Model-based Systems Engineering to Support Test-driven Development

摘要

Model-based systems engineering (MBSE) is shown to facilitate early test planning and execution. Testing autonomous systems is a challenge due to the near infinite test spaces and lack of autonomy specific measures. This paper identifies and demonstrates an end-to-end process for testing autonomous systems. Systems requirements are derived from the literature for autonomous systems implementing reactive architectures; these metrics were responsiveness, robustness, and perception accuracy. Multiple simulated unmanned aerial systems (UAS) are instantiated from the autonomous systems reference architecture (ASRA); these vehicles are tasked with performing wide area search (WAS). Using design of experiments, the test results are used to create a response surface model to characterize the system. Lastly, desirability was incorporated to determine an optimal system configuration that maximizes area searched, percent correct detected and confirmed, perception accuracy, and minimizes mission time for a given environment.