Linear and nonlinear system identification techniques for modelling of a remotely operated underwater vehicle

As opposed to classical mathematical-based modelling approach, this paper reports a black-box system identification technique for characterising the dynamics of a remotely operated vehicle (ROV). A linear system identification technique is employed to model the vehicle dynamics. However, use is also made of advance neural networks-based nonlinear system identification approach to model rudder-depth channel nonlinear behaviour. Different model validity tests are also employed to instil confidence in the identified linear and nonlinear ROV dynamic models. High fidelity models obtained for the multi-degree-of-freedom vehicle are of immense importance for developing ROV simulators, pilot training and autopilot design.