A real-time simulation environment architecture for autonomous vehicle design

Yusuf Özçevik; Özgür Solmaz; Eşref Baysal; Mert Ökten

Makale detayları

A real-time simulation environment architecture for autonomous vehicle design

Yazarlar:

Yusuf Özçevik | Özgür Solmaz | Eşref Baysal | Mert Ökten

Yayıncı:

Gazi University

Ana Başlıklar:

Mühendislik

Dergi:

Gazi Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi Vol.38 Issue 3, pp.1867-1878

Belge Tipi:

Makale

Yayın Tarihi:

6/2023

Belge Dili:

Türkçe

Document ID (Org No.):

2001282102

pISSN:

1300-1884

Anahtar kelimeler:

Özet:

 Various approaches proposed for autonomous driving basically involve an image processing and a machine learning process. It is extremely important to use appropriate image processing techniques and a comprehensive data set for the success of these processes. Moreover, in order to present a model that can be applied in the real world, the proposed model must work in real-time. On the other hand, the design, manufacture and testing processes of an autonomous vehicle model result in serious hardware costs. In addition, the design and manufacturing processes need to be repeated if serious hardware changes are required to implement and deploy new approaches. In this context, utilizing a real-time simulation environment can be seen as a suitable approach for a less costly pre-validation of such models. Thus, in this study, a real-time simulation architecture is developed with Unity framework and other necessary components to test an autonomous driving model. In addition, an autonomous driving model that includes a suitable combination of lane tracking and object recognition approaches is proposed, and an autonomous vehicle simulation is created on a track prepared in the proposed simulation environment architecture. Finally, the feasibility of the proposed simulation architecture is tested with the convolutional neural networks based YOLO algorithm versions on a data set in the literature. Considering the findings obtained from the study, it is argued that a vehicle simulation with an autonomous driving model has been successfully tested in the proposed system architecture.

 Various approaches proposed for autonomous driving basically involve an image processing and a machine learning process. It is extremely important to use appropriate image processing techniques and a comprehensive data set for the success of these processes. Moreover, in order to present a model that can be applied in the real world, the proposed model must work in real-time. On the other hand, the design, manufacture and testing processes of an autonomous vehicle model result in serious hardware costs. In addition, the design and manufacturing processes need to be repeated if serious hardware changes are required to implement and deploy new approaches. In this context, utilizing a real-time simulation environment can be seen as a suitable approach for a less costly pre-validation of such models. Thus, in this study, a real-time simulation architecture is developed with Unity framework and other necessary components to test an autonomous driving model. In addition, an autonomous driving model that includes a suitable combination of lane tracking and object recognition approaches is proposed, and an autonomous vehicle simulation is created on a track prepared in the proposed simulation environment architecture. Finally, the feasibility of the proposed simulation architecture is tested with the convolutional neural networks based YOLO algorithm versions on a data set in the literature. Considering the findings obtained from the study, it is argued that a vehicle simulation with an autonomous driving model has been successfully tested in the proposed system architecture.

 Various approaches proposed for autonomous driving basically involve an image processing and a machine learning process. It is extremely important to use appropriate image processing techniques and a comprehensive data set for the success of these processes. Moreover, in order to present a model that can be applied in the real world, the proposed model must work in real-time. On the other hand, the design, manufacture and testing processes of an autonomous vehicle model result in serious hardware costs. In addition, the design and manufacturing processes need to be repeated if serious hardware changes are required to implement and deploy new approaches. In this context, utilizing a real-time simulation environment can be seen as a suitable approach for a less costly pre-validation of such models. Thus, in this study, a real-time simulation architecture is developed with Unity framework and other necessary components to test an autonomous driving model. In addition, an autonomous driving model that includes a suitable combination of lane tracking and object recognition approaches is proposed, and an autonomous vehicle simulation is created on a track prepared in the proposed simulation environment architecture. Finally, the feasibility of the proposed simulation architecture is tested with the convolutional neural networks based YOLO algorithm versions on a data set in the literature. Considering the findings obtained from the study, it is argued that a vehicle simulation with an autonomous driving model has been successfully tested in the proposed system architecture.

