Model identification of an unmanned underwater vehicle via an adaptive technique and artificial fiducial markers

To control any kind of unmanned underwater vehicle (UUV) it is important to know the dynamic model of the system. Creating the model itself is only the first step, the second is identifying the model's parameters. Due to the nonlinear and coupled characteristics of UUVs this is a complex and operationally demanding task. Several methods to identify the model parameters have been published in recent years. One of the biggest challenges is providing a data set of the vehicle's true motion, which has not been solved completely. This paper presents and evaluates a new method to identify the motion model parameters of a UUV. The method uses the UUV's pose estimation, the on-board cameras, fiducial markers and an adaptive method to combine the resulting data. The new method is used to parameterize the model of the FlatFish AUV. The paper closes with a comparison of the new method for model identification and the traditional least-squares method using onboard sensors.