多传感器快速粒子滤波器融合的方法及其在导航中的应用

提出在高斯噪声条件下,将多传感器融合中多维测量的全局粒子滤波器的粒子权重表示成多个一维测量的子粒子滤波器的粒子权重的乘积的形式,从而避免了通常计算全局粒子权重需要的大计算量,达到提高测量融合速度的目的.本文还分析了现有方法和提出方法的计算复杂度,并将提出方法应用于导航.分析和仿真的结果均表明,所提出的方法明显提高了计算速度,且随着粒子数目的增加,优势愈加明显.     

高斯渐进贝叶斯滤波器

渐进贝叶斯方法将先验分布到后验分布的演化描述为一阶动态系统,通过在伪时间上连续地引入观测信息实现后验状态估计.该方法的一般形式解,即动态系统的时间导数,是难以得到的.本文提出一种高斯型渐进贝叶斯滤波器.首先在线性高斯条件下推导了时间导数的解析解;然后证明了在该条件下,由该解析解确定的一阶动态系统与常量状态估计的Kalman-Bucy滤波器是一致的,且由此导出的高斯渐进贝叶斯滤波器与卡尔曼滤波器是一致的.最后利用一阶Taylor展开推导了滤波器在非线性高斯条件下的近似解表达式,并采用Monte Carlo方法给出了具体实现方法.通过若干仿真算例表明,新滤波器具有较高的精度,且在一定精度条件下的时间复杂度低于一般粒子滤波器.     

机动目标跟踪双滤波器模型及自适应算法

现代机动目标跟踪的困难来自跟踪的快速生成精度在一定计算负荷约束下的协调难以令人满意,考虑依次处理快速性与精度的方案。采用滑动均值均匀分布描述目标的随机机动特性,分别采用宽带的均值预估滤波器和窄带的跟踪滤波器串联,实现机动速度大变动或突变的精确,快速跟踪,双滤波器的计算量适中,易于工程实现,对各种运动形式进行计算机模拟表明,这类算法对高度机动或弱机动或无机动均可给出较好的目标位置,速度及加速度估值。     

利用粒子流滤波的单通道BPSK信号盲分离算法

由于粒子滤波本身的"粒子贫化"问题,导致基于传统粒子滤波的单通道信号盲分离算法分离性能恶化以及计算量较大,本文提出了一种基于粒子流滤波的单通道BPSK信号盲分离新算法。根据由两路BPSK信号混合的单通道信号,构造了测量方程和状态方程。然后,通过将状态空间中服从先验分布的粒子移动到其对应的后验分布上,实现了粒子更新,其不同于粒子滤波采用重采样来更新粒子,避免了"粒子贫化"现象发生。最后,采用一种基于弱解形式的粒子流滤波器实现BPSK信号的盲分离。计算机仿真结果表明,与粒子滤波算法相比本文算法具有更低的误码率和计算复杂度。     

基于粒子群优化算法的滤波器优化

给出一种基于粒子群优化算法（PSO）的模拟滤波器优化设计方法。传统的模拟滤波器的精度与效率均较差,引入PSO算法可对滤波器参数进行寻优。将滤波器的设计问题转化为滤波器参数的优化问题,然后利用粒子群优化算法对整个参数空间进行高效搜索以获得最优解;通过变异、重新随机化及采用自适应的惯性权重,提高了算法的搜索效率及收敛性。实例计算表明了算法在该类问题中的有效性和可行性。     

数字滤波器设计的文化量子算法

有限脉冲响应(FIR)和无限脉冲响应(IIR)数字滤波器的设计实质可看作是多参数优化问题。为实现高效的数字滤波器,首先将滤波器的设计转化为滤波器参数的约束优化问题,然后提出文化量子(CQ)算法在参数空间进行并行搜索以获得滤波器设计的最优参数值。提出的文化量子算法结合文化原理,在量子种群空间更新中使用了量子旋转门的知识进化机制,是一种可用于实数解优化的快速多维搜索算法。计算机仿真实验表明在对FIR和IIR数字滤波器设计时,文化量子算法的收敛速度和性能都优于粒子群,量子粒子群以及自适应量子粒子群优化等算法,证明了该方法的有效性和优越性。     

一种带多步随机延迟量测高斯滤波器的一般框架解

提出了一种适用于线性和非线性系统的带多步随机延迟量测高斯滤波器的一般框架解. 为了完成状态的递归更新估计, 噪声向量和先前时刻状态向量被扩展到当前时刻状态向量中. 然后基于贝叶斯方法推导了扩展后状态向量的一般框架解. 对于非线性系统, 通过利用不同的数值计算方法计算贝叶斯解中的高斯加权积分可以推导获得不同的高斯近似滤波器. 最后本文利用三阶球径容积准则来实施提出的方法, 并通过量测被随机延迟多步的目标跟踪模型对所提出的方法进行了仿真, 仿真结果验证了提出方法的有效性和优点.     

