Scholarlink

For the movement of Autonomous Underwater Vehicles(AUVs), the impact of ocean currents is extremely important. Therefore, planning a navigation path that better conforms to the currents can greatly improve navigation efficiency. Although AUVs can measure nearby current changes with their sensors, it is impossible to measure the current speeds at all points in the entire sea area in real time while navigating, which means that a globally optimal path considering currents cannot be planned. In this paper, we propose an ocean current prediction method using Long Short-Term Memory (LSTM) networks with an attention mechanism to predict the current speed at any location in the sea area at any given time. The neural network is trained with data from the South China Sea. Experiments show that our method is more accurate compared to traditional LSTM and BP (Back Propagation) networks.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Ocean Current Prediction Algorithm Based on Attention LSTM Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Bowen Fu,
  • Jibo Bai,
  • Mingzhi Chen,
  • Daqi Zhu

摘要

For the movement of Autonomous Underwater Vehicles(AUVs), the impact of ocean currents is extremely important. Therefore, planning a navigation path that better conforms to the currents can greatly improve navigation efficiency. Although AUVs can measure nearby current changes with their sensors, it is impossible to measure the current speeds at all points in the entire sea area in real time while navigating, which means that a globally optimal path considering currents cannot be planned. In this paper, we propose an ocean current prediction method using Long Short-Term Memory (LSTM) networks with an attention mechanism to predict the current speed at any location in the sea area at any given time. The neural network is trained with data from the South China Sea. Experiments show that our method is more accurate compared to traditional LSTM and BP (Back Propagation) networks.