基于信息融合的深海水下机器人组合导航方法研究
文献类型:学位论文
作者 | 刘本![]() |
学位类别 | 硕士 |
答辩日期 | 2016-05-25 |
授予单位 | 中国科学院沈阳自动化研究所 |
导师 | 刘开周 |
关键词 | 水下机器人 LBL/DR组合导航系统 ASRCKF滤波 支持向量机 粒子群优化算法 |
其他题名 | Research of Deep Underwater Vehicle Integrated Navigation System with fusing algorithms |
学位专业 | 模式识别与智能系统 |
中文摘要 | 本文以国家863计划课题 “4500米载人潜水器控制系统设计与关键技术研究” (No.2009AA093302)等科研课题为依托,对水下机器人组合导航系统以及导航融合算法进行了深入系统的研究。就目前发展情况而言,长基线定位系统(Long Base Line,简称LBL)和航迹推算系统(Dead Reckoning,简称DR)是目前应用广泛的导航方法。组合导航系统主要包括声学定位设备、运动传感器、多普勒计程仪以及深度计等设备。本文对水下机器人导航传感器、水下机器人导航系统结构以及传统导航融合算法进行了研究。期望能够提出更加有效的方法为水下机器人提供精确的导航信息,最终能够为水下机器人的安全航行和有效作业提供必要、有效的信息。本文主要研究内容如下:1)本文首先对水下机器人LBL/DR组合导航系统进行了研究。并分别对LBL系统和DR系统的基本工作原理和误差产生原因进行了分析、研究。2) 针对水下机器人组合导航系统非线性和以及该系统在海底工作时容易受到噪声干扰的问题,本文深入研究了扩展卡尔曼滤波(Extended Kalman Filter,简称EKF)、无色卡尔曼滤波(Unscented Kalman Filter,简称UKF)、容积卡尔曼滤波(Cubature Kalman Filter,简称CKF)以及平方根CKF(Square-root Cubature Kalman Filter,简称SRCKF)四种非线性融合方法。在此研究的基础上引入了自适应SRCKF滤波(Adaptive SRCKF,简称ASRCKF)方法,该算法能够在线估计非线性系统噪声,并将其加入到滤波运算当中。通过数值仿真实验比较了自适应SRCKF方法和标准CKF算法的优缺点,实验证明本文引入的自适应SRCKF相比传统KF算法,在估计精度和计算速度方面均有一定程度提高。3) 针对由于LBL系统定位信息更新率过低,导致基于卡尔曼算法的组合导航系统稳定性和精确性不高的问题,研究了支持向量机(Support Vector Machine,简称SVM)及其他机器学习算法。并在研究的基础上,将SVM算法引入水下机器人组合导航框架当中,提出了基于自适应SRCKF和SVM算法的混合导航方法。同时,为了优化SVM算法的性能,本文引入了粒子群优化(Particle Swarm Optimization,简称PSO)算法对SVM参数进行寻优配置。最后,通过“蛟龙”号海试导航数据进行的实验证实了算法的优越性。 |
英文摘要 | This thesis is supported by National 863 key project of China (No.2009AA093302), and other projects on HOV. A deep and scientific research in UV navigation system is done in this thesis. Long base Line (LBL) positioning system and Dead Reckoning (DR) system are the two widely used navigation systems. The UV navigation system nowadays includes acoustic positioning devices, motion sensor, Doppler Log and depth log. This thesis aims to find and propose a proper method to fuse those various sensors and proved precise and stable navigation data for UV. The main contents of this thesis are as follows, 1) The LBL/DR integrated system of UV is studied. Then LBL positioning system and Dead Reckoning system are studied in depth. The working principle and errors of both systems are analyzed in theory. 2) To deal with the nonlinear UV navigation system and reduce the impact of underwater environment noise, nonlinear filtering algorithms including extended Kalman filter (EKF), unscented Kalman filter (UKF) ,cubature Kalman filter (CKF) and SRCKF are introduced and studied. Then an adaptive SRCKF is presented by bring the maximum a posterior (MAP) estimator into SRCKF and a numerical experiment is done to verify this method. 3) Due to the low update rate of LBL data, the traditional fusing frames based on Kalman filter method are not so stable and precise. Therefore, machine learning (ML) algorithms, such as support vector machine (SVM) are studied. ML algorithms are now very popular and effective in fusing different kinds of sensors. Then a new fusing frame based on SVM and Adaptive SRCKF is presented here to improve the performance of LBL/DR. Also, PSO algorithm is adopted to improve the performance of SVM. Experiments based on real data from “Jiaolong” HOV are conducted to verify the performance of this method presented here. |
语种 | 中文 |
产权排序 | 1 |
页码 | 52页 |
源URL | [http://ir.sia.cn/handle/173321/19650] ![]() |
专题 | 沈阳自动化研究所_水下机器人研究室 |
推荐引用方式 GB/T 7714 | 刘本. 基于信息融合的深海水下机器人组合导航方法研究[D]. 中国科学院沈阳自动化研究所. 2016. |
入库方式: OAI收割
来源:沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。