TY - GEN
T1 - CiThruS2
T2 - IEEE International Intelligent Transportation Systems Conference
AU - Galazka, Emilian
AU - Niemirepo, Teo
AU - Vanne, Jarno
N1 - Funding Information:
This paper is part of the NEWCONTROL project that has received funding within the ECSEL JU in collaboration with the European Union's H2020 Framework Programme (H2020/2014-2020) and National Authorities, under grant agreement 826653.
Publisher Copyright:
© 2021 IEEE.
jufoid=73675
PY - 2021/9/19
Y1 - 2021/9/19
N2 - The automotive and transport sector is undergoing a paradigm shift from manual to highly automated driving. This transition is driven by a proliferation of advanced driver assistance systems (ADAS) that seek to provide vehicle occupants with a safe, efficient, and comfortable driving experience. However, increasing the level of automation makes exhaustive physical testing of ADAS technologies impractical. Therefore, the automotive industry is increasingly turning to virtual simulation platforms to speed up time-to-market. This paper introduces the second version of our open-source See-Through Sight (CiThruS) simulation framework that provides a novel photorealistic virtual environment for vision-based ADAS development. Our 3D urban scene supports realistic traffic infrastructure and driving conditions with a plurality of time-of-day, weather, and lighting effects. Different traffic scenarios can be generated with practically any number of autonomous vehicles and pedestrians that can be made to comply with dedicated traffic regulations. All implemented features have been carefully optimized and the performance of our lightweight simulator exceeds 4K (3840 × 2160) rendering speed of 60 frames per second when run on NVIDIA GTX 1060 graphics card or equivalent consumer-grade hardware. Photorealistic graphics rendering and real-time simulation speed make our proposal suitable for a broad range of applications, including interactive driving simulators, visual traffic data collection, virtual prototyping, and traffic flow management.
AB - The automotive and transport sector is undergoing a paradigm shift from manual to highly automated driving. This transition is driven by a proliferation of advanced driver assistance systems (ADAS) that seek to provide vehicle occupants with a safe, efficient, and comfortable driving experience. However, increasing the level of automation makes exhaustive physical testing of ADAS technologies impractical. Therefore, the automotive industry is increasingly turning to virtual simulation platforms to speed up time-to-market. This paper introduces the second version of our open-source See-Through Sight (CiThruS) simulation framework that provides a novel photorealistic virtual environment for vision-based ADAS development. Our 3D urban scene supports realistic traffic infrastructure and driving conditions with a plurality of time-of-day, weather, and lighting effects. Different traffic scenarios can be generated with practically any number of autonomous vehicles and pedestrians that can be made to comply with dedicated traffic regulations. All implemented features have been carefully optimized and the performance of our lightweight simulator exceeds 4K (3840 × 2160) rendering speed of 60 frames per second when run on NVIDIA GTX 1060 graphics card or equivalent consumer-grade hardware. Photorealistic graphics rendering and real-time simulation speed make our proposal suitable for a broad range of applications, including interactive driving simulators, visual traffic data collection, virtual prototyping, and traffic flow management.
KW - Advanced driver-assistance systems (ADAS)
KW - driving simulation
KW - open-source software
KW - photorealism
KW - traffic imaging
U2 - 10.1109/ITSC48978.2021.9564751
DO - 10.1109/ITSC48978.2021.9564751
M3 - Conference contribution
AN - SCOPUS:85118466522
T3 - IEEE Conference on Intelligent Transportation Systems, Proceedings, ITSC
SP - 3284
EP - 3291
BT - 2021 IEEE International Intelligent Transportation Systems Conference, ITSC 2021
PB - IEEE
Y2 - 19 September 2021 through 22 September 2021
ER -