基于BP神经网络的GPS导航算法

GPS导航解算中常用最小二乘算法。随着高动态用户需求精度的不断提高，且由于线性化忽略高次项，初始值精度低以及差分后剩余或放大误差的存在。导航解精度很难满足高动态用户的需求。为此，本文基于BP神经网络的非线性逼近性能。给出了基于BP神经网络的GPS导航算法。实测数据计算结果表明该算法能够真实地反映载体运动轨迹，其导航解的精度和可靠性有明显的提高。     

神经网络辅助的GPS/INS组合导航自适应滤波算法

首先利用预报残差构造的最优自适应因子设计GPS/INS组合导航自适应滤波器。并针对BP神经网络存在的训练速度慢、容易陷入局部极小等问题,给出网络的改进算法。利用神经网络对自适应滤波器状态方程的预报值进行在线修正,给出神经网络辅助的GPS/INS组合导航自适应滤波算法。最后,利用实测数据进行验证。结果表明,改进的神经网络算法明显提高网络收敛速度;两种自适应滤波算法相对标准组合导航算法都能够可靠地反映载体运动轨迹;神经网络辅助的GPS/INS组合导航自适应滤波算法相对GPS/INS组合导航自适应滤波算法在精度和可靠性方面又有明显提高。     

顾及动力学模型误差影响的GPS/INS组合导航自适应滤波算法

在顾及动力学模型随机误差的情况下,设计了一种GPS/INS自适应滤波算法.针对BP神经网络存在的训练速度慢、易陷入局部极小值等问题,对神经网络学习算法进行了改进.利用神经网络进一步减小系统误差对导航解的影响,给出了顾及动力学模型随机误差和系统误差的GPS/INS自适应滤波算法,并利用实测数据验证了算法的有效性.     

GPS导航中非线性Kalman滤波的方法分析

介绍了非线性Kalman滤波常使用的扩展Kalman滤波、平淡Kalman滤波和基于Bancroft算法的两步滤波法,分析了三种方法的优点和不足,为GPS导航方法的选择提供了参考.     

Elman神经网络在GPS高程拟合中的应用

根据Elman神经网络模型能够逼近任意非线性函数的特点和具有反映系统动态特性的能力,将其运用到GPS高程拟合中,通过实例数据比较Elman神经网络与几何解析法、BP神经网络和广义回归神经网络的拟合效果,结果表明该方法拟合效果最佳。     

神经网络辅助的GPS/INS组合导航自适应UKF算法

GPS/INS组合导航非线性系统最优估计算法中,基于统计信息和假设检验理论的多渐消因子自适应滤波算法的应用前提条件是残差向量为高斯白噪声。本文针对观测异常会影响残差向量的数字特性分布,提出了一种神经网络辅助的多重渐消因子自适应SVD-UKF算法。该算法采用神经网络算法削弱观测异常对残差序列高斯白噪声分布特性的影响,利用奇异值分解抑制UKF中先验协方差矩阵负定性变化,同时构造多重渐消因子对预测状态协方差阵进行调整,使得不同的滤波通道具有不同的调节能力,高效地应用于多变量复杂系统。最后利用车载实测数据进行了验证。结果表明,神经网络算法极大削弱了观测粗差对残差序列高斯白噪声分布特性的影响,拓展了多重渐消因子的应用范围,使其能在观测值含有粗差的条件下自适应调节不同滤波通道,消除滤波状态中的异常,提高组合导航解的精度和可靠性。     

一种基于Bancroft算法的GPS动态抗差自适应滤波

基于抗差自适应滤波的思想，结合非线性Bancroft算法的特点，提出了一种基于Bancroft算法的GPS动态抗差自适应滤波。计算表明，该算法不仅在一定程度上减弱了由于线性化忽略高次项对导航解的影响，而且再次证实抗差自适应滤波在控制扰动异常的有效性和合理性。     

BP神经网络在GPS高程拟合中的应用探讨

神经网络算法一直是国内外研究的热点问题,BP神经网络算法具有更小的模型误差,因此,被广泛应用于GPS高程拟合。本文通过对同一区域GPS高程拟合的应用探究,运用迭代运算对比BP神经网络算法与多项式拟合数据,从而证明BP神经网络在一定条件下具有更高的精度,更加突出了BP神经网络算法的实用性。     

