Event-based Measurement Scheduling for Linear Quadratic Gaussian Control

This paper investigates measurement scheduling for linear quadratic Gaussian (LQG) control. The measurement collected via the local sensor is sent over the bandwidth-limited channel to the remote controller for regulating the system state. An event-based scheme is designed at the local scheduler to smartly choose the time for sending the data, so as to reduce the uncertainty of system state at the controller side, as well as to reduce the communication cost. Thanks to the fixed event-triggered rule, we show that the control law and the estimation algorithm can be separately designed. On that basis and by virtue of the introduced generalized closed skew normal distribution, a minimum mean-squared error estimation algorithm is proposed, which promotes the existing scheduling scheme to any communication rate.