【作者】 李兵；

【导师】 李世银；

【作者基本信息】 中国矿业大学 ， 电子信息（专业学位）， 2023， 硕士

【摘要】 可见光定位(Visible Light Positioning,VLP)系统利用广泛部署的发光二极管(Light Emitting Diode,LED)可以同时提供高精度定位和照明服务,因其在众多室内位置感知服务中的关键作用而备受关注,其中单基站VLP是实际应用中一个常见重要场景。然而单基站三维移动定位目前研究较少,同时接收机用户设备(User Equipment,UE)的朝向随机变化对定位精度的影响不可忽视,亟待更进一步地研究。针对上述问题,本文提出单基站LED下,基于多光电二极管(Photodiode,PD)接收信号强度(Received Signal Strength,RSS)和UE朝向任意的VLP系统。本文分别利用优化理论和机器学习方法设计了面向移动用户的单基站三维VLP算法和可稳定实现厘米级定位的集成化原型机。本文工作内容如下:1、针对实际场景下LED部署无法满足VLP系统有效接收多LED信号的难题,基于单基站LED下多PD的RSS,将UE任意朝向的三维VLP问题表示为非线性最小二乘(Non-Linear Least Square,NLS)的优化问题。针对NLS问题的非凸性,基于Wolfe-Powell规则改进了序列二次规划(Sequential Quadratic Programming,SQP)算法的步长更新规则,进而设计了有效的定位解算方法。经过实际测试,当UE始终垂直向上时,提出的方法平均定位误差为16.8 cm,对于UE任意朝向场景,平均定位误差为26.36 cm。2、针对采用SQP算法求解NLS问题时可能陷入局部最优解的问题,并为了进一步提高定位精度,设计实现了基于随机梯度下降的人工神经网络(Artificial Neural Network Stochastic Gradient Descent,ASGD)定位算法。ASGD算法的输入为多PD的RSS和UE自身朝向信息,输出为UE的估计位置。在UE任意朝向场景和UE垂直向上场景中,ASGD定位算法的平均定位误差分别不超过1.77 cm和0.7 cm。3、进一步,针对实际场景下可见光通信定位技术未实现集成小型化的问题,设计了轻量化VLP定位算法,结合可见光通信技术(Visible Light communication,VLC)搭建了可见光通信定位一体化(Visible Light Positioning and Communication,VLPC)系统原型机,其中采用微控制器控制基站LED发送通信定位信息,UE搭载ASGD定位算法和通信信号处理算法。通过采用自定义帧结构传输协议,解决多基站间的信息干扰问题,实现了将多个单站VLPC系统组建成VLPC网络。更多还原

【Abstract】 Visible Light Positioning(VLP)systems utilize widely deployed Light Emitting Diode(LED)to provide both high-precision positioning and illumination services.VLP systems are attracting attention because of the key role playing in many indoor location-aware services.The single-base station VLP is a common and important scenario in practical applications.However,there are few studies on single-base station 3D mobile positioning.Besides the impact of random changes in receiver user equipment(UE)orientation on positioning accuracy cannot be ignored.Thus the further research is urgently needed.For the above problems,the thesis proposes a VLP system based on multiple Photodiodes(PDs)Received Signal Strength(RSS)and UE orientation arbitrary under single-base station LED.A single-base station 3D VLP algorithm for mobile users and an integrated prototype that can stably achieve centimeter-level positioning are designed based on the optimization theory and machine learning methods.The works of the thesis are as follows:1.It is hard for LED deployment under practical scenarios to meet the VLP system to effectively receive multi-LED signals.For this problem,the 3D VLP problem with arbitrary UE orientation is expressed as a non-linear least squares(NLS)optimization problem based on the RSS of multiple PDs under a single-base station LED.To address the nonconvexity of the NLS problem,the thesis improves the step update rule of the Sequential Quadratic Programming(SQP)algorithm based on the Wolfe-Powell rule,and then designs an effective localization solution method.After practical tests,the average positioning error of the proposed method is 16.8 cm when the UE is always vertically upward,and 26.36 cm for the UE arbitrarily oriented scenario.2.To further improve the positioning accuracy,the Artificial Neural Network Stochastic Gradient Descent(ASGD)positioning algorithm is designed and implemented in the thesis to address the problem that the SQP algorithm may fall into local optimal solutions when solving the NLS problem.The inputs of the ASGD algorithm are the RSS of the multi-PDs and the UE’s orientation information,and the outputs are the estimated positions of the UE.The average positioning error of the ASGD positioning algorithm does not exceed 1.77 cm and 0.7 cm in the UE arbitrary orientation scene and the UE vertical up scene,respectively.3.Further,it is aimed at the problem that the visible light communication and positioning technology has not achieved integration and miniaturization in actual scenarios.The lightweight VLP positioning algorithm designed in the thesis is combined with Visible Light communication(VLC)technology to build an Visible Light Positioning and Communication(VLPC)system prototype,in which a microcontroller is used to control the base station LED to send communication and positioning information,and the UE is equipped with ASGD positioning algorithm and communication signal processing algorithm.By adopting a custom frame structure transmission protocol,the problem of information interference between multiple base stations is solved,and multiple single-base-station VLPC systems are formed into a VLPC network.更多还原