Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks
The autonomous vehicle (AVs) market is expanding at a rapid pace due to the advancement of information, communication, and sensor technology applications, offering a broad range of opportunities in terms of energy efficiency and addressing climate change concerns and safety. With regard to this last...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-09-01
|
| Series: | Vehicles |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2624-8921/3/3/36 |
| _version_ | 1797516969838116864 |
|---|---|
| author | Aditya Dixit Ramesh Kumar Chidambaram Zaheer Allam |
| author_facet | Aditya Dixit Ramesh Kumar Chidambaram Zaheer Allam |
| author_sort | Aditya Dixit |
| collection | DOAJ |
| description | The autonomous vehicle (AVs) market is expanding at a rapid pace due to the advancement of information, communication, and sensor technology applications, offering a broad range of opportunities in terms of energy efficiency and addressing climate change concerns and safety. With regard to this last point, the rate of reduction in accidents is considerable when switching safety control tasks to machines from humans, which can be noted as having significantly slower response rates. This paper explores this thematic by focusing on the safety of AVs by thorough analysis of previously collected AV crash statistics and further discusses possible solutions for achieving increased autonomous vehicle safety. To achieve this, this technical paper develops a dynamic run-time safe assessment system, using the standard autonomous drive system (ADS), which is developed and simulated in case studies further in the paper. OpenCV methods for lane detection are developed and applied as robust control frameworks, which introduces the factor of vehicle crash predictability for the ego vehicle. The developed system is made to predict possible crashes by using a combination of machine learning and neural network methods, providing useful information for response mechanisms in risk scenarios. In addition, this paper explores the operational design domain (ODD) of the AV’s system and provides possible solutions to extend the domain in order to render vehicle operationality, even in safe mode. Additionally, three case studies are explored to supplement a discussion on the implementation of algorithms aimed at increasing curved lane detection ability and introducing trajectory predictability of neighbouring vehicles for an ego vehicle, resulting in lower collisions and increasing the safety of the AV overall. This paper thus explores the technical development of autonomous vehicles and is aimed at researchers and practitioners engaging in the conceptualisation, design, and implementation of safer AV systems focusing on lane detection and expanding AV safe state domains and vehicle trajectory predictability. |
| first_indexed | 2024-03-10T07:08:17Z |
| format | Article |
| id | doaj.art-0c39f0a4e4ee434f9956ac4bf2a6d337 |
| institution | Directory Open Access Journal |
| issn | 2624-8921 |
| language | English |
| last_indexed | 2024-03-10T07:08:17Z |
| publishDate | 2021-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Vehicles |
| spelling | doaj.art-0c39f0a4e4ee434f9956ac4bf2a6d3372023-11-22T15:35:37ZengMDPI AGVehicles2624-89212021-09-013359561710.3390/vehicles3030036Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural NetworksAditya Dixit0Ramesh Kumar Chidambaram1Zaheer Allam2Center of Excellence for Autonomous Vehicle Research, Department of Mechanical Engineering, Vellore Institute of Technology, Vellore 632014, Tamil Nadu, IndiaAutomotive Research Center, Department of Mechanical Engineering, Vellore Institute of Technology, Vellore 632014, Tamil Nadu, IndiaEntrepreneuriat Territoire Innovation (ETI), Groupe de Recherche en Gestion des Organisations (GREGOR), IAE Paris—Sorbonne Business School, Université Paris 1 Panthéon-Sorbonne, 75013 Paris, FranceThe autonomous vehicle (AVs) market is expanding at a rapid pace due to the advancement of information, communication, and sensor technology applications, offering a broad range of opportunities in terms of energy efficiency and addressing climate change concerns and safety. With regard to this last point, the rate of reduction in accidents is considerable when switching safety control tasks to machines from humans, which can be noted as having significantly slower response rates. This paper explores this thematic by focusing on the safety of AVs by thorough analysis of previously collected AV crash statistics and further discusses possible solutions for achieving increased autonomous vehicle safety. To achieve this, this technical paper develops a dynamic run-time safe assessment system, using the standard autonomous drive system (ADS), which is developed and simulated in case studies further in the paper. OpenCV methods for lane detection are developed and applied as robust control frameworks, which introduces the factor of vehicle crash predictability for the ego vehicle. The developed system is made to predict possible crashes by using a combination of machine learning and neural network methods, providing useful information for response mechanisms in risk scenarios. In addition, this paper explores the operational design domain (ODD) of the AV’s system and provides possible solutions to extend the domain in order to render vehicle operationality, even in safe mode. Additionally, three case studies are explored to supplement a discussion on the implementation of algorithms aimed at increasing curved lane detection ability and introducing trajectory predictability of neighbouring vehicles for an ego vehicle, resulting in lower collisions and increasing the safety of the AV overall. This paper thus explores the technical development of autonomous vehicles and is aimed at researchers and practitioners engaging in the conceptualisation, design, and implementation of safer AV systems focusing on lane detection and expanding AV safe state domains and vehicle trajectory predictability.https://www.mdpi.com/2624-8921/3/3/36autonomous vehicleelectric vehicleslane detectionV2X communicationoperational design domainssafety and risk assessment |
| spellingShingle | Aditya Dixit Ramesh Kumar Chidambaram Zaheer Allam Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks Vehicles autonomous vehicle electric vehicles lane detection V2X communication operational design domains safety and risk assessment |
| title | Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks |
| title_full | Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks |
| title_fullStr | Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks |
| title_full_unstemmed | Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks |
| title_short | Safety and Risk Analysis of Autonomous Vehicles Using Computer Vision and Neural Networks |
| title_sort | safety and risk analysis of autonomous vehicles using computer vision and neural networks |
| topic | autonomous vehicle electric vehicles lane detection V2X communication operational design domains safety and risk assessment |
| url | https://www.mdpi.com/2624-8921/3/3/36 |
| work_keys_str_mv | AT adityadixit safetyandriskanalysisofautonomousvehiclesusingcomputervisionandneuralnetworks AT rameshkumarchidambaram safetyandriskanalysisofautonomousvehiclesusingcomputervisionandneuralnetworks AT zaheerallam safetyandriskanalysisofautonomousvehiclesusingcomputervisionandneuralnetworks |