Application of genetic algorithm for flight system verification and validation

Most complex systems nowadays heavily rely on software, and spacecraft and satellite systems are no exception. Moreover as systems capabilities increase, the corresponding software required to integrate and address system tasks becomes more complex. Hence, in order to guarantee a system's success, testing of the software becomes imperative. Traditionally exhaustive testing of all possible behaviors was conducted. However, given the increased complexity and number of interacting behaviors of current systems, the time required for such thorough testing is prohibitive. As a result many have adopted random testing techniques to achieve sufficient coverage of the test space within a reasonable amount of time. In this paper we propose the use of genetic algorithms (GA) to greatly reduce the number of tests performed, while still maintaining the same level of confidence as current random testing approaches. We present a GA specifically tailored for the systems testing domain. In order to validate our algorithm we used the results from the Dawn test campaign. Preliminary results seem very encouraging, showing that our approach, when searching the worst test cases, outperforms random search , limiting the search to a mere 6 % of the full search domain.