Using recurrent neural networks to build a data-driven model of an autonomous underwater vehicle

This paper describes an approach to preparing a data-driven model for simulating the behavior of an autonomous underwater vehicle (AUV), using recurrent neural networks (RNNs). The main task of the research system was to process data containing states and control signals in such a way as to then enable the start of the learning process of selected types of artificial neural networks. In the conversion process, the data are transformed into time sequences that will constitute the training set for Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks. Artificial neural network models, based on the above-mentioned architectures, were tested on the validation set to determine the discrepancy in prediction accuracy compared to real data. The obtained results were analyzed which allowed to evaluate the proposed solutions and draw conclusions.