REINS

Self-organizing emergent systems are known for their scalability, robustness, flexibility, and adaptivity rather than their efficiency, optimality, and predictability. However, several (mainly industrial) settings require guarantees for a certain degree of performance respectively efficiency in varying situations. But due to the non-deterministic and open nature of these systems the efficiency can not be guaranteed without further ado, in particular not in priorly unexpected situations. Thus, the objective of the REINS (Runtime Efficiency Improvement of Self-organizing Emergent Systems) project is to facilitate the improvement of the systems' efficiency at runtime by implementing a feedback control loop that takes into account the low observeability and poor controllability of the systems, preserves their basic self-organizing emergent behavior, and considers their opennes and autonomy. In this project, we therefore develop the concept of an advisor as dedicated system element performing the control loop. Instead of controlling the system rigidly, this element only provides advice to the system elements, how they can perform better in certain future situations, so that every problem solving decision is still locally made by each system element itself.