Deep learning model inspired by lateral-line system for underwater object detection

This study develops a deep learning-based object identification model that can identify objects through flow information measured from a moving sensor array, inspired by the lateral-line systems of various aquatic organisms capable of hydraulic imaging with ambient flow information. A hydrofoil navigates around four stationary cylinders in uniform flow, and two types of sensory data, flow velocity and pressure, are obtained numerically from an array of sensors located on the surface of the hydrofoil with potential flow assumption. Several neural network models based on the flow velocity and pressure are built to identify the positions of the foil and surrounding objects. The LSTM-based model, which is a type of recurrent neural network capable of learning order dependence in sequence prediction problems, outperforms the other network models. The optimization of the number of sensors is then performed, using feature selection techniques, LASSO and Elastic Net. Through sensor optimization, a new object identification model shows impressive accuracy in predicting the locations of the foil and objects with only 40% of the sensors used in the original model.