# 臺灣博碩士論文加值系統

(100.24.115.215) 您好！臺灣時間：2022/05/23 20:26

:::

### 詳目顯示

:

• 被引用:1
• 點閱:265
• 評分:
• 下載:27
• 書目收藏:0
 卡爾曼濾波器是一種遞迴式最佳估測演算法，常被廣泛應用在整合式導航設計中；理想的卡爾曼濾波器建基於準確的動態模型，而雜訊統計特性是零均值的白雜訊，將可獲致最佳的濾波效果。若系統建模不準確且雜訊的統計模型不能掌握時，將導致濾波器發散，此時，應用交互多模型演算法能有效抑制濾波器的發散。由於在導航過程中系統雜訊和量測雜訊是非常複雜的，使得雜訊的不確定性導致擴展型卡爾曼濾波器(EKF)或無跡卡爾曼濾波器(UKF)的效能降低，結合交互多模型演算法及無跡卡爾曼濾波器(IMM-UKF)來解決雜訊不確定性問題是一可行的方法。但決定系統雜訊的傳統方法往往是依賴人為經驗或是藉由電腦計算，為了解決這個缺點，本論文提出模糊交互多模型無跡卡爾曼濾波器(FUZZY-IMMUKF)演算法，並應用於整合導航系統之設計中。這個方法，引入一模糊邏輯自適應系統(Fuzzy Logic Adaptive System)，藉由模糊推論適時調整系統雜訊變動之上、下界，再利用交互多模型演算法來決定載體運動之系統狀態。此外，無跡卡爾曼濾波器採用確定性採樣，是無導數(derivative-free)計算的演算法，能有效處理導航非線性之問題。實驗顯示，於整合導航設計中， FUZZY-IMMUKF 演算法較傳統之UKF和IMMUKF具有較優越之估計準確度。
 The Kalman filter, a kind of recursive optimal estimator, has been widely applied in the design of integration navigation. An ideal Kalman filter can lead to an optimal filter when dynamics of model is completely known and noise is zero mean and white. In case of imprecise system modeling or deficiency of noise statistics, filter will tend to be divergent. In this case, interactive multiple model can be adopted to suppress the divergence of filtering.Owing to process and measurement noises are very complex during maneuvering, uncertainties induced by noise seriously degrade the performance of Extended/Unscented Kalman filter. The combination of interacting multiple models algorithm and unscented Kalman filter becomes a feasible way to resolve the problem of uncertainties of noise. A conventional way to decide the quantity of process noise heavily depends on expert’s experience or resorts to test by try and error via computer. To overcome this inconvenience, a fuzzy interacting multiple models unscented Kalman filter (FUZZY-IMMUKF) is proposed in this thesis, and further demonstrated in the design of integrated navigation. A fuzzy logic adaptive system is introduced in this approach to timely adjust upper and lower bounds of the variation of process noise, then the states of the maneuver can be estimated by interacting multiple model unscented Kalman filter. Moreover, unscented Kalman filter adopts deterministic sampling and it is known as a derivative free algorithm capable of dealing with nonlinear filtering problem. The results show that the proposed FUZZY-IMMUKF algorithm has superior estimation accuracy as compared to UKF or IMM-UKF approach in the application of integrated navigation.
 摘要 IAbstract IIIContents VList of Figures VIIList of Tables XIChapter 1　INTRODUCTION 1§ 1.1 General 1§ 1.2 Thesis Outline 4Chapter 2　NAVIGATION SYSTEM 5§ 2.1 Introduction 5§ 2.2 Overview of the GPS 6§ 2.3 The mathematics of the GPS 8§ 2.4 Pseudorange errors 10§ 2.5 GPS/INS integration 13Chapter 3 The Nonlinear Kalman filter for navigation processing 15§ 3.1 The Extended Kalman Filter 15§ 3.2 The Unscented Kalman Filter 18Chapter 4 Introduction of Interacting Multiple Model 25§ 4.1 The Interacting Multiple Model (IMM) Algorithm 25§ 4.2 The IMM based on Unscented Kalman Filter 29§ 4.3 The Fuzzy Logic Adaptive System (FLAS) 32§ 4.4 The Fuzzy Adaptive Interacting Multiple Model Unscented Kalman Filter 34Chapter 5 Simulation and Experiments with Analyses 36§ 5.1 Using IMM-EKF in GPS navigation 37§ 5.1.1 Simulation of two dimensional trajectory 39§ 5.1.2 Simulation of three dimensional trajectory 47§ 5.2 IMM-UKF in GPS/INS integrated navigation 55§ 5.2.1 Simulation for scenario 1 58§ 5.2.2 Simulation for scenario 2 64§ 5.3 FUZZY-IMMUKF in GPS/INS integrated navigation 70§ 5.3.1 Simulation for scenario 1 71§ 5.3.2 Simulation for scenario 2 77§ 5.4 Experiment results 83Chapter 6 Conclusions and Future Work 93References 95