Collective dynamics, control and imaging

Model order reduction for networked control systems

Peter Benner

Max Planck Institut Magdeburg
[SLIDES]

In an increasingly connected world, networked control systems are a key technology for robust and stable operation of dynamical systems acting together via a network structure. As such systems are often very complex by nature, model order reduction (MOR) has become an enabling technology to apply real-time control strategies to such complex network systems. In this talk, we provide some insight into MOR techniques for networked control systems, with a primary focus on the preservation of structural properties, like the interconnection topology, stability and/or synchronization. We will primarily focus on multi-agent systems and discuss some recent clustering-based MOR strategies and associated error bounds. Most results currently apply to linear time-invariant systems, but we also show some first steps for network with nonlinear agent dynamics.