 Otonom sürüş için önerilen çeşitli yaklaşımlar temelde bir görüntü işleme ve bir makine öğrenmesi sürecini içermektedir. Bu süreçlerin başarıya ulaşması için uygun görüntü işleme tekniklerinin ve kapsamlı bir veri setinin kullanılması son derece önemlidir. Bununla birlikte, gerçek dünyada uygulanabilir bir model ortaya koymak adına, önerilen modelin gerçek zamanlı çalışması gerekir. Öte yandan, bir otonom araç modelinin tasarlanması, imalatı ve test süreçleri ciddi donanım maliyetleri ile sonuçlanmaktadır. Ayrıca, yeni yaklaşımların geliştirilmesi ve uygulanması için ciddi donanım değişikliklerine ihtiyaç duyulması durumunda, tasarım ve imalat süreçlerinin tekrarlanması gerekmektedir. Bu bağlamda, gerçek zamanlı bir benzetim mimarisinden faydalanmak, modelin daha az maliyetle bir ön doğrulaması için uygun bir yaklaşımdır. Bu yüzden, bu çalışmada, bir otonom sürüş modelini sınamak üzere, Unity kütüphanesi ve gerekli diğer bileşenler ile gerçek zamanlı bir benzetim ortamı mimarisi önerilmektedir. Ayrıca, şerit takip ve nesne tanıma yaklaşımlarının uygun bir bileşimin içeren bir otonom sürüş modeli tanıtılarak, önerilen benzetim mimarisinde hazırlanan bir parkur üzerinde, bir otonom araç benzetimi oluşturulmaktadır. Son olarak, literatürde yer alan bir veri seti üzerinde, evrişimsel sinir ağları tabanlı YOLO algoritması versiyonları ile önerilen benzetim mimarisinin uygulanabilirliği sınanmaktadır. Çalışmadan elde edilen bulgular göz önüne alındığında, otonom sürüş modeline sahip bir araç benzetiminin, önerilen sistem mimarisinde başarılı bir şekilde sınandığı savunulmaktadır.

Makale detayları

A real-time simulation environment architecture for autonomous vehicle design

Yazarlar:

Yusuf Özçevik | Özgür Solmaz | Eşref Baysal | Mert Ökten

Yayıncı:

Gazi University

Ana Başlıklar:

Mühendislik

Dergi:

Gazi Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi Vol.38 Issue 3, pp.1867-1878

Belge Tipi:

Makale

Yayın Tarihi:

6/2023

Belge Dili:

Türkçe

Document ID (Org No.):

2001282102

pISSN:

1300-1884

Autonomous Vehicle

Object Recognition Algorithm

Simulation Environment Architecture

Unity Framework

Benzetim Ortamı Mimarisi;

Nesne Tanıma Algoritması

Otonom Araç

Unity Kütüphanesi

Sayı

Makale detayları

Yazarlar: Yusuf Özçevik | Özgür Solmaz | Eşref Baysal | Mert Ökten

Yayıncı: Gazi University

Ana Başlıklar: Mühendislik

Dergi: Gazi Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi Vol.38 Issue 3, pp.1867-1878

Belge Tipi: Makale

Yayın Tarihi: 6/2023

Belge Dili: Türkçe

Document ID (Org No.): 2001282102

pISSN: 1300-1884

Sayı

Yazarlar:

Yusuf Özçevik | Özgür Solmaz | Eşref Baysal | Mert Ökten

Yayıncı:

Gazi University

Ana Başlıklar:

Mühendislik

Dergi:

Gazi Üniversitesi Mühendislik Mimarlık Fakültesi Dergisi Vol.38 Issue 3, pp.1867-1878

Belge Tipi:

Makale

Yayın Tarihi:

6/2023

Belge Dili:

Türkçe

Document ID (Org No.):

2001282102

pISSN:

1300-1884