prikaz prve stranice dokumenta Simulacija upravljanja autonomnim vozilom
Access restricted to authenticated users
master's thesis
Simulacija upravljanja autonomnim vozilom
2021. urn:nbn:hr:179:265931
Rizvan, Matija
University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Department of Electronics and Computing

Cite this document

Rizvan, M. (2021). Simulacija upravljanja autonomnim vozilom (Master's thesis). Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture. Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:265931

Rizvan, Matija. "Simulacija upravljanja autonomnim vozilom." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2021. https://urn.nsk.hr/urn:nbn:hr:179:265931

Rizvan, Matija. "Simulacija upravljanja autonomnim vozilom." Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2021. https://urn.nsk.hr/urn:nbn:hr:179:265931

Rizvan, M. (2021). 'Simulacija upravljanja autonomnim vozilom', Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 24 April 2024, https://urn.nsk.hr/urn:nbn:hr:179:265931

Rizvan M. Simulacija upravljanja autonomnim vozilom [Master's thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2021 [cited 2024 April 24] Available at: https://urn.nsk.hr/urn:nbn:hr:179:265931

M. Rizvan, "Simulacija upravljanja autonomnim vozilom", Master's thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2021. Available at: https://urn.nsk.hr/urn:nbn:hr:179:265931

Please login to the repository to save this object to your list.
Metadata
TitleSimulacija upravljanja autonomnim vozilom
Title (english)Autonomous vehicle driving simulation
AuthorMatija Rizvan
MentorIvan Slapničar (mentor)
GranterUniversity of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
(Department of Electronics and Computing)
Split
Defense date and country2021-09-27, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Electrical Engineering
Electronics
Abstract
U prvom dijelu radu predstavljene su klasične neuronske mreže, odnosno njihov osnovni građevni element, neuron. Prikazan je biološki živčani sustav, model neurona te arhitektura mreže s osnovnim parametrima kao što su ulazi, izlazi, težine, slojevi i aktivacijske funkcije. U trećem poglavlju uveden je pojam učenja mreže i funkcije gubitka te su detaljno opisani algoritmi gradijentnog spusta i povratnog prostiranja pogreške. Potom, u četvrtom poglavlju predstavljene su konvolucijske neuronske mreže. Prezentira se operacija konvolucije na jednodimenzionalnim i dvodimenzionalnim signalima te se u nastavku daje opis svakog sloja mreže, objašnjavajući ulogu pojedinog sloja u arhitekturi. U praktičnom dijelu rada implementirana je metoda učenja imitacijom koristeći simulator online platforme za učenje Udacity. Opisuje se arhitektura implementacije, sučelje simulatora te princip učenja u trening načinu rada i testiranja u autonomnom načinu rada. Opisan je korišteni kod te metode optimizacije i regularizacije. Na samom kraju rada iznesen je zaključak u kojem se ukazuje na prednosti i nedostatke učenja imitacijom.
Abstract (english)
In the first part of the thesis, classical neural networks are presented, that is, their basic building element, the neuron.The biological nervous system, the model of neurons and the architecture of the network with basic parameters such as inputs, outputs, weights, layers and activation functions are presented. The third chapter introduced the concept of network learning and the loss function, describing the gradient descent algorithm and the backpropragation algorithm. Then, in the fourth chapter, convolutional neural networks were presented. The convolution operation on one and two-dimensional signals is presented within description of each layer of the network, explaining the role of each layer in the architecture. In the practical part of the thesis, the behavioral cloning method was implemented using the simulator of the online learning platform Udacity. The architecture of implementation, the simulator interface and the principle of learning in training mode and testing in autonomous mode are described. Then, the used code and optimization and regularization methods are described. At the very end, a conclusion was presented indicating the advantages and disadvantages of behavioral cloning.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:179:265931
Study programmeTitle: Electronics and Computer Engineering
Study programme type: university
Study level: graduate
Academic / professional title: magistar/magistra inženjer/inženjerka elektronike i računalnog inženjerstva (magistar/magistra inženjer/inženjerka elektronike i računalnog inženjerstva)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to authenticated users
Terms of useLicence In copyright icon
Created on2022-07-05 10:04:56