Deep learning/Kalman filter-based underwater localization in VLC systems

Abstract

There is a huge importance for the localization system in underwater visible light communication (VLC) systems as in petroleum, military and diving fields. To enhance the localization system, we use the Kalman filter (KF) algorithm with average received signal strength (RSS) method to obtain the nearest estimated positions. In this paper, two channel modeling weighted double Gamma functions (WDGF) are applied and a combination exponential arbitrary power function (CEAPF) for enhancing localization in VLC underwater systems. Using the proposed KF enhances the localization by ~ 60% as compared to the than average RSS technique for WDGF channel modeling and ~ 78% for the CEAPF channel modeling. Based on the estimate of received signal strength (RSS) by deep learning models (DLMs), underwater localization utilizing VLC is introduced. Our proposed framework is categorized into two phases. First, data collection is collected based on MATLAB software. Second, the training and testing of DLMs, SSD, RetinaNet, ResNet50V2 and InceptionResNetV2 techniques are applied. The channel gains are the DLMs’ input data set, while the DLMs’ output is the RSS intensity technique coordinates for each detector. The DLMs are then developed and trained using Python software. The ResNet50V2 based on average RSS technique hybrid with KF in CEAPF channel model achieves 99.98% accuracy, 99.97% area under the curve, 98.99% precision, 98.88% F1-score, 0.101 RMSE and 0.32 s testing time.