Detecting design flaws in control systems using optimisation methods

Complicated control systems are used in many safety-critical applications, such as in cars and airplanes. Due to the nature of these systems, verification can be very difficult to do analytically or algorithmically. The only feasible analysis and verification method is often simulation. The generation of good test cases that can expose flaws in the controller design is therefore of great importance. In this paper we investigate the use of optimisation methods for finding such test cases automatically. For this purpose we give a language to express assertions in control systems, as well as a translation of the assertions to a form suitable for optimisation. We also discuss different ways to generate the input signals for the systems to maximise performance of the optimisation. To evaluate if optimisation is a feasible approach, we provide a case study demonstrating that optimisation methods are beneficial for investigating properties of control system designs.