Comparison of Approaches for Self-Improvement in Self-Adaptive Systems

Various trends such as mobility of devices, Cloud Computing, or Cyber-Physical Systems lead to a higher degree of distribution. These systems-of-systems need to be integrated. The integration of various subsystems still remains a challenge. Self-improvement within self-adaptive systems can help to shift integration tasks from the static design time to the runtime, which fits the dynamic needs of these systems. Thus, it can enable the integration of system parts at runtime. In this paper, we define self-improvement as an adaptation of an Autonomic Computing system's adaptation logic. We present an overview of approaches for self-improvement in the domains of Autonomic Computing and self-adaptive systems. Based on a taxonomy for self-adaptation, we compare the approaches and categorize them. The categorization shows that the approaches focus either on structural or parameter adaptation but seldomly combine both. Based on the categorization, we elaborate challenges, that need to be addressed by future approaches for offering self-improving system integration at runtime.