Automatic Test Generation from Communicating Extended Finite State Machine (CEFSM)-Based Models

• Venue: ISORC '02 Proceedings of the Fifth IEEE International Symposium on Object-Oriented Real-Time Distributed Computing
• Year: 2002

We present a methodology based on communicating extended finite state machines (CEFSM) to generate tests automatically from system specifications. Unlike other model-based methods, it does not require the creation of any new models. Rather, it uses the existing available system models including both data and behavioral information. Our CEFSM-based test generation uses FSMs (finite state machines) to model behavior and events, their extension with variables to model data, and their interaction channels to model communication. Our method generates tests based on a combination of behavior, data, and communication specifications. Another important difference is that our method addresses branching coverage not only for data-related decision coverage, but also behavioral transition coverage. Sophisticated "priority" and "dominator" analysis is applied to generate efficient test cases to increase, as much as possible with as few tests as possible, the branching coverage of the system being tested. The generated tests are given in the XML format which can be input to a test driver with a standard XML parser for automatic test execution. We have implemented our test generation method on a tool called TestGen. We also demonstrate the effectiveness of our method by experimenting with a call-forwarding object-oriented real-time process. Results of our study suggest that our method can generate efficient tests directly from CEFSM-based system models to achieve high branching coverage in an effective way.