A Test Data Generation Approach for Automotive Software

Since automotive software contains many control flows, symbolic execution is an effective approach to generate test data for it. However, symbolic execution is cost expensive, so it is difficult to apply it directly. Moreover, parameters in automotive software are usually closely related to implement the same function, thus the constraints are dependent on other constraints in the entire path constraint set, which results in traditional optimization techniques, such as constraint independence optimization, could not be used for symbolic execution of automotive software. In this paper, we present a new test data generation approach for automotive software. In our approach, we combine symbolic execution and minimum cut to generate test data for automotive software. We firstly use minimum cut technique to divide the entire path constraint set into two constraint subsets. Then we solve the smaller subset and reuse the solution when solving the entire path constraint set. We believe this approach can not only be faster than solving the entire constraint set directly, but also increase the probability of hitting the cache.