The application of alternative modelling techniques to ROV dynamics

Simulation studies indicate that multi-input/multi-output recursive system identification (SI) and batch parameter estimation (PE) methods can be applied advantageously to the problem of modeling remotely operated underwater vehicles (ROV) dynamics. Least-squares (LS), extended-least-squares (ELS), and recursive prediction error (RPE) SI methods were used for SI. ELS produced slightly lower values of Akaike's information criterion (AIC) than did LS, indicating that the estimation of the noise model C(z/sup -1/) is worthwhile. RPE had relatively poor performance compared to LS and ELS. Statistical tests indicated that all of the nonlinear dynamics are conveniently lumped into low-order linear models. Real-time, recursive SI is a requirement for adaptive closed-loop control, where the performance of the ROV is estimated continuously and the controller gains are adjusted accordingly. Maximum-likelihood PE was used to produce a continuous LTI model of a ROV. Preliminary results are promising, although more work is required to define the estimated model structure.<
>