Today’s distributed applications require steady maintenance. To tackle this problem, so-called self-adaptive systems (SAS) can be used to change the behaviour automatically to adapt to a changing environment and context. Open challenges remain when those SAS get combined with Systems of Systems (SoS). SoS can get partitioned in multiple subsystems as a result of errors or connection faults which rises the need for a decentralized self-adaptation approach in SoS. In this doctoral paper, those open challenges are discussed and explained using a scenario of self-driving vehicles. Ideas for solving the problems are presented and the evaluation method of using the Webots simulation environment is explained. Solving the problems of self-adaptive SoS will enable robust adaptations in large-scale systems.