改进BP神经网络算法用于GPS高程转换的应用

介绍了用改进的BP神经网络算法转换GPS高程为正常高,并以青东煤矿首采工作面地表移动观测站的GPS数据为例,与标准的BP算法和多项式曲面拟合方法作比较,改进的BP算法在转换GPS高程中减少了BP神经网络的训练时间,提高了转换效率,且精度良好,可用于GPS高程转换。     

利用小波神经网络的GPS高程转换

介绍小波神经网络的基本结构及学习算法,并应用于GPS大地高转换为正常高。结合实际工程数据,与BP神经网络作比较分析,因小波网络较强的非线性使得它泛化性能更好,收敛速度更快,经实例论证,在同等条件下,小波神经网络方法用于GPS高程转换的精度优于BP神经网络,且其精度可满足常规工程需要,具有一定实用价值。     

Neural Networks in GPS Navigation

Neural networks have been proposed as nonlinear filters in a variety of applications that involve nonlinear processing of input signals; examples include blind signal separation, image registration, and blind deconvolution. The Global Positioning System (GPS) navigation equations are nonlinear (quadratic) in nature, and a direct closed form solution of the GPS navigation equations does not exist. This article presents a new approach to solving the GPS pseudorange equations using three-layer neural networks. A three-layer radial basis function (RBF) neural network is designed, which solves the non-linear GPS pseudorange equations directly as opposed to the linear least squares or extended Kalman filter approaches in traditional GPS receivers. For training the neural network, a carefully selected cost function is minimized using a variation of the classical conjugate gradient algorithm such that training time for the neural network is reasonable. Simulations have been performed at SiRF Technology Inc. that show stable behavior even under bad geometry conditions where the traditional recursive least squares and extended Kalman filter approaches show high sensitivity to measurement errors. Under good geometry conditions the neural network solution shows slightly improved noise performance compared to the expected performance of traditional leas squares solution. Simulations have been performed with additive white Gaussian noise and correlated noise models to evaluate the performance of the trained neural network. ? 2000 John Wiley & Sons, Inc.     

Unscented Kalman filter with nonlinear dynamic process modeling for GPS navigation

This paper preliminarily investigates the application of unscented Kalman filter (UKF) approach with nonlinear dynamic process modeling for Global positioning system (GPS) navigation processing. Many estimation problems, including the GPS navigation, are actually nonlinear. Although it has been common that additional fictitious process noise can be added to the system model, however, the more suitable cure for non convergence caused by unmodeled states is to correct the model. For the nonlinear estimation problem, alternatives for the classical model-based extended Kalman filter (EKF) can be employed. The UKF is a nonlinear distribution approximation method, which uses a finite number of sigma points to propagate the probability of state distribution through the nonlinear dynamics of system. The UKF exhibits superior performance when compared with EKF since the series approximations in the EKF algorithm can lead to poor representations of the nonlinear functions and probability distributions of interest. GPS navigation processing using the proposed approach will be conducted to validate the effectiveness of the proposed strategy. The performance of the UKF with nonlinear dynamic process model will be assessed and compared to those of conventional EKF.     

利用神经网络预测的GPS/SINS组合导航系统算法研究

提出了一种基于神经网络预测的GPS/SINS组合导航系统算法。GPS信号可用时,该算法分别将惯性传感器的输出以及卡尔曼滤波器的输出信息作为神经网络的输入及理想输出信息,并进行在线训练;当GPS信息失锁时,利用已经训练好的神经网络预测各导航参数误差,并校正SINS。地面静态实验与动态跑车实验结果证明了该方法的可行性与有效性。     

Adaptive estimation of multiple fading factors in Kalman filter for navigation applications

