AUV 3D Trajectory Prediction Based on CNN-LSTM

Abstract

When multiple AUVs perform formation tasks underwater, there is a delay in the follower receiving the leader’s information, so that the follower cannot accurately follow the leader. In response to this problem, this paper designs a short-term trajectory prediction scheme of CNN-LSTM. First, the data is processed, and then the CNN-LSTM neural network trajectory prediction model is constructed by mining the time series relationship in the historical data of the leader. Finally, the accuracy and robustness of the prediction of the CNN-LSTM model are verified by comparing with the prediction results of other models.