Least Squares Optimisation Algorithm Based System Identification of an Autonomous Underwater Vehicle

Abstract

This paper presents the identification of mathematical models governing dynamics in both the vertical and horizontal planes for a axisymmetric, torpedo-shaped Gavia class Autonomous Underwater Vehicle (AUV), based on a least squares optimisation algorithm. Rather than using the least squares algorithm to roughly estimate the mathematical models in a fixed time period, a simulator is developed based on least squares optimisation algorithm with the goal of accurately predicting the system response over time starting from initial conditions. The general equations for six degrees of freedom motions are decoupled into non-interacting longitudinal and lateral subsystems in the form of linear state space models with unknown parameters. These unknown parameters are initially determined by applying a least squares algorithm for experimental data collected from the AUV’s on-board sensors. The previously identified models are then optimised to form the simulator able to estimate the system response. The numerically simulated data from the simulator show a good agreement with the field measured data. The simulator provides a useful tool to examine the manoeuvrability of AUV. The verification process proved that the least squares algorithm could be utilised as an optimisation algorithm in the system identification of autonomous underwater vehicle.