Kalman filter is the most frequently used algorithm in navigation applications. A conventional Kalman filter (CKF) assumes that the statistics of the system noise are given. As long as the noise characteristics are correctly known, the filter will produce optimal estimates for system states. However, the system noise characteristics are not always exactly known, leading to degradation in filter performance. Under some extreme conditions, incorrectly specified system noise characteristics may even cause instability and divergence. Many researchers have proposed to introduce a fading factor into the Kalman filtering to keep the filter stable. Accordingly various adaptive Kalman filters are developed to estimate the fading factor. However, the estimation of multiple fading factors is a very complicated, and yet still open problem. A new approach to adaptive estimation of multiple fading factors in the Kalman filter for navigation applications is presented in this paper. The proposed approach is based on the assumption that, under optimal estimation conditions, the residuals of the Kalman filter are Gaussian white noises with a zero mean. The fading factors are computed and then applied to the predicted covariance matrix, along with the statistical evaluation of the filter residuals using a Chi-square test. The approach is tested using both GPS standalone and integrated GPS/INS navigation systems. The results show that the proposed approach can significantly improve the filter performance and has the ability to restrain the filtering divergence even when system noise attributes are inaccurate.     

GPS导航解算中几种非线性Kalman滤波的理论分析与比较

GPS导航解算中常采用离散线性Kalman滤波模型.由于线性化忽略高次项,加之线性化受初始值精度的影响,导致线性化模型精度很难满足高动态用户需求.为此,分别讨论了扩展Kalman滤波和Bancroft算法以及利用观测信息迭代精化观测方程三种算法,并结合算例进行了比较与分析.     

Hybrid derivative-free extended Kalman filter for unknown lever arm estimation in tightly coupled DGPS/INS integration

Differential carrier phase observations from GPS (Global Positioning System) integrated with high-rate sensor measurements, such as those from an inertial navigation system (INS) or an inertial measurement unit (IMU), in a tightly coupled approach can guarantee continuous and precise geo-location information by bridging short outages in GPS and providing a solution even when less than four satellites are visible. However, to be efficient, the integration requires precise knowledge of the lever arm, i.e. the position vector of the GPS antenna relative to the IMU. A previously determined lever arm by direct measurement is not always available in real applications; therefore, an efficient automatic estimation method can be very useful. We propose a new hybrid derivative-free extended Kalman filter for the estimation of the unknown lever arm in tightly coupled GPS/INS integration. The new approach takes advantage of both the linear time propagation of the Kalman filter and the nonlinear measurement propagation of the derivative-free extended Kalman filter. Compared to the unscented Kalman filter, which in recent years is typically used as a superior alternative to the extended Kalman filter for nonlinear estimation, the virtue of the new Kalman filter is equal estimation accuracy at a significantly reduced computational burden. The performance of the new lever arm estimation method is assessed with simulated and real data. Simulations show that the proposed technique can estimate the unknown lever arm correctly provided that maneuvers with attitude changes are performed during initialization. Field test results confirm the effectiveness of the new method.     

SINS／GPS级联组合滤波新算法

针对级联式SINS／GPS导航系统中组合滤波器的测量噪声与系统噪声的相关问题，提出了一种相关程度未知条件下的估计量最优融合算法，并由此得到了一个顾及噪声相关性的的卡尔曼滤波新算法；通过对一套SINS／GPS数据的计算，证明了新算法是成功的。     

神经网络在SINS/GPS组合定位中的应用

地籍测量中,单一系统无法满足定位要求,组合定位技术应运而生. 其中,捷联惯性导航系统(SINS)和GPS组合定位应用最为广泛.在卫星信号受到干扰失效区域,系统进入纯SINS解算,定位误差会逐渐累积,无法满足定位精度要求. 针对此问题,提出一种长短期记忆(LSTM)神经网络辅助的组合定位算法. 根据LSTM神经网络能够有效运用于长距离时间序列的特性,在GPS有效区域,用卡尔曼滤波(KF)算法对SINS/GPS信号进行数据融合得到精确定位信息,同时利用惯性测量单元(IMU)、GPS和SINS输出信息对神经网络进行训练;在GPS失效区域,利用训练好的神经网络预测GPS位置信息,使得系统能继续用卡尔曼滤波器滤波. 最后结合地籍测量特点,设计了仿真实验,证明了该算法在GPS信号失效时可以有效抑制系统误差发散、提高定位精度,在不同运动状态下依然可以满足定位精度要求、鲁棒性强.