Intelligent and Practical Deep Learning Aided Positioning Design for Visible Light Communication Receivers

Visible light positioning (VLP) systems can achieve high positioning precision. However, they are not compatible with visible light communication (VLC) systems. They require special positioning modules and could not reuse functional communication modules, while requiring more than two light emitting diodes (LEDs) to be deployed at user ends. In order to address the issues of weak compatibility and high complexity of VLP, we present a novel position estimation deep neural network (PE-DNN) and propose to add a PE-DNN aided module at the VLC receivers. The proposed module firstly learns features of the VLC channel from received pilot signals implicitly, then it can estimate receivers' 2-dimension positions intelligently with a single LED. Accordingly, VLC systems can simultaneously provide positioning and information transmission services with only one LED and one photodiode (PD), thus the compatibility and the practicality are greatly improved. Simulation results show that the proposed system achieves a centimeter-level positioning accuracy, and can provide intelligent and practical positioning services for the users.

I. Introduction

Visible light positioning (VLP) and visible light communication (VLC) have attracted a lot of research interests, since VLP can provide higher positioning precision than traditional methods in indoor scenarios, while VLC can provide higher data rate and higher security over visible light bands, which have advantages of low electromagnetic interferences, low costs, high bandwidth and high security [1], than traditional communication systems.