Towards a Metamorphic Testing Architecture for Software-Defined Drone Systems

Safety-critical systems, such as drones, are vulnerable to effects of uncertainty, where uncertainty can manifest as adverse weather conditions, unexpected human interactions, and misconfigured system settings. Moreover, it is humanly infeasible to test for all possible combinations of conditions that a system may experience through its lifetime during design. Additionally, drones are difficult to exhaustively test prior to deployment as real-world validation is costly in terms of time and equipment. Search-based testing is an approach for discovering new situations that a system may experience, however such tests can suffer from automatically inferring incorrect expected values/outcomes without domain knowledge (i.e., the oracle problem). As such, we propose a metamorphic testing framework for software-defined drone systems that uses exploratory search and aims to minimize the oracle problem at both design time and run time. We demonstrate our framework through a motivating example that illustrates each step of the process.