Multi-objective Search for Effective Testing of Cyber-Physical Systems.

We propose a multi-objective strategy for finding effective inputs for fault detection in Cyber Physical Systems (CPSs). The main goal is to provide input signals for a system in such a way that they maximise the distance between the system's output and an ideal target, thus leading the system towards a fault; this is based on Genetic Algorithm and Simulated Annealing heuristics. Additionally, we take into consideration the discrete locations (of hybrid system models) and a notion of input diversity to increase coverage. We implement our strategy and present an empirical analysis to estimate its effectiveness.