Anomaly Detection and XAI Concepts in Swarm Intelligence
Abstract
For human operators in swarm intelligence, decision support in critical situations is crucial. The large amount of data shared by the autonomous systems, can easily make the human decision makers too overwhelmed by it and hence there is a need for support in analysing the data in an intelligent way. For this purpose, automatic systems for assessing situations and indicating suspicious behaviour or statistical outliers is employed. This strengthen their situation awareness as well as decrease the work load. Therefore, in this work, we emphasize that the data fusion services developed for detecting anomalies in surveillance tasks, e.g. in the maritime domain, can be adapted to support operators in swarm intelligence. Furthermore, in order to make the behaviour of the swarm and the results of the data fusion services understandable to the human operator, explainable artificial intelligence (XAI) concepts are introduced. This makes the autonomous system's behaviour more intelligible and understandable to humans by providing explanations for certain decisions.
