無線定位追蹤訊號處理技術

Abstract

隨著位基服務商用價值俱增，即時定位演算法如何根據不同網路環境提升精準度是位基服務應用關鍵技術。本論文旨由統計訊號處理觀點提出適用於即時建置的無線定位追蹤演算法，並考慮以下無線環境：1)室外衛星定位系統；2)室內無線感測網路；3)手機蜂巢式系統。衛星定位系統在未被阻隔的室外環境中擁有較佳的效能。在衛星訊號微弱的環境中，無線感測網路適用於室內小範圍傳輸；手機蜂巢式系統可涵蓋前兩者訊號被遮掩或傳輸半徑不足的問題，並提供手機系統在資源管理中所需的基本精準度要求。本論文先以 4G 網路候選 IEEE 802.16m 系統討論定位精準度與其限制，作為如何在不同網路中啓動定位機制的範例。

為提供即時服務所需的定位資訊，本論文探討具封閉解特性、低演算複雜度的二階段最小方差法及卡曼濾波器。針對線性最小方差法推導其線性 Cramer-Rao lower bound (CRLB)，歸結出三項影響精準度的因素：1)幾何效應；2)訊號模型及雜訊變異性；3)定位訊號源數目。透過線性 CRLB 的分析，發現線性最小方差演算法手持裝置在訊號源所圍成幾何區域外的精準度因線性化產生偏差。本論文提出一幾何輔助線性定位演算(Geometry-assisted linearized localization, GALL) 彌補源於定位裝置與訊號的幾何效應所造成定位精準度差異。

以衛星及蜂巢式系統為例，考量兩系統不同的訊號模型及雜訊變異所提出的混合式定位架構來增進定位精準度。融合式 (Fusion-based hybrid, FH) 混合定位可有效整合兩系統定位結果，統一式(Unified-hybrid, UH) 混合定位架構包含混合訊號選擇及混合線性最小方差定位估測器，可採納異質網路訊號於單一估測器中。兩系統訊號結合可針對以下狀況增進精準度：1)市區：手機訊號密集但衛星訊號被阻隔；2)郊區：手機基地台訊號不足但衛星訊號品質佳。

當定位訊號源數目不足時，傳統定位追蹤演算法會因訊號源小於待估測參數的維度而導致無法定位的問題。針對行動裝置移動模型及透過速度與加速度的追蹤，以預測資訊作為幾何限制設計幾何輔助式預測性定位追蹤演算法 (Geometry-assisted predictive location tracking, GPLT)，在訊號源不足狀況下增進精準度。

根據合作式偵測，探討在混合直徑與非直徑環境中定位追蹤的問題。本論文提出一考量結合位置和通道狀況估測的自我導航演算法 (Cooperative self-navigation, CSN)，可有效結合其他行動裝置的量測訊號解決訊號源不足的問題，亦可在通道狀況追蹤有效整合不同訊號模型及雜訊變異。

With rising interests in location-based services (LBS) over the past decade, real-time localization algorithms with enhanced precision become critical for various applications under potentially challenging circumstances. Based on statistical signal processing theory, this dissertation solves the location estimation and tracking problems by incorporating the methods suitable for real-time implementation in different wireless network scenarios: 1) Global positioning system (GPS); 2) Wireless sensor network (WSN) system; and 3) Cellular-based positioning system. GPS is popular in commercial system and accurate in unobstructed outdoor environments. In environment where GPS coverage is either weak or absent, WSN can operate in a short-range indoor environment with high accuracy. On the other hand,cellular-based positioning based on the densely deployed base stations (BSs) in the city can cover the areas with weak GPS signal to maintain the LBS for resource management purpose. Note that the system level simulation of the fourth generation telecommunication system, e.g., IEEE 802.16m WiMAX system is first evaluated to discuss how to enable LBS and the performance limitation using the cellular system as an example.

In order to employ the location estimation and tracking for real-time services, closed-form signal processing techniques such as two-step least squares estimator and Kalman filter are adopted due to their lower computation complexities. Through the study of Cramer-Rao lower bound (CRLB), the limiting factors affecting the estimation accuracy include: 1) Geometric effect; 2) Signal model or noise variance; and 3) Number of signal sources. This dissertation first characterizes the linearization effect by the proposed linearized location estimation problem based CRLB (L-CRLB). As further suggested by the L-CRLB, higher estimation accuracy can be achieved if the mobile station (MS) is located inside the geometry confined by the BSs compared to the case that the MS is situated outside of the geometric layout. This result motivates the design of geometry-assisted linearized localization (GALL) algorithm in order to compensate the linearization lost from the geometric effect.

Hybrid location estimation schemes, which combine both the satellite- and cellular-based signal, are proposed to deal with the location estimation and tracking problem under various signal models or noise variances. The proposed fusion-based hybrid (FH) architecture integrates the estimation results acquired from both the satellite- and cellular-based systems. On the other hand, the unified hybrid (UH) architecture employs the proposed hybrid signal selection scheme and the hybrid least squares estimator, which is capable of conducting location estimation within a selected set of signal sources from the heterogeneous networks. The proposed hybrid location architectures can provide accurate location estimation by adapting themselves under different environments, e.g., urban or rural area.

The location estimator associated with the Kalman filter, known as the two-stage location tracking architecture, is exploited in this dissertation to acquire location estimation and tracking for the MS. However, most of the existing schemes become inapplicable for location tracking due to insufficient number of signal sources. The proposed predictive location tracking (PLT) scheme utilizes the predictive information obtained from the Kalman filter to provide additional signal inputs for location estimator. Furthermore, the geometry-assisted PLT (GPLT) scheme incorporates the geometric dilution of precision (GDOP) information into algorithm design to achieve persistent accuracy for location tracking.

The problem of cooperative localization for MSs in the mixed line-of-sight/non-line-of-sight (LOS/NLOS) environment is investigated based on cooperative sensing. The proposed cooperative self-navigation (CSN) with joint position and channel tracking takes advantage over the noncooperative methods with the extra cooperative measurements to overcome the insufficient number of signal sources problem and over the methods without LOS/NLOS channel tracking to consider the effect of different signal models or noise variances.

To summarize, the main contribution of this dissertation is to investigate the mobile location estimation and tracking problem. The methods suitable for real-time implementation is applied and analyzed based on statistical signal processing theory. In this dissertation, we have improved the estimation accuracy for real-time methods under three limiting factors as follows: 1) Geometric effect; 2) Signal model or noise variance; and 3) Number of signal sources.