Abstract :
Summary form only given, as follows. The growing popularity of smart devices and applications is accelerating the convergence of the physical and the digital worlds. Smart apps allow users, with the help of sensors and networks, to do a great variety of things, from tracking their friends to controlling remote devices. At the core of such smart systems are self-adaptive systems. Self-adaptive systems optimize their own behaviour according to high-level objectives and constraints to address changes in functional and nonfunctional requirements as well as environmental conditions. Self-adaptive systems are implemented using four key technologies: runtime models, context management, feedback control theory, and run-time verification and validation. The proliferation of smart applications challenges the software engineering community in re-thinking the boundary between development time and run time and developing techniques for adapting systems at run time. The key challenge is to automate traditional software engineering, maintenance and evolution techniques to adapt and evolve systems at run time with minimal or no human interference. Hitherto, most developers did not instrument their software with sensors and effectors to observe whether requirements are satisfied in an evolving environment at run time. One way to break out of this mold is to make the four key technologies readily accessible at run time. Challenges for the reverse engineering community are to instrument existing systems with manageability endpoints (i.e., sensors and effectors), extract design time models that are useful as run-time models (e.g., to manage dynamic context), and investigate analysis techniques for run-time testing and validation (e.g., to regulate requirements).
