节点文献

基于联邦强跟踪卡尔曼滤波的组合导航关键技术研究

Research on Integrated Navigation Technology Based on Federated Strong Tracking Kalman Filter

【作者】 唐娟

【导师】 熊海良;

【作者基本信息】 山东大学 , 信息与通信工程, 2018, 硕士

【摘要】 导航定位技术伴随科技发展而产生。传统的导航系统虽然单独使用都能实现定位导航,但是都存在着一些缺陷,如全球定位系统(Global Positioning System,GPS)短时间内定位精度高,然而定位精度易受外界环境干扰;惯性导航系统(Inertial Navigation System,INS)是一种自主式导航系统,然而其定位误差会随着时间累积。为了充分发挥各系统的优势,取长补短,组合导航技术应运而生。常见的组合导航模型是GPS++INS,即用GPS辅助INS消除累积误差,从而实现精确导航定位。而本文提出了一种新的三传感器组合导航系统,同时利用多普勒计程仪(Doppler Velocity Log,DVL)的速度信息和GPS的位置信息来消除INS误差。这种新的组合导航模型能够解决GPS+INS模型在GPS接收不到信号时无法修正INS误差的问题,同时适用的范围更广,精度更高。另一方面,导航估计算法也是近些年来研究的热点。传统的估计算法有适用于线性模型的卡尔曼滤波(KalmanFilter,KF)算法,适用于非线性模型的扩展卡尔曼滤波(Extended Kalman Filter,EKF)算法,具有鲁棒性能的强Sage-Husa自适应KF算法以及适用于信息融合的联邦卡尔曼滤波(Federated Kalman Filter,FKF)算法等。为了更好的比较各算法的估计性能,本文首先提出了一个新的算法性能指标:残差方差,并基于正交性原理推导验证了该指标的可靠性。其次本文还分别推导了 KF算法的残差方差在三种非理想条件下包括系统模型不准确、初始值设置有误、状态存在突变的具体表达形式,通过推导比较发现这三种非理想情况下KF算法的估计性能都不佳。而强跟踪滤波(Strong Tracking Filter,STF)算法由于引入了渐消因子,能够保证残差无论在哪种非理想条件下都能满足正交性原理,即STF算法具有更强的估计性能,估计效果更好。最后本文建立了基于航位推算的单GPS导航仿真模型来验证这两种算法的性能,通过仿真实验可以证明在非理想条件下STF算法的性能优于KF算法性能。基于三传感器组合导航模型的建立及算法性能的分析,本文提出了一种基于鲁棒自适应联邦强跟踪卡尔曼滤波算法(Robust Adaptive Federated Strong Tracking Kalman Filter,RAFSTKF)的GPS+INS+DVL组合系统来进行高精度导航定位。该算法主要包括两部分:局部滤波和主滤波器信息融合,其能充分利用GPS+INS+DVL各自的观测信息,其中两个局部滤波器均采用STF算法,状态变量都是INS的误差,不同的是观测输入量分别是GPS与INS的位置误差和DVL与INS的速度误差。主滤波器首先采用无权重的最小二乘滤波来得到全局次优估计,即主滤波器的输入量直接融合,其次再采用基于直接融合的加权最小二乘自适应滤波来得到全局最优估计,即主滤波器输入量的自适应融合,其中各局部估计值与一步预测值的权重系数依据其与全局次优估计的差向量的二范数来确定。最后得到的全局最优估计反馈给INS修正其误差,实现精确定位。仿真实验证明本文提出的基于RAFSTKF的GPS+INS+DVL组合导航定位系统精度更高,性能更稳定。

【Abstract】 Navigation and positioning technology comes from the technological development.Although traditional navigation systems are basically able to realize the positioning goal alone,they still have drawbacks.For instance,positioning accuracy of GPS is heavily disturbed by external environments in spite of high precision during a short period.In addition,INS is an autonomous navigation system while its location error will accumulate with time.In order to give full play to the advantages of each system and offset their weakness,integrated navigation systems have emerged.A common integrated navigation is GPS+INS,which utilizes GPS to correct the error of INS to achieve high location precision.In this paper,we propose a new integrated navigation system including three sensors.This new integrated system adds a DVL to assist GPS+INS system and use the position information of GPS and the velocity information of DVL to eliminate the error of INS.The GPS+INS+DVL integrated navigation system can solve the problem of GPS+INS system that GPS can’t correct the error of INS during the outage of GPS signal.As a result and has a wider applied range and higher precision.On the other hand,navigation estimation algorithm has also been a hot research focus in recent years.Traditional estimation algorithms include KF algorithm applied in linear model,EKF algorithm applied in nonlinear model,Sage-Husa adaptive KF algorithm with robustness and FKF algorithm applied in information fusion.To better parallel performance of different algorithms,this paper firstly proposes a new algorithm evaluation index---residual variance and demonstrates the reliability of this index through rigorous mathematical derivation based on the orthogonality principle.Then this paper deduces expression of KF’s residual variance under three unsatisfactory conditions:1)inaccurate system model;2)erroneous initial value;3)abrupt change of states.We find that KF has poor estimation performance under the above three cases.Nevertheless,the residual of STF algorithm can satisfy the orthogonality principle no matter in which unsatisfactory cases because of the introduction of fading factor.Finally,this paper establishes a single GPS navigation simulation model to verify the performance of KF and STF.Through the simulation it is easily found that the performance of STF outperforms that of KF in unsatisfactory cases.Based on the new navigation model and the analysis of estimation algorithm,this paper proposes a new GPS+INS+DVL integrated navigation system based on RAFSTKF algorithm to achieve high precision in navigation and positioning.This proposed algorithm includes two parts:local filter and information fusion of main filter.It can take full advantages of observations of all sensors.Both of the two local filters are STF and the stable variable is the error of INS.Observation inputs of two filters are position differences between GPS and INS as well as velocity differences between DVL and INS respectively.The master filter firstly gets the direct fusion based on the least square filter and the direct fusion result is suboptimal estimation.Then the master filter utilizes the weighting adaptive least square filter on account of the suboptimal estimation to get the global optimal estimation.The weighting factor lies on the two norm of the difference between the local filter estimation as well as the one step prediction and the suboptimal estimation.Finally,the estimated INS errors will be fed back to INS to correct errors and realize the precise positioning.Simulations demonstrate that the GPS+INS+DVL system based on RAFSTKF has higher positioning precision and more stable performance.

  • 【网络出版投稿人】 山东大学
  • 【网络出版年期】2019年 01期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络