Generating concrete test cases from vehicle data using models obtained from clustering

For testing systems, we need concrete test cases for executing a system under test. Furthermore, such test cases must be relevant to the application domain and cover critical situations a system has to handle. Tests can be repetitive when used for verifying non-functional properties like robustness. This paper introduces an approach using model-based testing for generating test cases from data. The approach relies on models represented by a graph we obtain from data clustering where the clusters correspond to the nodes. We use graph traversal to generate abstract test cases and data sequences from corresponding clusters to deliver concrete tests. Besides outlining the basic foundations of the approach, we discuss results obtained using a well-known driving data set. This use case shows that we can reproduce a test sequence that is reasonably close to the actual behavior of the vehicle stored in the data set.