A model-based approach for exploring the space of adaptation behaviors of safety-related embedded systems

Embedded systems are becoming ever more adaptive. One reason for this is the increasing need for faulttolerant systems that provide a certain level of service quality in all environmental situations and even in situations where erroneous signals are present. The adaptation behavior of a safetyrelated embedded system has to be predicable in order to assure at design time that the system will behave safely at runtime. This implies that the configurations of the system are predefined at design time. The system configurations are the target states that can handle different signal errors and environmental situations. As every system configuration causes costs, and as every system configuration provides another set of features, it is necessary to identify a set of system configurations that optimizes the tradeoff between utility and costs. This is a complex task as the number of possible system configurations explodes. The state of the art provides no solution for coping with this complex task. It only allows identifying some valid system configurations. The approach presented in this thesis enhances the state of the art as it provides means for identifying a utilitycost optimized set of system configurations at design time. First, it provides a language for modeling the information that defines which of the system configurations are reasonable. Second, it provides an automated analysis for identifying all reasonable system configurations. Third, it provides a language for modeling the environment of the system and some analyses for selecting reasonable system configurations with respect to the system's environment. Fourth, it introduces an approach for implementing an adaptive system that reconfigures onthefly between the selected system configurations. Fifth, it provides an engineering approach that explains how the provided approaches should be used in a scenario with realworld utility and cost objectives