Besides advanced modeling techniques to specify the cooperation of multiple individuals or subsystems, also the specification and execution of possible reconfiguration of the system architecture (e.g., in case of failures) of the individuals is studied. For the resulting models, automatic code generation as well as synthesis of optimized realizations are considered. Finally, we explore how the required dependability guarantees for the resulting complex and sometimes even emergent behavior can be provided by analyzing the models and testing the systems. Examples of such dependability guarentees are required real-time constraints that need to be fulfilled.
As a first result, we have presented an approach to extend AUTOSAR architectures with reconfiguration capabilities (cf. our publications at the official project website). There exist two different strategies for how and when to identify appropriate configurations. First, identification of possible system configurations is completely done at runtime or second, the identification is specified by the developer at design time. Our approach fits into the existing AUTOSAR framework and meets certain performance requirements (reconfiguration need to be applied under hard real-time constraints). Therefore, the second strategy has been chosen. Thus, the possible configurations as well as the decisions for when to switch between them are specified at design time. In such a way a set of possible variants in the form of configurations is defined before runtime and one of these is selected at runtime accordingly. In doing so the overhead added to the resulting reconfigurable architecture has been shown to be negligible, but the developer rewards an easier development of reconfiguration logic, which otherwise has to be done manually at the functional / implementation level. We have successfully shown that it is possible to use high-level architectural modeling techniques without generating massive runtime overhead.
Currently we investigate how to realize reconfiguration according to the first strategy, identification of possible system configurations completely at runtime, without jeapordizing required properties like in the case of real-time constraints.
You find our publications for this research area at the official project website.