- With the development of low power technology and battery industry, underwater gliders (UGs) have realized a remarkable endurance. As autonomous platforms for long-term observation in complex ocean environment beyond visual line of sight, their motion condition monitoring appears to be particularly important for adjusting flight parameters in real time or finding malfunction in time. This paper proposes a novel method of motion condition monitoring for UGs. Firstly, a dynamic model is established for data collection, which takes system errors and variables influenced by disturbances into consideration. Then, the deep learning networks are trained and tested with datasets from dynamic model, which is used to identify the variable parameters by importing the cleaned and standardized glider datasets based on the multi-objective optimization problem. Finally, the motion conditions of UGs in three-dimensional space can be monitored by analyzing the trend of identification parameters, which contributes to flight parameter adjustment for improving glide efficiency and trajectory accuracy, or fault diagnosis in time for avoiding loss and potential risks. The method proposed in this paper is verified by massive trial data of Petrel-L, which is also appropriate for other marine autonomous systems.
展开▼
