Video image processing and object classification using a Deep Learning Neural Network (DLNN) can significantly increase the autonomy of underwater vehicles. This paper describes the results of a project focused on using DLNN for Object Classification in Underwater Video (OCUV) implemented in a Biomimetic Underwater Vehicle (BUV). The BUV is intended to be used to detect underwater mines, explore shipwrecks or observe the process of corrosion of munitions abandoned on the seabed after World War II. Here, the pretrained DLNNs were used for classification of the following type of objects: fishes, underwater vehicles, divers and obstacles. The results of our research enabled us to estimate the effectiveness of using pretrained DLNNs for classification of different objects under the complex Baltic Sea environment. The Genetic Algorithm (GA) was used to establish tuning parameters of the DLNNs. Three different training methods were compared for AlexNet, then one training method was chosen for fifteen networks and the tests were provided with the description of the final results. The DLNNs were trained on servers with six medium class Graphics Processing Units (GPUs). Finally, the trained DLNN was implemented in the Nvidia JetsonTX2 platform installed on board of the BUV, and one of the network was verified in a real environment.

This article covers research on an innovative propulsion system design for a Biomimetic Unmanned Underwater Vehicle (BUUV) operating at low speeds. The experiment was conducted on a laboratory test water tunnel equipped with specialised sensor equipment to assess the Fluid-Structure Interaction (FSI) and energy consumption of two different types of propulsion systems. The experimental data contrast the undulating with the drag-based propulsion system. The additional joint in the drag-based propulsion system is intended to increase thrust and decrease energy input. The tests were conducted at a variety of fins oscillation frequencies and fluid velocities. The experiments demonstrate that, in the region of low-speed forward movement, the efficiency of the propulsion system with the additional joint is greater.

In this paper, an investigation on the development of a low-cost passive hydroacoustic system for passive detection of moving vessels to counteract possible collision with an unmanned underwater vehicle is presented. The main goal of this paper is to determine if moving vessels generating hydroacoustic signals/signature are present in the space being searched, and if so, to determine the time delay ΔT between two signals (V₁(t) and V₂(t)) and consequently to estimate the bearing on the source of the hydroacoustic signals, for example, screw propellers of a moving vessel. The acoustic signals V₁(t) and V₂(t) have been recorded by a two hydrophones mounted in an unmanned underwater vehicle. In practice, signals V₁(t) and V₂(t) are heavily corrupted by the additional noise. The noise comes from surrounding environment and from the measurement system errors. Moreover, real signals are often unsteady (nonstationary) and random (stochastic). That is why the different methods have been taken under consideration and the received results have been compared. An analysis has been made for time and frequency domain as well. Due to the planned application of the obstacles detection system in the unmanned underwater vehicle, the algorithm had to be feasible for implementation in digital signal processor.

Recently a new type of autonomous underwater vehicle uses artificial fins to imitate the movements of marine animals, e.g. fish. These vehicles are biomimetic and their driving system is an undulating propulsion. There are two main methods of reproducing undulating motion. The first method uses a flexible tail fin, which is connected to a rigid hull by a movable axis. The second method is based on the synchronised operation of several mechanical joints to imitate the tail movement that can be observed among real marine animals such as fish. This paper will examine the first method of reproducing tail fin movement. The goal of the research presented in the paper is to identify the parameters of the one-piece flexible fin kinematics model. The model needs further analysis, e.g. using it with Computational Fluid Dynamics (CFD) in order to select the most suitable prototype for a Biomimetic Underwater Vehicle (BUV). The background of the work is explained in the first section of the paper and the kinematic model for the flexible fin is described in the next section. The following section is entitled Materials and Methods, and includes a description of a laboratory test of a water tunnel, a description of a Vision Algorithm (VA)which was used to determine the positions of the fin, and a Genetic Algorithm (GA) which was used to find the parameters of the kinematic fin. In the next section, the results of the research are presented and discussed. At the end of the paper, the summary including main conclusions and a schedule of the future research is inserted.