A new input-output based model coverage paradigm for control blocks

Abstract

Safety critical control systems, especially flight control systems, have failed in trials and actual flights sometimes killing people. Independence is a key word in the verification and validation activity of such systems. Model based testing is used today in all flight control programs. Test cases are generated using qualified automated tools or manually to ensure code and requirements coverage. A new paradigm of coverage metric is developed here for the independent verification. This has been developed for control systems represented as Simulink blocks. The coverage metrics are defined as pairs of cells. One cell has a True/False discrete metric, which indicates if a particular functionality has been covered. The other cell has the continuous metric, which indicates a distance to nominal, or coverage. The efficacy of these metrics is determined by tests using mutant code. Matlab code corresponding to the model is mutated to generate several mutant files. It is shown that the test cases developed using this coverage metric bring out the errors introduced in the mutant files. It was possible to use these metrics to optimize test cases successfully using the Taguchi Design of Experiments methodology. 1 2