Modeling for the performance of navigation, control and data post-processing of underwater gliders

Abstract

Underwater gliders allow efficient monitoring in oceanography. In contrast to buoys, which log oceanographic data at individual depths at only one location, gliders can log data over a period of up to one year by following predetermined routes. In addition to the logged data from the available sensors, usually a conductivity-temperature-depth (CTD) sensor, the depth-average velocity can also be estimated using the horizontal glider velocity and the GPS update in a dead-reckoning algorithm. The horizontal velocity is also used for navigation or planning a long-term glider mission. This paper presents an investigation to determine the horizontal glider velocity as accurately as possible. For this, Slocum glider flight models used in practice will be presented and compared. A glider model for a steady-state gliding motion based on this analysis is described in detail. The approach for estimating the individual model parameters using nonlinear regression will be presented. In this context, a robust method to accurately detect the angle of attack is presented and the requirements of the logged vehicle data for statistically verified model parameters are discussed. The approaches are verified using logged data from glider missions in the Indian Ocean from 2016 to 2018. It is shown that a good match between the logged and the modeled data requires a time-varying model, where the model parameters change with respect to time. A reason for the changes is biofouling, where organisms settle and grow on the glider. The proposed method for deciphering an accurate horizontal glider velocity could serve to improve the dead-reckoning algorithm used by the glider for calculating depth-average velocity and for understanding its errors. The depth-average velocity is used to compare ocean current models from CMEMS and HYCOM with the glider logged data.