Hardware and low-level control of biomimetic underwater vehicle designed to perform ISR tasks

Abstract

This paper conveys some of the results of a research project involving the design and tests of biomimetic underwater vehicle which is intended for implementation of intelligence, surveillance and reconnaissance tasks. The research presented in this paper is based on the theoretical and practical study. All major subsystems of the robot are specified and briefly described. The mechanical design is based on modular structure. Two types of modules, dry and wet, are specified. The major parts of the tail module were checked with the use of the finite element method. Moreover, hardware and software parts of robot's control subsystem and Simulink model of static depth regulator are briefly presented. The final part presents main parameters and basic functionality of the built biomimetic underwater vehicle.

Using Pretrained AlexNet Deep Learning Neural Network for Recognition of Underwater Objects

“…One of the software toolbox for deep learning is Neural Network Toolbox included in Matlab 2018 [10]. This paper undertakes problem of recognition of underwater objects using deep learning networks and training algorithms included in Matlab 2018.…”

“…The stochastic gradient descent with momentum algorithm uses a single learning rate for all the parameters. This algorithm can be defi ned as [10]:…”

“…The root mean square propagation algorithm uses a learning rates that are diff erent for diff erent parameters and that can automatically adapt to the loss function being optimized. This algorithm can be defi ned as [10]:…”

“…The derived from adaptive moment estimation (Adam) uses a parameter update that is similar to RMSProp with momentum [5]. The update is calculated based on the following equation [10]:…”

“…If gradients over many iterations are similar, the moving average of the gradient lets the parameter updates establish momentum in a certain direction. If the gradients contain mostly noise, then the moving average of the gradient and the parameter updates become smaller [10].…”