A Measurement Framework for the Assessment of Self-Adaptation and Self-Organisation in Systems with Decentralised Configurations

In many areas of our lives, we find technical devices that are able to autonomously make decisions and adapt their behaviour to changing conditions without the intervention of a human being. A particularly prominent example is self-driving cars, which every major car manufacturer offers today. The ability to drive fully autonomously in any traffic situation is being worked on intensely. The next step in the evolution is the collaboration of multiple autonomous devices that collectively manage a given task. Examples include smart cameras that can trace objects in large areas. Starting from the hypothesis that unexpected changes in behaviour result from unusual self-adaptations, the question arises as to how unusual self-adaptations can be detected. As an answer to this question, this thesis presents a framework of different metrics that translates change in configuration parameters in a system to an assessment and quantification of the self-adaptation behaviour.
The metrics contained in the framework are formally defined, explained in detail and compared to existing metrics. Then, several evaluation scenarios from different domains are presented and evaluated with the framework. Here, a focus is set on specific events in the scenarios that lead to unusual self-adaptation behaviour. The results are then evaluated and brought into context.

The thesis then closes with a presentation of an extension to the framework that can aid in root cause analysis. The procedure for such an analysis is then illustrated in different scenarios, showing that the framework is eligible for this application.

Rights

Use and reproduction:


CC BY 4.0

Please note that individual components of the publication may be subject to other licensing or copyright conditions.

Cite

Citation style:
Could not load citation form.