摘 要
初始對(duì)準(zhǔn)是慣性導(dǎo)航系統(tǒng)的關(guān)鍵技術(shù)之一，對(duì)準(zhǔn)的精度直接影響慣導(dǎo)系統(tǒng)的工作性能。濾波技術(shù)在初始對(duì)準(zhǔn)中具有極其重要的作用，對(duì)系統(tǒng)精度具有至關(guān)重要的影響。
實(shí)際的慣導(dǎo)系統(tǒng)尤其是捷聯(lián)慣導(dǎo)系統(tǒng)，由于工作環(huán)境非常惡劣，同時(shí)又受到外部各種因素的干擾，其初始失準(zhǔn)角往往比較大，誤差模型呈現(xiàn)非線性，因此需要非線性的濾波方法，而且在使用濾波方法進(jìn)行初始對(duì)準(zhǔn)時(shí)，存在著觀測(cè)數(shù)據(jù)延時(shí)的情形，并且由于系統(tǒng)的隨機(jī)性，觀測(cè)信號(hào)出現(xiàn)延時(shí)的可能性也是隨機(jī)的。因此研究觀測(cè)信號(hào)隨機(jī)延時(shí)下的非線性濾波方法，并將其應(yīng)用于捷聯(lián)慣導(dǎo)系統(tǒng)的初始對(duì)準(zhǔn)，進(jìn)一步提高系統(tǒng)性能，具有十分重要的意義。
本文對(duì)觀測(cè)信號(hào)隨機(jī)延時(shí)下的捷聯(lián)慣導(dǎo)系統(tǒng)初始對(duì)準(zhǔn)進(jìn)行了研究，在建立觀測(cè)信號(hào)隨機(jī)一步延時(shí)模型的基礎(chǔ)上，推導(dǎo)了非線性離散系統(tǒng)觀測(cè)信號(hào)隨機(jī)一步延時(shí)的擴(kuò)展濾波和Unscented濾波方法，并將其應(yīng)用于捷聯(lián)慣導(dǎo)系統(tǒng)動(dòng)基座初始對(duì)準(zhǔn)中。
本文的主要研究工作和成果可歸納如下：
(1) 闡述了捷聯(lián)慣導(dǎo)系統(tǒng)初始對(duì)準(zhǔn)基本原理及捷聯(lián)慣導(dǎo)系統(tǒng)誤差產(chǎn)生的原因，詳細(xì)推導(dǎo)了慣性傳感器誤差模型，速度誤差方程，姿態(tài)誤差方程和位置誤差方程。
(2) 建立了帶加性噪聲的非線性離散隨機(jī)系統(tǒng)觀測(cè)信號(hào)隨機(jī)一步延時(shí)的非線性模型，討論了其濾波問題，推導(dǎo)了基于線性化系統(tǒng)方程的擴(kuò)展濾波方法，給出了其濾波方程。
(3) 在介紹UT理論、UT采樣策略的基礎(chǔ)上，推導(dǎo)了基于SUT的觀測(cè)信號(hào)隨機(jī)一步延時(shí)的Unscented濾波方法，并采用SSUT對(duì)Unscented濾波方法進(jìn)行改進(jìn)，得到了基于SSUT的SSUF濾波方法。
(4) 建立了捷聯(lián)慣導(dǎo)系統(tǒng)大方位失準(zhǔn)角下動(dòng)基座非線性對(duì)準(zhǔn)模型，包括系統(tǒng)狀態(tài)方程和觀測(cè)信號(hào)隨機(jī)一步延時(shí)的系統(tǒng)觀測(cè)方程，將擴(kuò)展濾波方法和Unscented濾波方法應(yīng)用于初始對(duì)準(zhǔn)中，對(duì)濾波算法進(jìn)行仿真對(duì)比，實(shí)驗(yàn)結(jié)果表明，Unscented濾波方法具有更高的對(duì)準(zhǔn)精度和更快的收斂速度，Unscented濾波的改進(jìn)算法與Unscented濾波精度相當(dāng)，但是計(jì)算量大大降低。

關(guān)鍵詞 觀測(cè)值隨機(jī)延時(shí)；擴(kuò)展卡爾曼濾波；Unscented卡爾曼濾波；捷聯(lián)慣導(dǎo)系統(tǒng)；初始對(duì)準(zhǔn)





Abstract
Initial alignment is one of the key technologies of the inertial navigation system, it’s alignment accuracy directly affect the navigation system working performance. Filtering technique plays an extremely crucial role in the initial alignment and has an important effect to the system accuracy.
Actual inertial system always works in rugged environment and often suffers interferences from various kinds of fators. Especially to the strap-down inertial navigation system (SINS), the intial heading error angle is relatively large and the error model is generally nonlinear. The nonlinear error model needs nonlinear filtering methods. In SINS initial alignment, there exist situation in which the arrival of the measured data is delayed and due to the system of uncertainty, the possibility of delay in each measurement is random. Therefore, it is very significant to research the nonlinear filtering methods using randomly delayed observations and it’s application in SINS initial alignment in roder to improve the system performance.
In this paper, SINS initial alignment with randomly delayed observations is researched. On the base of establishment of the model with one-step randomly delayed observations, Extended Filter and Unscented filter for nonlinear discrete-time systems using one-step randomly delayed observations contaminated by additive white noise are developed and are applied to initial alignment of SINS.
The main research work in this thesis can be summarized as follows:
Firstly, the basic theory of SINS initial alignment and the causes of SINS error are expatiated. The inertial sensor error models, velocity error equations, attitude error equations and position error equations of SINS are given respectively.
Secondly, the nonlinear model of nonlinear discrere-time stochasic system using observations with stochastic delays contaminated by additive white noise is established and it’s filtering promble is discussed. The Extended Filter (EF) based on linearizing system equations is derived and it’s filtering equations are given.
Thirdly, on the base of Unscented Transform (UT) theory and UT sampling strategy, Unscented Filter (UF) using one-step randomly delayed observations is derived and improved Unscented Filter based on SSUT is investigated.
Fourthly, the nonlinear alignment model of SINS on moving bases is established, including system state equation and system observation equation using one-step randomly delayed observations. The EF and UF are applied and simulated to initial alignment of SINS. The simulation results show that the UF has faster convergent rate and better aligment precision than EF. Improved filtering has the same aligment precision with UF and it’s computions are greatly reduced.

Key words randomly delayed observations; EFK; UKF; SINS; initial aligment

目 錄
摘 要 I
Abstract III
第1章 緒論 1
1.1 研究背景與意義 1
1.2 初始對(duì)準(zhǔn)技術(shù)的發(fā)展 2
1.3 濾波理論的發(fā)展 3
1.4 本文研究的主要內(nèi)容 5
第2章 捷聯(lián)慣導(dǎo)系統(tǒng)初始對(duì)準(zhǔn)基本原理 7
2.1捷聯(lián)慣導(dǎo)系統(tǒng)概述 7
2.1.1常用坐標(biāo)系的定義 7
2.1.2捷聯(lián)慣導(dǎo)系統(tǒng)工作原理 8
2.2 捷聯(lián)慣導(dǎo)系統(tǒng)初始對(duì)準(zhǔn) 9
2.2.1 初始對(duì)準(zhǔn)的特點(diǎn) 10
2.2.2 初始對(duì)準(zhǔn)的分類 10
2.2.3 初始對(duì)準(zhǔn)的原理 11
2.3 本章小結(jié) ..