Path Optimization of Underwater Glider Based on Depth-averaged Current Prediction Model

Abstract

With the wide application of underwater gliders in the field of ocean surveying and underwater acoustic detection, accurate and efficient control of their path is important for refined ocean observation. In view of the problem that the underwater glider has a large path deviation due to the influence of current, the least-squares support vector machine (LSSVM) method is used to predict the depth-averaged current. The minimum path deviation of a single profile is taken as the objective function, with the constraint condition that the difference between the actual and planned heading does not exceed a certain value. A nonlinear constraint extremum model is constructed, and the optimal target heading and outlet point coordinate are calculated, to realize the goal of path optimization. The historical data of the Petrel-II glider are used for verification, and the following results are obtained. 1) The LSSVM method has high accuracy in predicting the depth-averaged current, however, its prediction accuracy is poor when the local current direction changes significantly. The prediction accuracy is higher when the first three historical profile data are used as training samples. 2) Following path optimization with the proposed method, the path of the glider is more stable, and the average path deviation is 281.1 m.

Keywords

• underwater glider
• path optimization
• depth-averaged current
• least squares support vector machine
• constrained extremum
• target heading