Multi input-multi output system identification of AUV systems by neural network

Accurate identification of non-linear time variant multi input-multi output systems, especially for AUV systems are essential for implementation of control algorithms and navigation. Nonlinearities, which come from coupling effects between different motion modes of an AUV and also environmental effects, are very complicated and cannot be modeled easily by using conventional dynamics methods. In the paper, by implementation of a neural network identifier (NNI) a general idea for dynamic modeling of any six-degree of freedom rigid body is discussed, and then as an example, the proposed idea is applied to a four degree of freedom AUV system, named Twin Burger 2. This vehicle has four motion modes, called surge, sway, heave, and yaw. It is equipped with an inertial navigation system that can detect state variables of its motion. The NNI is made of two main parts. The first part is made of six independent single degree of freedom neural network identifiers (SDFNNI), which in the case of Twin Burger 2, is four. After training each independent SDFNNI, the second part or main part, named coupled model neural network identifier (CMNNI) is trained. In order to train all of the proposed neural network identifiers, a series of experiments were done and training data packs were collected. From the simulation results it can be seen that the proposed idea is satisfactorily able to identify dynamic behavior of the AUV system under consideration, and also it can be concluded that the proposed method will be useful for system identification of similar systems