Advancing Verification and Validation for ADAS and ADS

Emerging automotive technologies like advanced driver assistance systems (ADAS) and automated driving systems (ADS) hold promise for improving safety for the traveling public; however, effective verification and validation (V&V) of these systems has proven to be challenging. Traditional testing methodologies may serve in limited cases for systems exhibiting low levels of automation, but recent studies show that these systems that have been brought to market perform poorly in practice. Further, these traditional methods do not serve for testing systems with high levels of automation where a human driver simply serves as a fallback ready user or is out of the loop altogether. New V&V methods are required to assess whether these systems can perform their intended functions in their intended operating environments, and to assess whether they can do so safely across the expansive and variable operating space. This paper presents an overview of ADAS and ADS challenges and novel approaches to address these challenges. It will dive into scenario-based testing approaches spread across testing methods like modeling and simulation (M&S), controlled track testing, and open-road testing, and will outline new methods and tools to sample the test space to compose test cases in a way that addresses the complexities and tradeoffs of testing coverage and efficiency. It will also dive into new methods tools for ADAS and ADS performance measurement and metrics generation. A proposed independent "ground truth" system will be described that can collect real-time testing or operational data for, among other things, conducting ADAS and ADS "driver's license tests."