Ji-Feng Zhang

Fundamental Problems and Recent Progresses of Quantized Systems: From Parameter Identification to Adaptive Controls

Systems with quantized data widely exist in practice. Typical examples include systems with finite-bandwidth communication channels and systems with binary-valued, set-valued, feature, or censored data. Control of such systems is essentially different from control with accurate data because quantizers and the coupling of quantization error with control inputs may introduce strong nonlinearities and large, or even unbounded, quantization errors into closed-loop systems.

To achieve satisfactory or optimal control performance, the controller and the quantizer should be designed in a unified manner. This leads to two fundamental questions: how much information is really needed to obtain a satisfactory parameter estimate, and how much information is really needed to obtain good control?

The talk presents recent progress by Professor Zhang's research group, including parameter identification of typical linear and nonlinear systems with binary-valued output observations or quantized inputs, adaptive tracking control with binary-valued output observations and time-varying thresholds, distributed dynamic consensus with quantized communication data, output-feedback quantized observer-based synchronization, and coordination over multi-agent networks with finite-level quantization.