Data-driven moment-based control of linear ensemble systems

The problem of controlling ensembles of similar dynamical systems appears in many scientific domains, ranging from quantum physics and neuroscience to robotic engineering. These applications have led to the development of a variety of ensemble control methods. These methods, although effective, rely on a critical assumption, namely, the availability of an accurate parametric model for the entire population. In this paper, we relax such assumptions and present a data-driven framework for designing both open-loop and feedback control schemes for ensemble systems, which only uses the measurement of a small number of subsystems within the population. We validate our approach through numerical analysis and simulation, showing its effectiveness in regulating large populations with just a few of their subsystems measured.