Identifying Adaptation Changes in Collections of Self-Adaptive Systems
Martin Goller, Sven Tomforde · 2022
Self-adaptive and self-organised behaviour of technical systems should lead to higher robustness, performance and less administration effort. However, the resulting increased autonomy of the systems also leads to new kinds of challenges: Continuous self-monitoring and evaluation of their own behaviour becomes necessary. Part of such a self-assessment is the identification of causal events for abnormal or even disturbed system behaviour – as a basis for explanations to users and as a basis for learning processes. In this paper, we present an approach that uses metrics to self-assess the distributed system behaviour of autonomous and self-organised subsystems based on external monitoring to determine the timing and location of causal events. Using a swarm simulation, we show that the corresponding events can be identified with high accuracy.