","[{\"question\":\"What problem does the study address in underwater robot dynamics modeling?\",\"answer\":\"Modeling soft, fin-actuated underwater robot dynamics is difficult due to complex nonlinear fluid–structure interactions and the cost of collecting labeled ground-truth data for state estimation and control.\"},{\"question\":\"How does the proposed method achieve transfer learning between robots?\",\"answer\":\"It adapts a model trained on a larger source robot to a smaller target robot by learning a shared latent representation via an autoencoder-based domain adaptation approach that aligns the dynamics across domains.\"},{\"question\":\"What key experimental outcome is reported for the target robot?\",\"answer\":\"The method produces accurate state estimation of body-frame velocities on the target platform without requiring labeled target data, demonstrating efficient cross-robot dynamics transfer for morphologically similar robots.\"}]",1784198780,15,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"efficient-transfer-learning-of-robot-dynamic-models-using-morphological-similarity","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/efficient-transfer-learning-of-robot-dynamic-models-using-morphological-similarity/84846/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the study address in underwater robot dynamics modeling?","Question",{"text":75,"@type":76},"Modeling soft, fin-actuated underwater robot dynamics is difficult due to complex nonlinear fluid–structure interactions and the cost of collecting labeled ground-truth data for state estimation and control.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed method achieve transfer learning between robots?",{"text":80,"@type":76},"It adapts a model trained on a larger source robot to a smaller target robot by learning a shared latent representation via an autoencoder-based domain adaptation approach that aligns the dynamics across domains.",{"name":82,"@type":73,"acceptedAnswer":83},"What key experimental outcome is reported for the target robot?",{"text":84,"@type":76},"The method produces accurate state estimation of body-frame velocities on the target platform without requiring labeled target data, demonstrating efficient cross-robot dynamics transfer for morphologically similar robots.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]