The Application Of FCMAC Neural Network For Motion Control Of AUVs

In this paper, the AUV is the object of the research. In order to overcome the negative effects of the non-linear part of kinematic model, sea current and wave during motion control, a new type of neural network: Fuzzy Cerebellar Model Arithmetic Controller(FCMAC), together with an on-line learning scheme based on the Lyapunov Stability Theory, has been adopted in designing the new motion control system. The simulation results have been compared with those generated by a classic PID controller.After the construction of coordinate and kinematic model( which includes the effects of sea currents and waves), FCMAC controller is applied to the depth and pitch control of submarine in the vertical plane and the 5-degree motion control of AUV. As the premise of implementing actual control, chapter 3 completes the program implementation of FCMAC in C++ language after a brief analysis of the FCMAC's working theory, the program includes initialization, feed-forward scheme, learning scheme and so on.In FCMAC's application in the vertical plane motion control of submarine, its output is used to compensate the effect of the non-linear part of the kinematic model, this function is realized by means of FCMAC's on-line learning function. The rest of the model can be seen as a linear system, and is controlled by a PID controller. The simulation results show that the FCMAC controller outperforms the classic PID controller in control accuracy as well as robusticity.In FCMAC's application in the 5-degree motion control of AUV, the function of FCMAC remains compensating the effects of the non-linear part of the kinematic model. Because the disposition of execute components goes against the implementation of control, the control scheme is adjusted in corresponding part, meanwhile a predictive direction ascertaining arithmetic is added to the control scheme. During the debugging of this control scheme, some constructive work has been done: the ascertaining of structural parameters of FCMAC, the relations between the prediction number and the state of control object, the adjustment of learning scheme when actual rudder( in which inertia must be taken into account)is adopted. Simulation results show that the performance of FCMAC controller is better than that of PID controller in control accuracy and robusticity while sea current is confined within some certain amount.