A Survey of Driving Risk Assessment for Autonomous Vehicles
Summary:
The existing driving risk assessment methods are divided into three categories, including single target-oriented, reachable set-based and potential field theory-based assessment methods. Five evaluation dimensions are proposed, including real-time calculation, timeliness of results, application feasibility, content adequacy, and scene versatility, and a comprehensive comparison of evaluation methods is carried out to reveal their characteristics and application. The problems faced by the risk assessment of autonomous driving and the future development trend are analyzed and forecasted.
With the promotion and popularization of intelligent driving, it is expected that functional safety will be a key technology direction that needs breakthroughs in the automotive field in the next 3-5 years. In order to provide you with more targeted learning materials and technical exchange space, SASETECH will launch the Expected Functional Safety Month in May, continue to share the content related to expected functional safety, and work together with you to learn and understand issues related to expected functional safety. SASETECH anticipates a more in-depth exploration in the Functional Safety Tribe.
Abstract: Driving risk assessment is crucial to the safe operation of automatic driving systems. The existing driving risk assessment methods are divided into three categories, including single target-oriented, reachable set-based and potential field theory-based assessment methods. Five evaluation dimensions are proposed, including real-time calculation, timeliness of results, application feasibility, content adequacy, and scene versatility, and a comprehensive comparison of evaluation methods is carried out to reveal their characteristics and application. The problems faced by the risk assessment of autonomous driving and the future development trend are analyzed and forecasted.
Key words: autonomous vehicle; risk assessment; driving risk; scene; review
Safety is the primary condition for autonomous driving systems to operate on the road. In recent years, with the improvement of the level of driving automation, the scenarios encountered by self-driving cars during driving are becoming more and more complex, and the driving tasks that need to be performed are also becoming more and more diverse. Complex scenarios and diverse driving tasks pose enormous safety challenges to autonomous driving systems. As an important means to ensure and test the safe operation of automatic driving system, driving risk assessment has always been the focus of research in the field of automatic driving.