Optimized deep learning/kalman filter-based underwater localization in VLC systems

Abstract

Several applications depend on the localization technique in underwater visible light communication (UVLC) systems, as military, petroleum, and diving fields. Recent research aims to develop the localization system by different methods to obtain the optimum position of the receiver. In this paper, we use Kalman Filter (KF) algorithm with average Received Signal Strength (RSS) technique using optimization. Optimized Deep Learning Models (DLMs) are utilized to improve the system performance, including such as ResNet50V2, InceptionResNetV2, SSD, and RetinaNet. Two channel modeling Weighted Double Gamma Function (WDGF) with a Combination Exponential Arbitrary Power Function (CEAPF) are used for sea water to enhance the UVLC localization system. The obtained results show that using CEAPF channel modeling with ResNetV2 strategy achieves the best accuracy of the localization for different methods. Also, the ResNetV2 outperforms other strategies for using RSS average technique. The RSS with KF and DLM achieves a higher accuracy with ResNetV2 than InceptionResNetV2, RetinaNet and SSD. Using WDGF achieves accuracy less than that in CEAPF where for using KF with average RSS method. Applying the RSS with KF with CEAPF channel modeling improves the performance than using WDGF. We use an automatic hyper-parameter (HP) approach to the Bayesian optimization models ResNet50V2, InceptionResNetV2, SSD, and RetinaNet. The ResNet50V2 based on average RSS technique hybrid with KF in CEAPF channel model achieves 99.99% accuracy, 99.99% area under the curve (AUC), 99.98% precision, 99.89% F1-score, 0.099 RMSE and 0.43 s testing time.