Validation of Control Software Specification Using Design Interests Extraction and Model Checking 2012-01-0960

Automotive control systems such as powertrain control interact with the open physical environment, and from this nature, expensive prototyping is indispensable to capture a deep understanding of the system requirements and to develop the corresponding control software. Model-based development (MBD) has been promoted to improve productivity by virtual prototyping. Even with MBD, systematic validation of the software specification remains as a major challenge and it still depends heavily on individual engineers' skill and knowledge. Though the introduction of graphical software modeling improved the situation, it requires much time to identify the primal functions, so-called “design interests”, from a large complex model where irrelevant components are mixed with, and to validate it properly. In addition, since software models have to coexist with legacy assets during the transition to MBD, the difficulty mentioned above should be overcome in terms of not only models, but also legacy C codes. As a remedy to this problem, we propose an approach based on design interests extraction from software model and/or legacy code. Design interests are mechanically extracted in the form of functional models and model checking is applied to exercise them. Currently, with our first tool implementation, a kind of dataflow graph is extracted from C codes as a functional model, and model checking is applied to prove specified properties on the dataflow graph and to generate execution traces which stimulate the specific parts of the design in a pinpoint manner. Such conditions are derived from the functional model.