An empirical study of requirements-based test generation on an automobile control system

Requirements-based test generation techniques have been widely used in industry to help practitioners generate appropriate test cases to ensure that their software systems behave according to customer expectations. These tests provide a solid foundation for functional testing -- a critical, essential step that must be performed properly for quality assurance of any software system. In this paper, we present a study, in collaboration with Hyundai Motor Company, on a real-life industrial software system used to control the driver's demand torque of an automobile. Since any hazardous incident caused by this mechanism may result in significant property loss or even fatalities, its safe and reliable operation becomes absolutely imperative. To overcome this problem, we propose a framework including different requirements-based test generation techniques such as equivalence class partitioning (ECP), boundary value analysis (BVA), a choice relation framework, and predicate testing-based BOR, BRO, and BRE strategies. In addition, a tool named C-Set has been developed to support the BOR, BRO and BRE-based automatic test generation. Results from our study indicate that weaknesses in any of the above techniques can be mitigated by exploiting the strengths of the other techniques. As a result, a set of high quality test cases for the software being studied can be efficiently generated to help Hyundai engineers improve their productivity.