A hybrid search and model-based approach for testing the self-adaptive unmanned aircraft system software

In recent years, there has been a significant increase in the deployment of unmanned aircraft systems (UAS) in critical missions like search and rescue, surveillance, and environmental monitoring. During a mission, the UAS experiences changes, referred to as interruptions, requiring self-adaptation, i.e., an adjustment in its behavior at run-time. This adaptation is crucial due to the mission’s inherent critical nature involving engagement with humans, structures, and neighboring unmanned aerial vehicles (UAVs). Testing the application software that defines its mission and behavior is mandatory to ensure the accurate adaptation of its behavior. For this, the primary challenge involves the flight of a UAS, followed by the identification of test cases ensuring the execution of necessary self-adaptive behaviors during a UAS flight. The current industrial practice of testing self-adaptive behaviors in UAS involves manual testing, a time-consuming method that restricts the execution to a limited set of test cases. To address this problem, we propose a hybrid approach to test the self-adaptive behavior of UAS application software. As part of the approach, we propose a modeling methodology to capture the application requirements for the UAS mission and the self-adaptive behavior. We then use the developed models and a search-based algorithm to automate the generation and execution of test cases. We have created a prototype tool to facilitate the automation of testing activities. The work is conducted in collaboration with an industrial partner and demonstrated through a case study of UAS formation flight application software. We have effectively modeled the case study concepts with the proposed modeling methodology. Employing our testing approach, we have detected ten unique faults within the formation flight application software. Additionally, statistical analysis indicates that the proposed approach outperforms the baseline random search in fault detection using a genetic algorithm.