A Multiple Models Approach for Adaptive Predictive Control of Networked Control Systems

A real-time, self-learning, multiple models adaptive predictive control algorithm is proposed for networked control systems(NCS), which can deal with different distributions of time delay introduced by a communication network or a field bus. The algorithm keeps updating the dynamic models in a model bank, and uses an adaptive model to track the parameter changes of NCS. In the mean time, it uses another adaptive model to ensure the stability of the overall NCS. The convergence analysis shows that the algorithm will finally switch to and stop at a stable, adaptive model.