Comprehensive Testing Methodologies and Metrics for Reliable Autonomous Driving

Autonomous driving has markedly improved road safety in the last decade, but concerns persist about its reliability. The shift from testing conventional vehicles to autonomous ones marks a fundamental change in testing approaches. Safeguarding autonomous driving system (ADS) safety is a complex challenge, demanding a meticulous examination of concepts and testing procedures. Despite the recent proliferation of various ADS testing methods, there's a lack of clarity regarding a comprehensive list of these methods and the metrics used for performance assessment. To address this gap, a comprehensive review of the state-of-the-art ADS testing methodologies and the crucial metrics employed within each is presented. It recommends five testing methods and introduces five types of metrics to expedite the development of autonomous vehicles. Additionally, it unravels the complexities of testing autonomous vehicles, pinpointing both scenario creation hurdles and technical limitations, before charting a course towards reliable deployment through targeted solutions like AI-powered tooling and data labeling automation. This paper not only clarifies the current testing landscape but also empowers developers with a practical roadmap for accelerating reliable ADS deployment on the road.