An Indoor Visible Light Positioning System Using Artificial Neural Network

We propose a visible light positioning system based on an artificial neural network (ANN) and optical camera communications. The receiver's position is approximately and precisely estimated based on the decoded block coordinate and a typical back propagation ANN, respectively. The experimental results show that the proposed scheme offers a mean positioning error of 1.49 cm, which is required in many indoor positioning scenarios where high accuracy is essential.

I. Introduction

Indoor positioning systems (IPSs) have attracted a considerable amount of attentions in recent years in a range of applications including highly automated manufacturing, shopping stores, medical environments, underground etc. The radio frequency (RF) based techniques such as zigbee, ultra wideband (UWB), bluetooth and RF identification (RFID) have been widely studied to provide indoor positioning with accuracy from tens of centimeter to several meters [1]–[2]. However, these methods are prone to high signal attenuation and multipath effects. However, there is an alternative IPS for indoor environments, which is based on visible light communications (VLC) that uses the already installed light emitting diodes (LEDs) lighting fixtures [3]. The VLC based IPS, which offers advantages such as not affected by the RF electromagnetic interference, high positioning accuracy, and low cost front-ends have been widely investigated [4]. Since in VLC-IPS the transmission data rate is not an issue, therefore both cameras and photo-detectors (PDs) based receivers (Rxs) are used.