A Runge-Kutta MLP Neural Network Based Control Method for Nonlinear MIMO Systems

In this paper, Runge-Kutta MLP based self-adaptive controller (SAC) is proposed for nonlinear multi-input multi output (MIMO) systems. The controller parameters are optimized by considering K-step ahead future behavior of the controlled system. The adjustment mechanism is composed of an online Runge-Kutta identification block which estimates a forward model of the system, an adaptive multi-input multi-output (MIMO) proportional-integral-derivative (PID) controller and an adjustment mechanism realized by separate online Runge-Kutta MLP neural networks to identify the dynamics of each tunable controller parameter. The performance of the introduced adjustment mechanism has been examined on a nonlinear three tank system for different cases, and the obtained results indicate that the RK-MLP-NN based adjustment mechanism and Runge-Kutta model acquire good control and identification performances.