广义联邦滤波器的全局最优性

基于矩阵理论和信息分配原理导出集中卡尔曼滤波、分散化滤波和联邦滤波之间的解析关系, 证明联邦滤波只有当其主滤波器和局部滤波器的维数都相同时, 其全局滤波才是最优的, 并用信号流图直观清晰地说明联邦滤波较分散化滤波结构更简单, 计算量小. 当联邦滤波的主滤波器和局部滤波器的维数不相同时, 只能得到次优解. 文中提出一种广义联邦滤波器的结构, 按信息分配原理重置其一步预测状态误差信息阵和一步预测状态, 获得全局滤波次优解, 并进一步利用全局滤波次优解作为观测量, 反馈修正其一步预测状态得到全局滤波最优解. 文中对最优反馈增益矩阵进行了数学推导, 从理论上证明其滤波结果同集中卡尔曼滤波是等价的, 并通过一个双SINS/GPS组合导航系统的仿真实验结果验证了算法的有效性.     

分布式并行粒子滤波算法结构分析与研究

粒子滤波器是解决非线性/非高斯系统状态估计的有效技术,广泛应用于目标跟踪、无线通信.自动控制等领域.但因其计算复杂、计算量庞大等缺陷,无法满足实时系统的应用需求.针对粒子滤波器计算量大、实时性差的问题,提出了一种基于MPI的分布式并行粒子滤波算法,给出了Master-Slave并行模式下任务分配、数据划分与负载平衡策略.实验结果表明,若忽略通信代价,加速比基本呈线性增长.     

基于粒子滤波的AUV组合导航方法

讨论了粒子滤波器和RB(Rao-Blackwellised)粒子滤波器两种滤波方法在组合导航中的应用,给出了组合导航算法用于自治水下航行器(AUV)的具体数学模型,并且与拓展卡尔曼滤波器的导航结果进行比较.利用AUV湖上试验验证了3种算法的导航性能,试验结果表明RBPF组合导航算法能够获得最好的导航精度;然而通过对算法进行分析,发现其计算复杂度高于其余两种滤波算法.     

基于粒子滤波的检测前跟踪改进算法

针对低信噪比环境下微弱目标的实时检测与跟踪,提出一种基于粒子滤波的检测前跟踪改进算法.该算法在粒子滤波的基础上融合不敏卡尔曼滤波(uKF1)算法,融合后的新算法在利用重要性密度函数产生粒子时充分考虑当前时刻的量测,从而引导粒子向高似然区域移动,使得粒子的分布更接近状态的后验概率分布.仿真实验表明,改进算法的检测与跟踪性能优于标准的粒子滤波算法.     

Eccentric electrophoretic motion of a sphere in circular cylindrical microchannels

The eccentric electrophoretic motion of a spherical particle in an aqueous electrolyte solution in circular cylindrical microchannels is studied in this paper. The objective is to investigate the influences of separation distance and channel size on particle motion. A theoretical model is developed to describe the electric field, the flow field and the particle motion. A finite element based direct numerical simulation method is employed to solve the model. Numerical results show that, when the particle is eccentrically positioned in the channel, the electric field and the flow field are not symmetric, and the strongest electric field and the highest flow velocity occur in the small gap region. It is shown that the rotational velocity of the particle increases with the decrease of the separation distance. With the decrease of the separation distance, the translational velocity increases in a smaller channel; while it decreases first and then increases in a relatively large channel. When a particle moves eccentrically at a smaller separation distance from the channel wall, both the translational velocity and the rotational velocity increase with the decrease of the channel size.     

Finite element computation of turbulent flows with the discontinuity-capturing directional dissipation (DCDD)

The streamline-upwind/Petrov-Galerkin (SUPG) and pressure-stabilizing/Petrov-Galerkin (PSPG) methods are among the most popular stabilized formulations in finite element computation of flow problems. The discontinuity-capturing directional dissipation (DCDD) was first introduced as a complement to the SUPG and PSPG stabilizations for the computation of incompressible flows in the presence of sharp solution gradients. The DCDD stabilization takes effect where there is a sharp gradient in the velocity field and introduces dissipation in the direction of that gradient. The length scale used in defining the DCDD stabilization is based on the solution gradient. Here we describe how the DCDD stabilization, in combination with the SUPG and PSPG stabilizations, can be applied to computation of turbulent flows. We examine the similarity between the DCDD stabilization and a purely dissipative energy cascade model. To evaluate the performance of the DCDD stabilization, we compute as test problem a plane channel flow at friction Reynolds number Re_τ = 180.     

A Stochastic Filtering Technique for Fluid Flow Velocity Fields Tracking

In this paper, we present a method for the temporal tracking of fluid flow velocity fields. The technique we propose is formalized within a sequential Bayesian filtering framework. The filtering model combines an Itô diffusion process coming from a stochastic formulation of the vorticity-velocity form of the Navier-Stokes equation and discrete measurements extracted from the image sequence. In order to handle a state space of reasonable dimension, the motion field is represented as a combination of adapted basis functions, derived from a discretization of the vorticity map of the fluid flow velocity field. The resulting nonlinear filtering problem is solved with the particle filter algorithm in continuous time. An adaptive dimensional reduction method is applied to the filtering technique, relying on dynamical systems theory. The efficiency of the tracking method is demonstrated on synthetic and real-world sequences.     

