[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119695-en":3,"doc-seo-119695-105":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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},119695,1099513958607,"Jiven","https://ap-avatar.wpscdn.com/avatar/100002390cf8733938c?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778829742770036399",8,"Research & Report","Road Surface Estimation Using Machine Learning - Thesis","Vehicle motion control systems improve safety and driving comfort, yet they require timely road-condition knowledge to handle poor weather scenarios such as icy roads or heavy rain. Because commercial vehicles typically lack direct friction-sensing hardware, an online road-surface classification approach is needed. This thesis evaluates two distinct machine-learning methods: an LSTM-based artificial neural network and Nadaraya-Watson kernel regression. Results show both methods achieve reasonable accuracy after hyperparameter tuning, with trade-offs in generalization, prediction cost, and interpretability.","Road Surface Estimation Using Machine Learning  \nby  \nBrian Mao  \nA thesis  \npresented to the University of Waterloo in ful􀀌llment of the thesis requirement for the degree of  \nMaster of Mathematics  \nin  \nApplied Mathematics  \nWaterloo, Ontario, Canada, 2023  \n􀀍c Brian Mao 2023  \nAuthor's Declaration  \nI hereby declare that I am the sole author of this thesis. This is a true copy of the thesis, including any required 􀀌nal revisions, as accepted by my examiners.  \nI understand that my thesis may be made electronically available to the public.  \nAbstract  \nVehicle motion control systems are present on commercial vehicles to improve safety and driving comfort. Many of these control systems could be further improved given accurate online information about the road condition to accommodate for driving under poor weather conditions such as icy roads or heavy rain. However, sensors for direct friction measurement are not present on commercial vehicles due to production costs. Hence, it is bene􀀌cial to incorporate an online estimation scheme for road surface classi􀀌cation.  \nThis thesis focuses on investigating two fundamentally di􀀋erent machine learning-based methods for road surface classi􀀌cation. The 􀀌rst is an arti􀀌cial neural network that provides a global function approximation of the underlying dynamics. In particular, Long ShortTerm Memory (LSTM) units are used to capture temporal relationships within the training data and to mitigate the vanishing gradient problem. The second is an instance-based learning method referred to as Nadaraya-Watson Kernel Regression, where local function approximations are generated around the input data.  \nResults indicated that both machine learning-based methods were able to classify road conditions to a reasonable degree of accuracy after tuning associated hyperparameters. However, each method has di􀀋erent bene􀀌ts and drawbacks. The LSTM network model was capable of making accurate predictions on inputs drastically di􀀋erent from data points within the training data set, was generally more accurate on predictions associated with new driving maneuvers, required less storage for implementation, and had relatively short prediction times. Conversely, the Nadaraya-Watson Kernel Regression model was generally more accurate at making predictions on inputs that were very similar to data points within the training data set, did not require any training time to incorporate newly collected data into the model, and generated predictions that were more easily explainable.  \nAcknowledgements  \nFirst and foremost, I'd like to thank my supervisors Professor Kirsten Morris, and Professor Amir Khajepour for their guidance and support. This thesis certainly would not have been possible without them. I'd also like to thank all of my wonderful friends across the Mathematics and Engineering faculties throughout my time at the University of Waterloo. They certainly made the entire experience far more enjoyable than it would have been otherwise. Finally, I'd also like to thank my other friends and family for all of their encouragement and support over the years.  \nTable of Contents  \nList of Figures vii  \nList of Tables xi  \n1 Introduction 1  \n2 Background and Literature Review 4  \n2.1 Vehicle Parameters ............................... 4  \n2.2 Classi􀀌cation of Road Surfaces ......................... 7  \n2.3 Regression Application For Friction Coe􀀎cient Estimations ......... 8  \n2.4 Model-Based Tire-Road Friction Estimation ................. 10  \n2.5 Machine Learning-Based Approaches for Other Vehicle Parameters ..... 11  \n3 Arti􀀌cial Neural Networks 13  \n3.1 Feedforward Neural Networks ......................... 13  \n3.2 Training Arti􀀌cial Neural Networks ...................... 17  \n3.3 Recurrent Neural Networks ........................... 21  \n3.4 Long Short-Term Memory ........................... 24  \n3.5 Friction Estimation on an Inverted Pendulum Using LSTM ......... 26  \n4 Instance-Based Learning 33  \n4.1 K-Neares","cbCaifJsYriwVydv","https://ap.wps.com/l/cbCaifJsYriwVydv","pdf",4700001,1,114,"English","en",105,"# Introduction\n# Background and Literature Review\n## Vehicle Parameters\n## Classification of Road Surfaces\n## Regression Application For Friction Coefficient Estimations\n## Model-Based Tire-Road Friction Estimation\n## Machine Learning-Based Approaches for Other Vehicle Parameters\n# Artificial Neural Networks\n## Feedforward Neural Networks\n## Training Artificial Neural Networks\n## Recurrent Neural Networks\n## Long Short-Term Memory\n## Friction Estimation on an Inverted Pendulum Using LSTM\n# Instance-Based Learning\n## K-Nearest Neighbors\n## Nadaraya-Watson Kernel Regression\n## Implementation of Nadaraya-Watson Kernel Regression\n## Friction Estimation on an Inverted Pendulum using Nadaraya-Watson Kernel Regression\n# Road Surface Classification\n## Feature Selection and Data Sets\n## LSTM Network Architecture and Estimation Results\n## Nadaraya-Watson Kernel Regression Hyperparameter Selection and Estimation Results\n## Comparison of Approaches\n# Conclusions and Future Work","[{\"question\":\"Why is online road-surface estimation needed for vehicle motion control?\",\"answer\":\"Accurate road-condition information helps control systems adapt to poor weather such as icy roads or heavy rain. Commercial vehicles often cannot measure friction directly, so an online estimation scheme is beneficial.\"},{\"question\":\"What are the two machine learning methods studied for road surface classification?\",\"answer\":\"The thesis examines an LSTM-based artificial neural network for global function approximation and a Nadaraya-Watson kernel regression method that builds local approximations around input data.\"},{\"question\":\"How do the LSTM and Nadaraya-Watson kernel regression methods differ in performance and behavior?\",\"answer\":\"Both can classify with reasonable accuracy after tuning. LSTM tends to generalize better to inputs far from training data and often has shorter prediction time, while Nadaraya-Watson is usually more accurate for inputs similar to training samples, requires no training time for new data, and produces more explainable predictions.\"}]","Road Surface Estimation Using Machine Learning - Thesis | PDF",1785725808,287,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"road-surface-estimation-using-machine-learning-thesis","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"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/road-surface-estimation-using-machine-learning-thesis/119695/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",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},"Why is online road-surface estimation needed for vehicle motion control?","Question",{"text":75,"@type":76},"Accurate road-condition information helps control systems adapt to poor weather such as icy roads or heavy rain. Commercial vehicles often cannot measure friction directly, so an online estimation scheme is beneficial.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What are the two machine learning methods studied for road surface classification?",{"text":80,"@type":76},"The thesis examines an LSTM-based artificial neural network for global function approximation and a Nadaraya-Watson kernel regression method that builds local approximations around input data.",{"name":82,"@type":73,"acceptedAnswer":83},"How do the LSTM and Nadaraya-Watson kernel regression methods differ in performance and behavior?",{"text":84,"@type":76},"Both can classify with reasonable accuracy after tuning. LSTM tends to generalize better to inputs far from training data and often has shorter prediction time, while Nadaraya-Watson is usually more accurate for inputs similar to training samples, requires no training time for new data, and produces more explainable predictions.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"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":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":106,"slug":138},19,"General","general"]