Predict the Future: Preventing unanticipated changes is the ultimate challenge for self-adaptive systems

Systems interact with their environment. This might lead to unanticipated events as system models usually cover only to a certain extent the dependencies of a system with the environment. We argue that many of these unanticipated events might become predictable in case we handle current and in particular past behavior of the environment by digital twins. We propose to refine the traditional MAPE-K architecture towards a MAPE-Twin architecture. Thus, the traditional knowledge component becomes an analyzable repository of behavior which allows to predict potential events in the future and to deal with them in a predefined way. Thus, the ultimate challenge for self-adaptive systems are not unanticipated changes, but the prediction of future behavior of a system and its environment.