基于MCMC方法的虚拟试验贝叶斯校准研究

虚拟试验利用计算模型研究复杂物理过程,并预测其性能.计算模型输入参数中通常包含部分固定但未知参数,可利用计算模型结果和少量有限的试验数据校准未知参数,并研究存在建模不确定性时虚拟试验的预测问题.提出了一种贝叶斯统计方法,采用高斯过程为仿真计算模型以及模型不确定性建模,利用Markov chain蒙特卡罗抽样方法计算校准参数和仿真模型预测后验分布.设计测试算例演示所提出方法的高效性.     

Consecutive tracking for ballistic missile based on bearings-only during boost phase

This paper proposes a modified centralized shifted Rayleigh filter(MCSRF) algorithm for tracking boost phase of ballistic missile(BM) trajectory with a highly nonlinear dynamical model based on bearings-only.This paper contributes three folds.Firstly,the mathematical model of an MCSRF for multiple passive sensors is derived.Then,minimum entropy based onedimensional optimization search to adaptively adjust the probability of the different filters for real time state estimation is deployed.Finally,the unscented transform(UT) is introduced to resolve the asymmetric state estimation problem.Simulation results show that the proposed algorithm can consecutively track the BM precisely during the boost phase.In comparison with the unscented Kalman filter(UKF) algorithm,the proposed algorithm effectively reduces the tracking position and velocity root mean square(RMS) errors,which will make more sense for early precision interception.     

Simulation method for reliability of TT&C mission with high redundancy and small time horizon

The tracking, telemetry and command (TT&C) mission is extremely reliable for its characters of small time horizon and high redundancy. The combined forcing and failure biasing (CFFB) method that is usually used for simulating the unreliability of the highly dependable mission system seems not so efficient for the TT&C mission. The concept about the importance of failure transition is proposed based on the logical relationship between TT&C mission and its involved resources. Then, the importance is used for readjusting the transition rate of the failure transition when using the forcing and failure biasing during the simulation. Examples show that the improved CFFB method can evidently increase the occurrence of the TT&C mission failure event and decrease the sample variance. More redundancy of the TT&C mission leads to the improved CFFB method more efficient.     

基于强跟踪滤波器估计的最优融合方法

在分布式雷达数据处理模式中 ,数据融合是获得较精确的目标轨迹的主要环节。为克服卡尔曼滤波器对初始值敏感、鲁棒性差和对机动目标跟踪性能差的缺陷 ,通过利用各传感器的观测数据 ,采用强跟踪滤波器对目标进行跟踪 ,以改善目标状态估计的精度。对判定源于同一目标的状态估计值 ,给出了一种估计状态线进行性组合的最优融合准则。得出了实际数据的实验结果。     

双基阵纯方位水下被动目标跟踪性能仿真分析

研究了基于纯方位信息的双基阵声纳信息融合的跟踪性能，并进行了详细的仿真实验。仿真结果表明，在平台—目标相遇这种较为常见的态势下应用扩展卡尔曼滤波可以快速、稳定地完成目标运动要素的估计，同时还揭示了平台与目标运动的几何态势对跟踪性能的影响，为工程应用提供了参考。     

时变转移概率IMM-SRCKF机动目标跟踪算法

给出了一种交互多模型(interacting multiple model,IMM)算法中Markov转移概率矩阵在线修正的方法,并将平方根容积卡尔曼滤波器(square-root cubature Kalman filter,SRCKF)引入到IMM算法中,提出一种时变转移概率的机动目标跟踪IMM-SRCKF算法。该算法利用当前量测中包含的模式信息,对IMM算法中的转移概率矩阵进行实时递推估计,避免了常规IMM算法中转移概率先验确定的困难,提高了模型切换速度和跟踪精度;同时,SRCKF以目标状态协方差的平方根进行迭代更新,确保了滤波过程中协方差矩阵的对称性和半正定性,改善了数值精度和稳定性。仿真实验结果表明,该算法对机动目标的跟踪性能优于常规的IMM及IMM-CKF算法。     

多目标数据关联的神经网络解算

从组合优化的角度,建立了密集回波条件下多目标数据关联的数学模型,并提出连续型Hopfield神经网络解算方案,对结果进行了仿真。通过与联合概率数据关联(JPDA)方法的对比仿真,表明了该算法的有效性和快速性。     

有限记忆ISODATA算法在航迹关联中的研究

针对多传感器多目标航迹关联的特点,提出了将基于聚类分析的ISODATA算法应用于航迹关联的解决方法。将表征航迹特征的参量构成待分类的样本空间,利用ISODATA算法对来自不同传感器的航迹进行关联。构建了基于ISODATA算法的有限记忆关联模型,利用有限步的历史航迹状态数据,有效地解决了目标密集环境下的误关联问题。通过仿真研究说明了算法的有效性。     

AN EXPONENTIAL INEQUALITY FOR AUTOREGRESSIVE PROCESSES IN ADAPTIVE TRACKING

A wide range of literature concerning classical asymptotic properties for linear models with adaptive control is available, such as strong laws of large numbers or central limit theorems. Unfortunately, in contrast with the situation without control, it appears to be impossible to find sharp asymptotic or nonasymptotic properties such as large deviation principles or exponential inequalities. Our purpose is to provide a first step towards that direction by proving a very simple exponential inequality for the standard least squares estimator of the unknown parameter of Gaussian autoregressive process in adaptive tracking.     

基于自适应初始搜索点预测的目标跟踪算法

为了得到恰当的初始搜索点以使得目标跟踪算法避开背景干扰并缩短搜索距离,提出了一种自适应初始搜索点预测的算法。该算法通过对坐标变换参数的变化率进行Kalman滤波来更好地预测初始搜索点;更重要的是,该算法有效地在线估计Kalman滤波器中的模型噪声功率,而非先验地对它们的取值做出假设,因而能够在没有任何人工干预的情况下动态地根据不同的目标运动状况和搜索精度进行实时调整。大量实景视频流上的实验结果均证实了该算法显著提高了跟踪稳定性,并且大幅降低了计算量。