Abstract
The system test of Safety Control Units in the vehicle includes the feeding of recorded real vehicle dynamics data which means that the extent of possible maneuvers is very limited. The new maneuverbased testing approach with vehicle dynamics and environment simulation provides the means to simulate almost every imaginable scenario. This freedom also implicates the challenge to find relevant boundary values or characteristic scenarios. Many approaches, like error guessing make use of very unreliable practices in order to find the desired test cases. Evolutionary algorithms such as Differential Evolution offer a more systematic and finite way. The approach offered here combines evolutionary algorithms with vehicle dynamics and environment simulation in order to find desired parameters for simulated test scenarios. Evolutionary algorithms are search algorithms that behave like the processes of selection, recombination and mutation of genes in biology. An initial population, which means an initial set of parameters for a maneuverbased test case, is automatically optimized by recombining, mutating and rerunning the test case until the desired aim, which is a fitness function, is reached. For example a braking maneuver in combination with a steering wheel angle could be optimized in order to find a scenario with maximum lateral acceleration. This would then lead to an especially critical lateral test case.
KEYWORDS – Evolutionary algorithms, maneuverbased testing, safety control unit, test methods, system test