Intelligently Applying Levenberg-Marquardt Algorithm to Optimize Mobile Positioning

• Main Authors: Cheng-PangLee, 李晟榜
• Other Authors: Jen-Fa Huang
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/54h5f5

碩士 === 國立成功大學 === 電腦與通信工程研究所 === 107 === The methods on using inverse matrix to calculate weighted geometric dilution of precision (WGDOP) have been widely used on communications positioning. Due to the complicated calculation of the inverse matrix, searching the optimal solution is time-consuming. Therefore, this thesis proposes searching the optimal solution by the artificial neural network (ANN) utilizing the computing ability of counterfeiting biological neural connection. In this thesis, we propose and use the program that can approximate the value of WGDOP by neural network with Levenberg-Marquardt algorithm (LMA). By selecting the base station set with the minimum of WGDOP to locate the mobile station position, it can reduce the effects of geometric distribution and improve positioning accuracy. In this thesis, we select the four base stations with minimum of WGDOP, and use time of arrival (TOA) to generate four measuring circles in cellular wireless communication system, and then estimate the position of MS by training circular intersection points with Levenberg-Marquardt algorithm. It will reduce the computational complexity and the location error. According to simulation and analysis result, this thesis firstly uses Levenberg-Marquardt algorithm to approximate the WGDOP, and then select base station set. Finally, train and get the positioning result by the Levenberg-Marquardt algorithm. It is confirmed having more accurate positioning accuracy and more efficient mobile station position estimation. In this thesis, the architecture we proposed in this research can be applied to the global positioning system, wireless sensing network and mobile communication system.