Modeling for automated test generation - A comparison

In this contribution, we compare and analyze different methodologies of modeling for test generation. As an example, we use an industrial requirement specification given in natural language, which describes a safety function in a hybrid car. We model these requirements with three different paradigms and languages: as the specification imposes several timing constraints, we choose abstract State Machines, Timed Automata and UML2 State Machines to formalize the given requirements. From these models, we employ different tools for generating test cases, and compare the resulting test suites with respect to coverage and fault detection capabilities. We discuss the process of designing the models and the implications for professional software testing.