《多传感器系统的网络化实现及信息融合算法研究论文论文》由会员分享，可在线阅读，更多相关《多传感器系统的网络化实现及信息融合算法研究论文论文（64页珍藏版）》请在金锄头文库上搜索。
1、国内图书分类号：TP273 UDC：681.5学校代码：10213 密级：公开工学硕士学位论文多传感器系统的网络化实现及信息融合算法 研究硕 士 研究生： 郭亮导师： 刘国良副教授申 请 学 位： 工学硕士学 科 、 专 业： 控制科学与工程所 在 单 位： 航天学院答 辩 日 期： 2011 年 6 月授予学位单位： 哈尔滨工业大学Classified Index: TP273 U.D.C: 681.5Dissertation for the Master Degree in EngineeringDEVELOPMENT OF NETWOKED MULTI-SENSOR SYSTEM AND
2、 RESEARCH ON INFORMATION FUSION ALGORITHMCandidate:Supervisor:Academic Degree Applied for:Specialty:Affiliation:Date of Defence:Degree-Conferring-Institution:Guo LiangAssociate Prof. Liu GuoliangMaster of EngineeringControl Science and EngineeringSchool of AstronauticsJune, 2011Harbin Institute of T
3、echnology哈尔滨工业大学工学硕士学位论文摘要随着 Internet 的发展和普及，基于 Internet 的有线或无线传感器网络技术已经 成为工业和民用领域实现远程监测和控制的重要手段。现有的以 RS232/RS485 协 议以及以工业现场总线为传输手段的多传感器系统已经远远不能满足现代工业系统远程监测和控制的要求，目前，基于 Internet 的多传感器网络化的监测和控制技术呈现出较高需求态势，特别是物联网技术在中国的迅速发展。因此开发基于Internet 的远程监测和控制系统，并能兼容现有的 RS232/RS485 以及现场总线系统是势在必行的。其次，对于一个复杂的多传感器系统，在
4、进行网络传输之前有必要对多传感器的数据进行融合，以缓解网络系统的负荷。对于基于多传感器系统的智能车辆，以及空天飞行器，最主要的融合手段仍然是卡尔曼滤波技术，但是对于较复杂的系统，随着状态变量维数的增加，集中式卡尔曼滤波算法的计算量会呈指数上升。针对上述情况，本项目给出了相应的解决方案，主要包括以下研究内容：(1) 通过对常用的嵌入式 Internet 实现方法的学习，确定了微处理器结合以太网控制器的设计方案，并根据该方案设计了以 ATmeaga128 为控制核心，以RTL8019AS 为以太网控制芯片的数据转换器。通过对比常用的嵌入式操作系统，确定了以 NUT/OS 操作系统为开发平台，并在该
5、平台的基础上开发了满足工程需求的数据转换程序。使用该系统用户可以通过 Internet 实现对现场的监控。(2)对于一个复杂的多传感器系统，在进行网络传输之前有必要对多传感器的 数据进行融合，以缓解网络系统的负荷。本文对基于多传感器的分布式信息融合算法进行了研究，主要研究了联邦卡尔曼滤波算法。通过对 INS/GPS/SS 组合导航系统建模，使用联邦卡尔曼滤波算法对 INS，GPS 和星敏感器多传感器的数据进 行了融合处理，通过仿真实验验证了联邦滤波算法的有效性。(3) 其次，针对复杂导航系统中联邦卡尔曼滤波器的各局部滤波器在高维状态变量情况下滤波计算量大的弱点，在各个局部滤波器中使用 EM 算
6、法，即在 EM 算法的 E 步使用卡尔曼滤波算法给出公共状态变量的估计，在 EM 算法的 M 步给出各个局部滤波器特有的误差状态变量的更新，M 步的更新过程可以和主滤波器进 行公共状态变量的融合过程同时进行，从而节省了局部滤波器的滤波时间，最后通过仿真实验验证了该 EM-FKF 算法在减少计算量方面的优越性。关键词:传感器网络；多传感器融合；联邦卡尔曼滤波算法；EM 算法； EM-FKF 算法I哈尔滨工业大学工学硕士学位论文AbstractWith the development and popularity of Internet, wired or wireless sensor netw
7、ork technology based on the Internet has becomed an important means of romote monitoring and control in industrial and domestic field. The existing RS232/RS485 protocol and industrial field bus for the main transmission of multisensor system cannot satisfy modern industrial monitoring and control re
8、quirements.At present,the demond for the technology of multi-sensor network based on Internet for monitoring and control shows an upword trend.Therefore the research and development of the technology of multi-sensornetworkbased onInternetformonitoringandcontrol is urgent.Secondly,for the purpose of
9、easing the load of network, it is necessary to intergrate the multi-sensor data before making the network transmission for a complex multi-sensor system.For vehicle tracking system based on mutli-sensor system,the traditional methord of information fusion is Kalman filter.But for more complex system
10、s, the computation of Kalman filter will increase exponentially with the increase of the dimention of state variables.Based on the discussion of above, this project provides the corresponding solutions. This project includes the following: (1) The embedded Internet implementations are learned and th
11、e method of microprocessor plus Ethernet controller is choosed.This project relized the design of Data conversion board in which ATmeaga128 is the control core and RTL8019AS is the Ethernet controller.Learned and compared some popular Embedded operating systems and choosed NUT/OS operating system as
12、 the Development Platform.Developed the data conversion programs which meet the engineering requriements based on the operating system. Users can use this system to achieve on-site monitoring through Internet. (2) For a complex multi-sensor system, in order to ease the load on the network system,it
13、is necessary to make data fusion.The centralized Kalman Filter(CKF) algorithm can result in a high computional load and a low levelel of fault-tolerance.We do some research on Federal Kalman Filter(FKF) algorithm which can used to integrate the information from multi-sensor systmes.The Federal Kalma
14、n Filter(FKF) algorithm is used on the model of INS/GPS/SS intergrated navigation sysytem to integrate the data from this three different sensors.Finally,some simulatios are done toprove that the Federal Kalman Filter(FKF) algorithm is effective. (3) Secondly, the computional burden increase sharply
15、 when Federal Kalman Filter(FKF) algorithm is used to deal with high-dimentional state variables in loacl filters.So we propose using an EM algorithm to improve Federal Kalman Filter(FKF) algorithm performance.More precisely,the common states of local filters are estimated in the E-step.The own uniq
16、ue sensor bias states of local filters are updated in the M-stepII哈尔滨工业大学工学硕士学位论文and the M-step is performed with the fusion of common states in the master filter simultaneously.Finally,some simulatios are done to prove that the computational burden is reduced.Keywords: Sensor network, multi-sensor fusion, Federal Kalman filter, Expectation Maximization algorithm(EM), Expectation Maximization-Federal Kalman Filter algorithm(EM-FKF).III哈尔滨工业大学工学硕士学位论文目录摘要 .
