The Collaborative Research Center 614 "Self-Optimizing Concepts and Structures in Mechanical Engineering" deals extensively with challenges related to these innovative systems. The paradigm of self-optimization may thereby be used as an opportunity for increasing the dependability if and when the additional functionality and variability of self-optimizing systems is used to counter threats during runtime. However, self-optimization also increases the complexity of the mechatronic system, resulting in a greater risk of systematic errors and making systems more prone to error. Another new challenge resulting from self-optimizing systems is also that, as a result of the self-adaption of objectives, the resulting behavior of the systems either may not or may only with great difficulty be determined in advance.
A continuous analysis of the system’s state and the environmental conditions, as well as the ability to respond appropriately to changes, enables a self-optimizing system to consider and include objectives that increase dependability. Using self-optimizing methods can prevent the occurrence of errors by detecting potential errors in advance and adapting system behavior accordingly. On the other hand, such a system can also react to occurring failures to ensure continuing dependability, depending on the type of error.
The multi-level dependability approach is the central concept for connecting self-optimization with monitoring of system behavior. Using the four areas shown, the importance of dependability is determined depending on the current situation. A detailed description can be found here.
You can purchase Dr. Sondermann-Wölke's dissertation thesis here.
His publications are listed here.