Data-driven networked predictive control of a class of nonlinear systems with random packet dropouts

This paper investigates the data-driven networked control problem for a class of nonlinear systems with random packet dropouts in the feedback and forward channels. To compensate for the two-channel packet dropouts, a new data-driven networked predictive control method is proposed by using a dynamic linearization approach. In the controller design, a general dynamic linearization data model, i.e., full form dynamic linearization model, is employed instead of the compact form dynamic linearization model to predict future system outputs. Then based on such a model, a sequence of future control increment predictions are generated and transmitted to the actuator through networks to compensate for the random packet dropouts in both channels. The proposed method merely depends on the input and output data of the controlled plant, without using any explicit or implicit information about its structure and dynamics. Finally, the effectiveness of the proposed method is demonstrated through simulation results.