Abstract:

While major advances have been made in the R&D and commercialization of autonomous driving (AD) in the past decade, there still exists significant challenges in the large-scale commercial deployment of AD in complex open-road scenarios, such as longtail perception problem and limited operational design domain (ODD). Information from vehicles, traffic and the underlying infrastructure (V2X) can be used to enhance the overall system safety and accelerate the deployment, with integration of multi-scaled, multi-dimensional diverse sources. This integration would enable cooperated perception, decision-making, and control, expanding single-vehicle intelligence’s capability boundaries. By using this combined knowledge, some of the obstacles encountered in the commerciliazation of autonomous driving can be addressed. This paper introduces a unified framework for autonomous driving known as Vehicle-Infrastructure-Cloud Autonomous Driving (VICAD). VICAD combines the diverse collaborative deployment strategies related to vehicles, infrastructure, and the cloud with autonomous driving algorithms via an integrated framework. Simulations and evaluations are conducted to evaluate the performance of VICAD system, and evaluation results are then feedback as input of the VICAD system. This iterative process enables the continuous optimization of collaborative deployment strategies and autonomous driving algorithms, thereby enhancing the capabilities of autonomous driving. Moreover, this paper describes the key role of VICAD in fostering the large-scale commercial deployment of autonomous driving with practical cases and industrial applications, and concludes with suggestions for further VICAD development.

Key words: autonomous driving(AD), vehicle-infrastructure cooperation, safety of the intended functionality(SOTIF), Vehilcle-Infrastructure-Cloud Autonomous Driving(VICAD), operational design domain(ODD)