基于容积粒子滤波的时延差定位估计算法

在基于粒子滤波的时延差定位估计方法中, 重要密度函数的选取将直接影响估计的性能, 为此, 提出了基 于容积粒子滤波的时延差估计(BCPF-TDE) 算法. 该算法利用最新的数据检测信息, 通过容积卡尔曼滤波(CKF) 获 取粒子滤波的重要性密度函数. 仿真实验表明, 在粒子数目相同的情况下, 基于容积粒子滤波的时延差估计(BCPF- TDE) 方法与基于扩展粒子滤波的时延差估计(BEPF-TDE) 方法相比, 定位估计误差只有后者的50% 左右, 而运行时 间相当.

     

Numerical investigation of erosion threshold velocity in a pipe with sudden contraction

This paper deals with erosion prediction in a pipe with sudden contraction for the special case of two-phase (liquid and solid) turbulent flow with low particle concentration. The pipe axis was considered vertical and the flow was either in direction of gravity (downflow) or against it (upflow). The mathematical models for the calculations of the fluid velocity field and the motion of the solid particles have been established and an erosion model was used to predict the erosion rate. The fluid velocity (continuous phase) model was based on the time-averaged governing equations of 3-D turbulent flow and the particle-tracking model (discrete phase) was based on the solution of the governing equation of each particle motion taking into consideration the effect of particle rebound behavior. The effects of flow velocity and particle size were investigated for one contraction geometry considering water flow in a steel pipe. The results showed the strong dependence of erosion on both particle size and flow velocity but with little dependence on the direction of flow. The effect of flow direction was found to be significant only for large particle size and moderate flow velocity. The erosion critical area was found to be the inner surface of the tube sheet (connecting the two pipes) in the region close to the small pipe. The results also indicated the presence of a threshold velocity below which erosion is insignificant for all particle sizes.     

Satellite derived estimates of the normal and tangential components of near-surface flow

Abstract

The determination of both the normal and tangential components of the total velocity is important in the study of cross- and along-isopycnal transport processes in the ocean. A pattern-matching method is used to determine objectively the total velocity. Sensitivity of this method to pattern and search tile sizes and to correlation threshold also is examined. Three methods for estimating the cross-isopycnal or normal component of the total flow are compared and discussed: Marr-Ullman, optical flow and minimum norm. It is also shown that optical flow and minimum norm are equivalent when the parameter a in the optical flow formulation is set to zero. The direct computation of the tangential component is not possible because it lies in the null-space of the solution set of the basic constraint equations used in velocity estimation methods which are based on the rate of change of image brightness (or temperature). A new method for indirectly estimating the tangential component of the total flow based on vector subtracting of the total flow and the normal component of flow is introduced. Several sequences of satellite images are analysed and the resulting total flow, normal component of flow, and the tangential component determined using this new method are consistent with motion inferred from edge maps. Recommendations are then made for the best normal component of flow to use in the determination of the tangential component.     

Experiment and simulation of mixed flows in a trapezoidal microchannel

This paper presents experimental and numerical results of mixed electroosmotic and pressure driven flows in a trapezoidal shaped microchannel. A micro particle image velocimetry (μPIV) technique is utilized to acquire velocity profiles across the microchannel for pressure, electroosmotic and mixed electroosmotic-pressure driven flows. In mixed flow studies, both favorable and adverse pressure gradient cases are considered. Flow results obtained from the μPIV technique are compared with 3D numerical predictions, and an excellent agreement is obtained between them. In the numerical technique, the electric double layer is not resolved to avoid expensive computation, rather a slip velocity is assigned at the channel surface based on the electric field and electroosmotic mobility. This study shows that a trapezoidal microchannel provides a tapered-cosine velocity profile if there is any pressure gradient in the flow direction. This result is significantly different from that observed in rectangular microchannels. Our experimental results verify that velocity distribution in mixed flow can be decomposed into pressure and electroosmotic driven components.     

一种基于GHF 的高斯粒子滤波算法

高斯粒子滤波算法重要性权值方差不会随迭代次数的增加而增加, 能够较好地解决粒子退化问题, 但其重要性密度函数没有考虑最新的量测信息, 导致有效粒子数减少, 算法滤波性能下降. 针对该问题, 提出一种基于Gaussian-Hermite 滤波(GHF) 的高斯粒子滤波算法, 采用GHF构造高斯粒子滤波的重要性密度函数, 考虑最新的量测信息, 增加有效粒子数, 提高算法的滤波精度. 仿真结果表明, 所提出算法的滤波精度明显优于高斯粒子滤波算法.

     

目标跟踪状态估计排序粒子群算法研究

基于视觉的无人机地面目标跟踪状态估计为非线性滤波问题,针对使用一般粒子滤波算法存在粒子退化和计算量大的缺陷问题,提出了一种基于排序的粒子滤波算法,对粒子依误差大小进行排序并计算粒子权重。仿真试验表明,该方法减轻了粒子贫化的影响,提高了状态估计精度。