Generalized predictive control of a class of MIMO models via a projection neural network

The system identification and generalized predictive control of a class of multiple input multiple output models are studied. The generalized predictive control problem with unknown parameters is first addressed by finding a control sequence for control performance as a goal. Then, the unknown parameters of the models are estimated by a new stochastic gradient algorithm providing high estimation accuracy. Third, the generalized predictive control problem is formulated to a quadratic programming problem with linear inequality constraints. Finally, the constrained quadratic programming problem is solved through a generalized projection neural network with simple structure and small number of neurons, while previous projection neural networks have complex structure and require more neurons. Numerical simulations are provided to reinforce our theoretical results.