Self-organized network development: Control of complex nonlinear dynamics

• Applicants: Professor Dr.-Ing. Thomas Meurer; Professor Dr. Martin Ziegler
• Subject Area: Electronic Semiconductors, Components and Circuits, Integrated Systems, Sensor Technology, Theoretical Electrical Engineering
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 567385006

In our nervous system, billions of sensory inputs are processed per second, enabling us to adapt to constantly changing environmental conditions. Motivated by this, the goal of this project is to develop a neuromorphic controller that enables efficient processing of sensory information, adapts its behavior to changing environmental conditions, and predicts the behavior of complex dynamic systems to achieve their conceptually model-free control. To achieve this, system theoretic approaches for controlling complex dynamic systems will be combined with memristive neuromorphic architectures, which allow for efficient implementation of neural functionalities in real-time capable energy-efficient hardware. In doing so, we aim to bridge the paradigms of structurally dynamic relationships in biological systems and those in memristive neuromorphic systems. To address the complexity of the problem and to enable a comparison with existing machine learning approaches, we consider the concept of Reservoir Computing. In this context, mechanisms for self-organization of connectivity between neurons in the reservoir and/or output layer are investigated and tailored to the specific properties of memristive devices. We combine techniques for controlling nonlinear systems with Reservoir Computing architectures that functionally exploit the memory and dynamic properties of memristive devices to predict the dynamic behavior of quasi-chaotic systems. We will test the performance of the developed memristive neuromorphic controller in various control tasks, such as stabilizing a MEMS-based Duffing oscillator between unstable fixed points or in a desired state, speech recognition, and sound source localization under changing environmental conditions and the influence of background noise.

• DFG Programme: Research Grants
• Co-Investigator: Dr. Kristina Nikiruy