Hybrid Heuristic-aided Multi-scale Deep Learning Model for Maximizing the Received Signal Strength in Visible Light Communication

Abstract

LED-based Visible Light Communication (VLC) technology has excellent positional accuracy; according to some articles, it can reach centimetre-level accuracy. VLC is unique because it uses a lot of unlicensed bandwidth and is naturally energy-efficient. Notably, VLC proves suitable in environments like hospitals, aircraft and mining zones, where concerns about electromagnetic contamination arise. As a result, indoor localization has identified a promising alternative: Indoor Visible Light Positioning (VLP), which is a hinging on VLC techniques. The implications extend to its potential integration in safety and monitoring systems within nuclear mining operations, where robust, non-RF communication is beneficial. It's important to note that theoretical evaluations of the accuracy of indoor VLC positioning systems based on the Received Signal Strength Indicator (RSSI) algorithm are still lacking. Some research projects have looked into integrating deep learning to improve translation procedures. A deep learning approach is proposed for VLC networks to accomplish effective performance. The main aim of this system is to maximize the RSS in the network. Firstly, the data attributes regarding the signal are collected and fed into the model of a Multiscale One-Dimensional Convolutional Neural Network with Symmetric Convolution (M1D-CNN-SC), in which the hyperparameters are optimally selected using the Hybrid Position of Barnacles and Grasshopper Algorithm (HPBGA). The enhanced model is estimated using diverse metrics in contrast with other approaches. Thus, the findings illustrate that it attains impressive results in increasing the RSS for locating the LED effectively.