Robustness-guided temporal logic testing and verification for Stochastic Cyber-Physical Systems

We present a framework for automatic specification-guided testing for Stochastic Cyber-Physical Systems (SCPS). The framework utilizes the theory of robustness of Metric Temporal Logic (MTL) specifications to quantify how robustly an SCPS satisfies a specification in MTL. The goal of the testing framework is to detect system operating conditions that cause the system to exhibit the worst expected specification robustness. The resulting expected robustness minimization problem is solved using Markov chain Monte Carlo algorithms. This also allows us to use finite-time guarantees, which quantify the quality of the solution after a finite number of simulations. In a Model-Based Design (MBD) process, our framework can be combined with Statistical Model Checking (SMC). Finally, we present a case study on a high fidelity engine model where the goal is to verify the air-to-fuel ratio problem.