Simulation-based validation and defect localization for evolving, semi-formal requirements models

When requirements models are developed in an iterative and evolutionary way, requirements validation becomes a major problem. In order to detect and fix problems early, the specification should be validated as early as possible, and should also be revalidated after each evolutionary step. In this paper we show how the ideas of continuous integration and automatic regression testing in the field of coding can be adapted for simulation-based, automatic revalidation of requirements models after each incremental step. While the basic idea is fairly obvious, we are confronted with a major obstacle: requirements models under development are incomplete and semi-formal most of the time, while classic simulation approaches require complete, formal models. We present how we can simulate incomplete, semi-formal models by interactively recording missing behavior or functionality. However regression simulations must run automatically and do not permit interactivity. We therefore have developed a technique where the simulation engine automatically resorts to the interactively recorded behavior in those cases where it does not get enough information from the model during a regression simulation run. Finally, we demonstrate how the information gained from model evolution and regression simulation can be exploited for locating defects in the model.