Model-Based Testing for MAPE-K adaptation control loops

Self-adaptation is an effective solution for complex systems that need to adapt themselves to external changes while keeping internal goals; it can be realized in terms of MAPE-K feedback control loops, organized according to specific MAPE patterns in case of decentralized and distributed control. The MSL-centric framework allows the definition of the structure of MAPE-K loops using the MAPE Specification Language (MSL), and the description of its adaptation logic in terms of self-adaptive Abstract State Machines (ASMs). These models can be tested by exploiting classical model-based testing for ASMs that allows structural coverage of ASM models, but which is not adequate to test executions that are peculiar of the structure of MAPE-K loops. To overcome this limitation, in this paper, we propose a MAPE-based testing approach, which generates, from the MSL model, test goals specifying requirements on the control loop executions, translates these into concrete test predicates for the corresponding self-adaptive ASM, and then generates (by using a model checking based technique) abstract tests covering these test predicates. Such abstract tests are suitable to test the desired executions of MAPE control loops. The application of the approach is demonstrated on a smart home case study.