Packetized MPC with dynamic scheduling constraints and bounded packet dropouts
We study a Networked Control System architecture which uses a communication network in the controller-actuator links. The network is affected by packet dropouts and allows access to only one plant input node at each time instant. This limits control performance significantly. To mitigate these limitations we propose a control and network protocol co-design method. Succinctly, the underlying features of the proposed method are as follows: a sequence of predicted optimal control values over a finite horizon, for an optimally chosen input node, is obtained using Model Predictive Control ideas; the entire resulting sequence is sent to the chosen input node; a smart actuator is used to store the predictions received and apply them accordingly. We show that if the number of consecutive packet dropouts is uniformly bounded, then partial nonlinear gain @?"2 stability and also a more traditional linear gain @?"2 stability can be ensured via appropriate choice of design parameters and the right assumptions. Whilst our results apply to general nonlinear discrete-time multiple input plants affected by exogenous disturbances, for a disturbance-free case we prove that Global Asymptotic Stability follows from our main result. Moreover, we show that by imposing stronger assumptions, Input-to-State Stability is achievable as well. Finally we demonstrate the potential of the proposed method via simulations.