Statistical Model Checking for Safety Critical Hybrid Systems: An Empirical Evaluation

As more computing systems are utilized in various areas of our society, the reliability of computing systems becomes a significant issue. However, as the complexity of computing systems increases, conventional verification and validation techniques such as testing and model checking have limitations to assess reliability of complex safety critical systems. Such systems often control highly complex continuous dynamics to interact with physical environments. To assure the reliability of safety critical hybrid systems, statistical model checking (SMC) techniques have been proposed. SMC techniques approximately compute probabilities for a target system to satisfy given requirements based on randomly sampled execution traces. In this paper, we empirically evaluated four state-ofthe-art SMC techniques on a fault-tolerant fuel control system in the automobile domain. Through the experiments, we could demonstrate that SMC is practically useful to assure the reliability of a safety critical hybrid system and we compared pros and cons of the four different SMC techniques.