Neural network control approach of a midwater trawl system based on feedback linearization

It is almost impossible to obtain the precise mathematical model of a Midwater trawl system because of the unknown model structure and uncertain parameters. Therefore, conventional model-based control method does not work out for the depth tracking of a Midwater trawl system. Nonlinear feedback linearization is an effective control strategy for the nonlinear system, however, it´s hard to implement since the dynamic mathematical model is usually not accurate. A neural network controller based on feedback linearization was adopted for the depth control of the towed net within this paper. NARMA-L2 model was used for the training of multilayered models to study the dynamics of trawl system. In using a NARMA-L2 model the control strategy is nonlinear and online adaptation of the model is possible to capture unknown time-varying dynamics. A simplified mathematic model of a Midwater trawl system is presented for the simulation which takes the parameters uncertainties and disturbance into consideration. The simulation results show that the NARMA-L2 control method is effective.