Modeling the adaptation behavior of adaptive embedded systems
This thesis introduces a model-based approach for the specification of the adaptation behavior of adaptive embedded systems. Adaptive systems are capable to dynamically adapt to changing environmental conditions. While appropriate run time frameworks for the actual adaptation of the system are already available, the methodical specification of the adaptation behavior is still an open problem. Dynamic adaptation is very complex as the adaptation of one component usually requires the adaptation of further components. It is very difficult to identify these interdependencies manually. As no modeling support has been available so far, it has been hardly possible to specify how the system must adapt. The approach that is introduced in this thesis resolves these complex dependencies between components by attaching an appropriate quality description to the exchanged data. By this means, all information that is required in order to adapt a component are available at its local interface. This enables the modular modeling of reusable adaptive components, which is the major prerequisite for the modeling of complex systems. Based on an executable specification of the adaptation behavior of single components, the overall system can be automatically analyzed, simulated, verified, and, if necessary, optimized from a global point of view by centrally constraining the adaptation behavior. The introduced modeling approach has been practically applied in the development of the electronic stability program ESP of the Robert Bosch GmbH. The advantages of the approach over the development without an explicit modeling of the adaptation behavior, as well as the practical applicability and the suitability of the approach are underlined by the fact that the elementary concepts of the approach are integrated to the development process of the ESP-development.
Zugl.: Kaiserslautern, TU, Diss., 2005