藉由資料融合及交互式多模型卡門濾波技術開發車輛精準定位系統

Development of a Precision Road Vehicle Positioning System through Data Fusion and Interactive Multiple Model-Kalman Filter Techniques

吳棖立(Cheng-Li Wu)

指導教授 ： 李綱

國立臺灣大學/工學院/機械工程學研究所/碩士(2012年)

https://doi.org/10.6342/NTU.2012.10815

摘要

本論文探討使用資料融合(data fusion)技術，以整合全球衛星導航系統(Global Navigation Satellite System, GNSS)、常用車輛運動感測器(如:陀螺儀、速度計、電子羅盤)與電子地圖，開發車輛精準定位系統，以提供車輛主動式安全系統(Active Safety System)及通訊為基礎之協同式安全系統(Communication-based Cooperative Safety System)開發使用。傳統上使用全球定位系統(Global Positioning System, GPS)及低價之慣性導航感測系統(Inertial Navigation System, INS)所開發之車輛定位系統無法提供高可靠度、高精確度之車輛即時定位功能，特別是在高樓大廈集中的城市峽谷(urban canyons)、隧道或是山區等地方，GPS/INS定位系統受GPS訊號遮蔽、中斷或多重路徑誤差(multi-path errors)之影響，導致定位系統性能大幅降低。本研究嘗試運用資料融合之概念，整合多重感測器及電子地圖資訊，以提升GPS/INS車輛定位系統之精準度與可靠度。
我們所提出之車輛精準定位技術主要可分為兩大部分：1.以交互式多模型擴展式卡門濾波(Interactive Multiple Model–Extended Kalman Filter, IMM-EKF)技術整合GPS和低價慣性導航感測器資料，提供初步車輛位置與航向角估測值; 2.運用地圖匹配(Map Matching)技術、電子地圖和電子羅盤感測器資料，可進一步修正車輛航向角與側向定位誤差。
本研究所開發之車輛精準定位系統於台灣大學校園內進行測試驗證，實驗結果顯示IMM-EKF based GPS/INS可即時偵測車輛之運動狀態(縱向或側向運動)，有助於確認使用圖資校正定位與航向角誤差之時機，並可在車輛直行時避免偏航角陀螺儀(yaw gyro)之誤差影響GPS/INS之定位精確度；此外，電子羅盤提供之方位角資訊可彌補GPS heading於低速下之誤差，並可結合電子地圖提供之道路方位角及EKF-based GPS/INS方位角估測值，以求取最佳之車輛方位角估測值，有效提升整體定位系統之精確度。

關鍵字

資料融合 ； 交互式多模型理論 ； 全球定位系統 ； 慣性導航系統 ； 電子地圖

並列摘要

This thesis documents the study on the data fusion techniques with the aim to integrate the Global Navigation Satellite System (GNSS), common vehicle motion sensors (e.g., gyro, speed sensor and electronic compass), and the digital map to develop a high-precision vehicle positioning system for vehicle active safety and communication-based cooperative safety applications. The conventional vehicle positioning system by integrating the Global Positioning System (GPS) and the low cost inertial navigation system (INS) cannot achieve real-time precision positioning with high reliability and data integrity, especially in urban canyons, tunnels or mountains, where the GPS is often severely degraded by signal blockages, outages or multipath errors. By exploiting the concept of data fusion which, in principle, aims for information/data redundancy, cooperation and complementarity, this research attempts to fuse the information/data from multiple sensors and the digital map to enhance the accuracy and reliability of the GPS/INS positioning system.
The proposed vehicle precision positioning techniques include two parts: Firstly, the Interactive Multiple Model-Extended Kalman Filter is applied to integrating the GPS and economical dead-reckoning (DR) sensors to provide the preliminary vehicle position and heading angle estimate. Secondly, by making use of the map matching (MM) technique, digital map and the electronic compass data, the errors in the heading angle and lateral position estimate can be further reduced.
The ground vehicle precision positioning system developed in this study was tested and validated on the campus of the National Taiwan University. Experimental results show that the IMM-EKF based GPS/INS can detect the motion of the host vehicle (lateral or longitudinal motion) in real-time, which is helpful for determining the timing to apply map corrections to the position and heading angle estimate and preventing bias errors in the yaw gyro data from affecting the accuracy of the GPS/INS system. In addition, the heading angle measurement provided by the electronic compass can compensate for GPS heading angle errors at low speeds, and it can be combined with the GPS/INS heading angle, and the road segment azimuth to obtain an optimal vehicle heading angle estimate, so that the overall accuracy of the positioning system can be significantly improved.

並列關鍵字

Data Fusion ； Interactive Multiple Model Theorem ； Global Positioning System ； Inertial Navigation System ； Digital Map

參考文獻

[1] Hao Xu, Hongchao Liu, Chin-Woo Tan, and Yuanlu Bao, “Development and

Application of an Enhanced Kalman Filter and Global Positioning System Error Correction Approachfor Improved Map Matching”, Journal of Intelligent Transportation Systems, 14(1):27–36, 2010.

[2] Kang Li, Han-Shue Tan, and J. K. Hedrick, “Map-Aided GPS/INS Localization

Using a Low-Order Constrained Unscented Kalman Filter”, Decision and Control, 2009 held jointly with the 2009 28th Chinese Control Conference. CDC/CCC 2009. Proceedings of the 48th IEEE Conference on 15-18 Dec. 2009.

[3] Kang Li, Han-Shue Tan, and J. K. Hedrick, “An Enhanced GPS-Based Vehicle Positioning System through Sensor Fusion of Digital Map Data”, in Proc. 9th International Symposium on Advanced Vehicle Control, Kobe, Japan, Oct. 6-9, 2008.