Event-Based Networked Model Predictive Control With Overclocked Local Nodes

We improve a recently proposed networked MPC approach by reducing the memory and computational requirements for the local nodes and by reducing the network use. The networked MPC setup consists of a powerful central node that provides one or multiple local nodes with regionally optimal affine feedback laws. Whenever the region of optimality of a current affine law is left on one of the local nodes, this local node requests a new optimal law from the central node. Since the local node uses lean hardware, computational and memory resources are restricted. Similarly, network usage should be as small as possible. The proposed method essentially increases the sampling time and reduces the horizon of the underlying MPC, which results in the desired reductions on the local node and the network usage. We show that the feedback laws can be overclocked on the local nodes to compensate for the loss of performance due to longer sampling times and shorter horizons. Results are obtained from hardware-in-the-loop simulations with a micro controller and a PC as local and central nodes, respectively, and with a wireless network.