A dynamically adaptive system (DAS) is a software system that monitors itself and its environment at run time to identify conditions that require self-reconfiguration to ensure that the DAS continually satisfies its requirements. Self-reconfiguration enables a DAS to change its configuration while executing to mitigate unexpected changes. While it is infeasible for an engineer to enumerate all possible conditions that a DAS may experience, the DAS must still deliver acceptable behavior in all situations. This dissertation introduces a suite of techniques that addresses assurance for a DAS in the face of both system and environmental uncertainty at different levels of abstraction. We first present a technique for automatically incorporating flexibility into system requirements for different configurations of environmental conditions. Second, we describe a technique for exploring the code-level impact of uncertainty on a DAS. Third, we discuss a run-time testing feedback loop to continually assess DAS behavior. Lastly, we present two techniques for introducing adaptation into run-time testing activities. We demonstrate these techniques with applications from two different domains: an intelligent robotic vacuuming system that must clean a room safely and efficiently and a remote data mirroring network that must efficiently and effectively disseminate data throughout the network. We also provide an end-to-end example demonstrating the effectiveness of each assurance technique as applied to the remote data mirroring application.
Read
