[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86572-en":3,"doc-seo-86572-105":30,"detail-sidebar-cat-0-en-105":92},{"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},86572,1099514068365,"Aurelia","https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?_k=1776742907772140068",8,"Research & Report","Learning Residual Kinematic Corrections for Continuous Neural Decoding via Reinforcement Learning","Decoding continuous three-dimensional (3D) motor imagery (MI) with noninvasive EEG-based brain–computer interfaces (BCIs) remains difficult because EEG signals vary and decoding trajectories contain systematic residual errors. A two-stage framework is proposed that applies reinforcement learning (RL) to correct residual kinematics from a CNN–LSTM decoder output. The RL agent is trained offline without direct EEG input, optimizing movement accuracy relative to target trajectories. Offline evaluation uses data from ten participants across ten sessions with 2D and immersive VR feedback, improving correlation and reducing RMSE versus CNN–LSTM alone.","arXiv :2607 . 1 1530v 1 [ cs .AI] 13 Jul 2026  \nLearning Residual Kinematic Corrections for Continuous Neural Decoding via Reinforcement  \nLearning  \nJiamian Li 1 , Niall McShane2 , Attila Korik3 , Naomi du Bois3 , Karl McCreadie2 , Leen Jabban4 , Benjamin Metcalfe4 , Özgür ¸Sim¸sek 1 , Damien Coyle3  \n1Department of Computer Science, University of Bath, Bath, BA2 7AY {jl4668, [os435}@bath.ac.uk](os435}@bath.ac.uk)  \n2Intelligent Systems Research Centre, University of Ulster, BT48 7JL {N .McShane, [k.mccreadie}@ulster.ac.uk](k.mccreadie}@ulster.ac.uk)  \n3Bath Institute for the Augmented Human, University of Bath, Bath, BA2 7AY {ak3825, ndb36, [dhc30}@bath.ac.uk](dhc30}@bath.ac.uk)  \n4Department of Electronic and Electrical Engineering, University of Bath, Bath, BA2 7AY {lj386, [bwm23}@bath.ac.uk](bwm23}@bath.ac.uk)  \nAbstract  \nDecoding continuous three-dimensional (3D) motor imagery (MI) using noninvasive electroencephalography (EEG)-based brain–computer interfaces (BCIs) remains challenging due to signal variability and residual decoding errors. Deep learning architectures such as convolutional neural network–long short-term memory (CNN–LSTM) models can capture spatial and temporal dynamics for continuous kinematic decoding; however, systematic residual errors persist in predicted trajectories. We propose a two-stage decoding framework that applies reinforcement learning (RL) to perform residual kinematic correction on the outputs of a CNN–LSTM decoder (CNN–LSTM–RL) . The RL agent is trained offline without direct EEG input and instead operates on predicted kinematic trajectories to optimize movement accuracy relative to target trajectories. This design enables targeted correction of systematic decoder errors while preserving the primary neural decoding pipeline. The proposed framework was evaluated offline using data from ten participants across ten sessions of online continuous 3D motor imagery, with feedback provided in both 2D and immersive virtual reality (VR) environments.  \nDecoding performance was quantified using Pearson correlation coefficients (r) and Root Mean Square Errors (RMSE) along the x, y, and z axes. Compared to CNN–LSTM applied alone, CNN–LSTM–RL improved the mean correlation from 0.5076 to 0.7181 (p = 0 .0005) in 2D and from 0.6420 to 0.7780 (p = 0 .0059) in VR, with relative gains of 41 .5% and 21 .2%, respectively. Correspondingly, RMSE was reduced from 0.0890 to 0.0532 (2D, p \u003C 0.0001) and from 0.0714 to 0.0441 (VR, p \u003C 0.0001), representing relative reductions of 40 .2% and 38.2% . These findings demonstrate that this scalable framework enhances 3D BCI MI decoding by correcting kinematic errors via offline residual RL without extra neural data, advancing neurorehabilitation, prosthetics, and virtual interaction.  \nPreprint.  \n1 Introduction  \nBrain-computer interfaces (BCIs) have demonstrated remarkable success across a wide range of domains, including cognitive and physical rehabilitation [Millán et al., 2010][Prasad et al., 2010][Cervera et al., 2018], augmentative communication [Willett et al., 2021][Willett et al., 2023][Anumanchipalli et al., 2019], assessing consciousness [Du Bois et al., 2026], entertainment [Marshall et al., 2013][van de Laar et al., 2013], and robotic control [Hochberg et al., 2012][Ajiboye et al., 2017][Edelman et al., 2019], by establishing a direct signaling pathway between neural activity and external devices. Non-invasive BCIs have become increasingly popular in the BCI field, as they acquire neural signals, which are subsequently decoded in real-time to execute user-intended commands, via sensors mounted on the human scalp and mitigate the clinical risks associated with invasive approaches that require the surgical implantation of intracortical electrodes, despite the latter offering higher resolution neural interfacing [Lebedev and Nicolelis, 2006] . Electroencephalography (EEG) remains a predominant non-invasive modality due to its superior temporal resolution and c","cbCaiemwWGOXsD8l","https://ap.wps.com/l/cbCaiemwWGOXsD8l","pdf",3217608,5,1,14,"English","en",105,"# Introduction\n## Brain–computer interfaces and EEG decoding\n## Deep learning for continuous 3D trajectory decoding\n## Reinforcement learning for BCI optimization","[{\"question\":\"What problem does the paper address in continuous 3D BCI decoding?\",\"answer\":\"Systematic residual errors remain in predicted trajectories when decoding continuous 3D motor imagery from EEG, due to signal variability and decoder bias.\"},{\"question\":\"How does the proposed CNN–LSTM–RL framework correct kinematic decoding errors?\",\"answer\":\"It uses reinforcement learning as a residual kinematic correction stage on top of a CNN–LSTM decoder, optimizing predicted trajectories toward target movement accuracy.\"},{\"question\":\"How was the framework evaluated and what improvements were reported?\",\"answer\":\"The framework was evaluated offline using ten participants across ten sessions of online continuous 3D motor imagery with 2D and immersive VR feedback, improving mean Pearson correlation and reducing RMSE compared with CNN–LSTM alone.\"}]",1784212710,35,{"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":87,"head_meta":89,"extra_data":91,"updated_unix":28},"learning-residual-kinematic-corrections-for-continuous-neural-decoding-via-reinforcement-learning","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/learning-residual-kinematic-corrections-for-continuous-neural-decoding-via-reinforcement-learning/86572/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What problem does the paper address in continuous 3D BCI decoding?","Question",{"text":76,"@type":77},"Systematic residual errors remain in predicted trajectories when decoding continuous 3D motor imagery from EEG, due to signal variability and decoder bias.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How does the proposed CNN–LSTM–RL framework correct kinematic decoding errors?",{"text":81,"@type":77},"It uses reinforcement learning as a residual kinematic correction stage on top of a CNN–LSTM decoder, optimizing predicted trajectories toward target movement accuracy.",{"name":83,"@type":74,"acceptedAnswer":84},"How was the framework evaluated and what improvements were reported?",{"text":85,"@type":77},"The framework was evaluated offline using ten participants across ten sessions of online continuous 3D motor imagery with 2D and immersive VR feedback, improving mean Pearson correlation and reducing RMSE compared with CNN–LSTM alone.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":20,"slug":138},19,"General","general"]