Computer-aided design for safe autonomous vehicles

摘要: This paper details the design of an autonomous vehicle CAD toolchain, which captures formal descriptions driving scenarios in order to develop a safety case for (AV). Rather than focus on particular component AV, like adaptive cruise control, toolchain models end-to-end dynamics AV way suitable testing and verification. First, domain-specific language capable describing that occur day-to-day operation is defined. The allows description composition traffic participants (e.g., other vehicles control devices), specification correctness requirements. A scenario described this executable can be processed by specification-guided automated test generator (bug hunting), exhaustive reachability tool. user exploit integrate strengths both reachability, not possible when each run alone. Finally, given execution violates requirements, visualization tool display counter-example generate labeled sensor data. effectiveness approach demonstrated three drawn from collection 36 account over 95% accidents nationwide. These studies demonstrate robustness-guided verification heuristics reduce analysis time, counterexample identifying controller bugs discrete decision logic low-level analog (continuous) dynamics, identification modeling errors behaviors) lead unrealistic environment behavior.