176

We observe an inefficient execution order of tests in the way that for a given set of test cases, a significant number of test steps occur in more than one test case. During test executions these duplicates increase the test load while providing none or limited additional test results. Removing such test steps interrupts the initial execution chain of steps inside a test case. To solve this issue, we propose a test case synthesis. After removing redundant steps, all (non-redundant) test steps are rearranged into a new set of test cases. This is achieved by using a clustering technique to group similar test steps into new test cases. A Path finding algorithm is used to find an optimized test step execution order for each test case. By applying this method in a case study at Mercedes-Benz Passenger Car Development, we observe a test load reduction of 15% due to removing redundant test steps and an additional reduction of at least 3% for rearranging test steps. This totals up to at least 18% overall test load reduction for the proposed method including removing redundant elements. We see this as strong indication for the usefulness of our approach.

Automotive systems are constantly increasing in complexity and size. Beside the increase of requirements specifications and related test specification due to new systems and higher system interaction, we observe an increase of redundant specifications. As the predominant specification language (both for requirements and test cases) is still natural text, it is not easy to detect these redundancies. In principle, to detect these redundancies, each statement has to be compared to all others. This proves to be difficult because of number and informal expression of statements. In this paper we propose a solution to the problem of detecting redundant specification and test statements described in structured natural language. We propose a formalization process for requirements specification and test statements, allowing us to detect redundant statements and thus reduce the efforts for specification and validation. Specification Pattern Systems and Linear Temporal Logic provide the base for our process. We did evaluate the method in the context of Mercedes-Benz Passenger Car Development. The results show that for the investigated sample set of test statements, we could detect about 30% of test steps as redundant. This indicates the savings potential of our approach.