Automatic test-data generation for testing simulink models

Software testing is costly, labour-intensive, and time-consuming. For most practical systems it will not be possible to perform ‘exhaustive testing’. Test sets must be effective (i.e. they reveal faults) but also easily generated (i.e. the process of generation must be efficient). We generally aim to develop small sets with high fault detection ability. Systematically generating effective test-data is one of the most interesting and practically relevant topics in the testing domain. Modern testing requires faults to be discovered at the earliest possible stage, i.e. specification or architecture design stage rather than the coding stage, because the cost of fixing an error increases with the time between its introduction and detection. We need to generate test cases to exercise our high-level models. Again, as with all testing, we wish to develop effective test sets, and to do so efficiently. One means of capturing high-level behaviour of systems is provided by the Matlab/Simulink toolset. Matlab/Simulink is popularly used in embedded systems engineering as an architectural-level design notation. Engineers like using it, finding it intuitively appealing. In this report we show how certain techniques, taken largely from automated code-level testing, can be adapted for Matlab/Simulink models and applied to generate test sequences that satisfy identified testing aims. In our work, these aims are mainly concerned with analogues of code-level structural and fault-based coverage criteria. We describe the functionality of a toolset developed to automatically generate effective and efficient test sets for the architectural models of interest. We describe also the techniques applied in developing the toolset. Preliminary experimental results show that the toolset can facilitate automatic test-data generation for architectural level models to a certain extent.