面向可见光无线定位的指纹推理模型校准  

Fingerprinting Inference Model Calibration Towards Visible Light-Based Positioning

作　　者：周炳朋 陈光森 王鑫 ZHOU Bing-peng;CHEN Guang-sen;WANG Xin(School of Electronics and Communication Engineering,Sun Yat-Sen University,Shenzhen,Guangdong 518107,China;Shenzhen Key Laboratory of Navigation and Communication Integration,Sun Yat-Sen University,Shenzhen,Guangdong 518107,China)

机构地区：[1]中山大学电子与通信工程学院,广东深圳518107 [2]中山大学深圳市导航通信一体化重点实验室,广东深圳518107

出　　处：《电子学报》2024年第6期1925-1937,共13页Acta Electronica Sinica

基　　金：国家自然科学基金(No.62001526);广东省自然科学基金(No.2021A1515012021);广东省重大人才工程青年项目(No.2021QN02X074);中山大学中央高校基本科研业务费专项基金(No.23QNPY22);深圳市科技计划(No.ZDSYS20210623091807023)。

摘　　要：针对可见光无线定位指纹推算技术遇到的指纹赋权模型校准难题,本文利用贝叶斯推断理论和随机优化方法,提出了一种高效的指纹赋权模型校准优化方法.首先,利用可见光接收信号强度及其对应的位置和姿态标签建立指纹数据库;其次,依据指纹观测模型及最大熵理论,构建高斯型指纹赋权模型,用于推算可见光无线终端的位置和姿态;然后,依据随机优化理论,将高斯赋权模型校准难题建模为随机优化问题,并设计了基于连续凸函数逼近的赋权模型优化算法解决其中的非凸函数优化难题,从而实现对指纹赋权模型参数的高效校准,提升可见光定位的指纹推算精度.通过对指纹赋权过程进行建模与优化,本文提出的指纹定位方法能够有效抑制非视距传播干扰、弥补指纹匹配误差,极大地提升了可见光指纹定位的精度.仿真数据表明,本文基于指纹赋权模型优化的可见光无线定位算法相比于主流的指纹定位方法具有更高的定位性能.In order to address the difficult fingerprinting-based visible light positioning(VLP)issue,an efficient fingerprinting interference model(FIM)calibration algorithm is proposed via leveraging Bayesian inference and stochastic optimization approaches.Firstly,a fingerprinting database is built by collecting received signal strength of visible light from various observation grids with known location and pose angles.Secondly,a Gaussian-form FIM is developed as per maximum entropy theory,and then FIM calibration is treated as a stochastic optimization problem.Finally,a successive convex approximation-driven optimization algorithm is proposed for calibrating FIM parameters by exploiting hidden convex substructures of FIM,thus improving the fingerprinting-based VLP performance.With our problem-specific algorithm design,the proposed FIM calibration-enhanced VLP method can alleviate the disturbance from non-line-of-sight propagation interference and fingerprinting model mismatch.It is verified by simulation results that our FIM calibration-enhanced VLP method outperforms the state-of-the-art baseline methods.

关 键 词：可见光无线定位 指纹模型 室内定位 非视距干扰 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]