Intelligent Systems




Technical systems should take over more and more tasks autonomously and reliably. As far as possible, they should act in the background, serve people and make intelligent decisions. This applies above all to appropriate reactions to unknown situations or the provision of services even under severely disturbed conditions. Tasks that people use their creativity to solve. The working group "Intelligent Systems" of the Christian-Albrechts-Universität zu Kiel develops approaches and procedures that enable technical systems to find appropriate and continuously improving answers to challenges even without human creativity.

What is an "intelligent system"?



We define the term "intelligent system" as follows: "An intelligent system is a computing system that achieves or maintains a certain level of performance, even when operating in environments that change over time and even if it is exposed to disturbances or emergent situations. Such an intelligent system is self-adaptive and improves its behaviour over time."

This requires in particular that technical systems have mechanisms of autonomous learning, i.e. methods for independent, opportunistic learning without (or with only minimal) involvement of the human user and with only minimal prior knowledge from design time.


Our activities are divided into the following columns:

1. Design of intelligent systems

  • Architectures of intelligent systems

  • Transfer of design decisions to the systems and into the runtime

  • Integration of machine learning techniques, safety concepts and self-organisation patterns

  • Interwoven system structures and self-integration processes of autonomous (sub)systems



2. Autonomous learning: methods for learning at runtime without/with very little external control (by users)

  • Reinforcement learning (learning through feedback, comparison of observation and expectation) based on utility functions

  • Anomaly detection

  • Transfer Learning

  • Active Learning (active request of knowledge from oracles), independent evaluation of existing knowledge



3. System description and analysis

  • Quantification of system properties, i.e. metrics for self-organisation, degree of adaptation, emergence, etc.

  • Modelling the perception of a technical system based on sensor data

  • Data preparation, analysis and processing at runtime

  • Recognition of mutual influencing effects and consideration of these in learning and decisionmaking processes


4. Trust and security

  • Techniques for detecting attacks and evaluating conspicuous behaviour

  • Methods for establishing Technical Trust among autonomous subsystems

  • Trust-based system organisation

  • Methods for detecting mutual influencing effects

  • Safety oriented communication protocols and self-organisation patterns




The work is based on mastering complex information and communication technology (ICT) systems and solving concrete problems. Our work is therefore based on a wide range of applications, the most prominent of which are:

  • Traffic control and management

  • Data communication networks

  • Future decentralised energy systems / Smart Grid

  • Internet of Things

  • Surveillance networks

  • Intelligent devices