An Event-triggered Output-based Model Predictive Control Strategy

In many of today's control systems, the control loop is closed over a communication network with limited resources that requires explicit consideration of the network utilization in the controller design. At first, this paper deals with the decentralized control of discrete-time linear time-invariant systems consisting of multiple dynamically coupled subsystems subject to state and input constraints where the sensors send the measurements over a network to the controllers. A novel design for an event trigger mechanism for reducing the communication effort over the network in combination with an output-based model predictive controller is proposed. The approach is enhanced with another event trigger for a configuration where additionally the controllers send the inputs to the actuators over the network. Furthermore, it is discussed how the scheme can be extended for the control of sampled-data systems and how the network imperfections bounded time-varying delays, and sampling intervals can be considered in the design. Recursive feasibility of the control law and exponential stability of the origin are guaranteed for all considered setups. Illustrative examples demonstrate the clear reduction of the communication effort compared to a time-triggered output-based model predictive controller.