Modeling of Unmanned Underwater Vehicles based on System Identification Techniques with experimental data

Abstract

This talk will present a dynamic modeling technology of an AUV. The presentation briefly introduces mathematical model of the AUV considering its shape. The hydrodynamic coefficients were estimated by empirical formulas and data base of similar-shaped vehicles. A least square method were implemented to estimate unknown nonlinear parameters of the AUV. The parameters were finely tuned by comparing the experiments and the simulated responses generated by adopting the same control inputs of the experiment. The AUV’s speed for the test was fixed with 3.5 knots. Additional simulation was conducted with an experiment including variable speed commands. The simulated result has shown good consistent being similar with those of the experiment even under speed variation. This presentation will also discuss the feasibility of applying a deep learning algorithm, LSTM (Long Short-Term Memory Models), to system identification issue for dynamic modellingof underwater vehicles based on experimental data.