Stefano Percivati
Virtual validation toolchain for automotive in-cabin sensing.
Rel. Massimo Violante. Politecnico di Torino, Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica), 2024
|
PDF (Tesi_di_laurea)
- Tesi
Licenza: Creative Commons Attribution Non-commercial No Derivatives. Download (45MB) | Preview |
Abstract: |
The significant reduction in fatal accidents and serious injuries on roads, made possible by advances in vehicle safety, represents a crucial goal for the European Union. In 2018, the EU declared its intention to reduce such accidents by 50% by 2030, with the ambitious target of "zero fatalities" by 2050. The integration of Advanced Driver Assistance Systems (ADAS) plays a fundamental role in achieving this goal. The project aims to develop a validation framework compliant with EU regulations for driver and occupant monitoring systems. Leveraging the powerful capabilities of Unreal Engine, the framework allows for testing and training algorithms in an extremely realistic virtual environment, reducing the need for extensive field testing. The project pays particular attention to creating a wide range of avatars, aiming to reflect human diversity as accurately as possible. To make the user experience more authentic, an accurate reproduction of avatar facial movements has been implemented, using prerecorded videos of real people. This approach has been crucial in ensuring that avatars interact consistently and realistically with the surrounding virtual environment, thus ensuring a sense of immersion for users. A crucial aspect lies in integrating avatars into a coherent environmental context. This required meticulous data collection to enable detailed control and precise customization of avatar actions. With the introduction of the eye gazing feature, the user experience has been further enriched, allowing them to accurately track the avatars' gaze and thereby increasing interactivity. To test camera functionality, a virtual bench test has been developed, providing an alternative and accurate method for simulating cameras in a virtual environment. This allows for verifying the proper operation of cameras under controlled conditions, without the need to transfer data and images to the camera control system. Finally, a random scene generation has been implemented in the project during the validation phase, allowing for exploration of a wide range of situations and scenarios, ensuring a comprehensive and accurate assessment of system performance in different contexts. |
---|---|
Relators: | Massimo Violante |
Academic year: | 2023/24 |
Publication type: | Electronic |
Number of Pages: | 83 |
Subjects: | |
Corso di laurea: | Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica) |
Classe di laurea: | New organization > Master science > LM-25 - AUTOMATION ENGINEERING |
Aziende collaboratrici: | SANTER Reply S.p.a. |
URI: | http://webthesis.biblio.polito.it/id/eprint/31009 |
Modify record (reserved for operators) |