Timeliness in Autonomous Driving: Hype or Reality?

Tianze Wu, Weisong Shi

Abstract: Autonomous driving (AD) has witnessed rapid advancements in recent years, with numerous high-performance algorithms emerging for various subtasks. However, the underlying vehicle computing systems that support these algorithms remain unfamiliar to most researchers. In reality, most current AD solutions are still limited to low-speed scenarios, making successfully deploying these solutions in the real world a significant challenge. This paper comprehensively analyzes AD systems from the two aspects of computing hardware and software solutions and further discusses the challenges faced by defining and guaranteeing the real-time performance of AD. By analyzing real AD application scenarios, we abstract a typical software and hardware architecture for autonomous vehicles and want this article to serve as a starting point to study and explore the timeliness of autonomous driving.