An Autonomic Computing Approach for Systemic Self-Regulation

Project Leaders: Sebastian Wrede, Marc Hanheide, Martin Heckmann
PhD Student: Raphael Golombek

To face the challenges that arise from the ever growing complexity of software-intensive systems, novel scientific and technological methodologies are mandatory. Robots with cognitive abilities that operate under varying environmental conditions in the real world are instances of these systems where new approaches to analyze, understand and control their complex runtime behavior are needed. Hence, this project is located at the crossroads of software engineering, machine learning and human machine interaction technology that exploits the paradigm of autonomic computing to build artificial cognitive systems that feature an increased level of dependability.

Dynamic environments forced living organisms to develop self-regulating mechanisms in order to survive. Consequently, biology provides many inspirations for the design of analogue algorithms and system structures. The vision of this project is to mimic the functions of the autonomous nervous systems for a new class of self-managing and self-aware software architectures that will allow, for instance a robot to autonomously monitor its internal dynamics, detect, and as well adapt to new situational contexts or malfunctions. This shall increase the level of safety and autonomy of such systems while at the same time minimising the amount of human intervention. The scientific aspect is to combine ideas from machine learning and artificial intelligence with advanced concepts from software engineering, resulting in an autonomic computing architecture for interactive cognitive systems.

Within the research roadmap of the CoR-Lab's graduate school, the notion of natural and therefore varying environments as the operational conditions for future robotic systems is particularly emphasized in the manifesto of the Resource-bounded Adaptation research activity. Autonomic computing in this context shall support the development and understanding of the more and more complex systems we work on and significantly increase the level of reliability these systems will feature operating in the real-world.