Design Verification and Validation for Reliable Safety-Critical Autonomous Control Systems

Abstract

Providing guarantees on the system behavior is mandatory for safety-critical autonomous vehicles. Among these guarantees, proving the fulfillment of real-time constraints and reliability requirements on the system is a key issue, as their violation could result in unexpected and unsafe behaviors. The violation may come from the complicated interaction between software and hardware modules, or transient hardware faults. AUTOSAR, the most popular industrial standard in the automotive domain, provides an open standardized architecture for software development, where an application can be deployed on multiple electronic control units (ECUs). We present a verification and validation method for the design of such safety-critical autonomous control systems that could tolerate transient faults. The embedded implementation of an AUTOSAR model is transformed into a three-layer system model in timed automata, so that system behavior can be evaluated and checked with hard real-time constraints and the implementing architecture. We demonstrate the feasibility of the method with a simplified controller developed for the autonomous vehicles.