Efficient Statistical Validation of Autonomous Driving Systems
Today’s automotive vehicles are often equipped with powerful data processing systems for driver assistance and/or autonomous driving. To meet the rigorous safety standard, ensuring extremely small failure probability over all possible operation conditions is one of the critical tasks for an autonomous driving system. In this chapter, we describe a novel methodology to validate vision-based autonomous driving systems over different circuit corners with consideration of temperature variation and circuit aging. Our approach seamlessly integrates the image data recorded under nominal conditions with comprehensive statistical circuit models to synthetically generate the critical corner cases for which an autonomous driving system is likely to fail. As such, a given automotive system can be robustly validated for these worst-case scenarios that cannot be easily captured by physical experiments. To efficiently estimate the rare failure rate of an autonomous system, we further propose a novel Subset Sampling (SUS) algorithm. In particular, a Markov Chain Monte Carlo algorithm based on graph mapping is developed to accurately estimate the rare failure rate with a minimal amount of test data, thereby minimizing the validation cost. Our numerical experiments show that SUS achieves 15.2× runtime speed-up over the conventional brute-force Monte Carlo method.
