Underwater localization system based on visible-light communications using neural networks

Area of Science:

• Optical Communications
• Underwater Navigation
• Machine Learning

Background:

• Underwater localization is crucial for various applications.
• Visible-light communication (VLC) offers a promising alternative for underwater data transmission.
• Accurate positioning in underwater environments remains a challenge.

Purpose of the Study:

• To propose and evaluate a novel underwater localization system.
• To leverage neural networks (NNs) for received signal strength (RSS) based positioning.
• To assess the performance and robustness of the proposed NN model.

Main Methods:

• Data collection using Zemax OpticStudio Monte Carlo ray tracing software with 40,000 receivers.
• Channel gain measurement in seawater to create input datasets for NNs.
• NN system development and training using Orange data mining software.

Main Results:

• Achieved high performance metrics: 99.1% Area Under the Curve (AUC), 98.7% classification accuracy (CA), F1 score, precision, and recall.
• Logloss of 7.3% and specificity of 99.3% indicate a robust and accurate model.
• Optimized NN parameters including training algorithms, activation functions, and neuron count.

Conclusions:

• The proposed NN-based underwater localization system demonstrates high accuracy and robustness.
• Visible-light communication combined with NNs is effective for underwater positioning.
• The study provides a validated methodology for improving underwater navigation systems.