In recent times, several applications requiring highly accurate indoor positioning systems have been developed. Since the global positioning system is unavailable/less accurate in the indoor environment, alternative techniques such as visible light positioning (VLP) are considered. The VLP system benefits from the wide availability of illumination infrastructure, energy efficiency and the absence of electromagnetic interference. However, there is a limited number of studies on three-dimensional (3D) VLP and the effect of multipath propagation on the accuracy of the 3D VLP. This study proposes a supervised artificial neural network to provide accurate 3D VLP whilst considering multipath propagation using receiver diversity. The results show that the proposed system can accurately estimate the 3D position with an average root mean square (RMS) error of 0.0198 and 0.021 m for line-of-sight (LOS) and non-LOS link, respectively. For 2D localisation, the average RMS errors are 0.0103 and 0.0133 m, respectively.

中文翻译：

---

利用接收器分集和多层感知器神经网络进行精确的室内三维可见光定位

近年来，已经开发了需要高精度室内定位系统的几种应用。由于全球定位系统在室内环境中不可用/精度较差，因此需要考虑使用其他技术，例如可见光定位（VLP）。VLP系统得益于照明基础设施的广泛可用性，能效和无电磁干扰。但是，关于三维（3D）VLP以及多径传播对3D VLP精度的影响的研究数量有限。这项研究提出了一种有监督的人工神经网络，以提供准确的3D VLP，同时考虑使用接收器分集的多径传播。结果表明，对于视距（LOS）和非LOS链接，该系统可以准确估计3D位置，其平均均方根（RMS）误差为0.0198和0.021 m。对于2D定位，平均RMS误差分别为0.0103和0.0133 m。