论文部分内容阅读
本文介绍一种用于具有附加高斯噪声的一般非线性量测模型的分布式参数估计算法。我们证明:当扩展为多传感器情况得出一个线性融合规则时,由Kulhavy′提出的贝叶斯-闭式估计算法与局部的后验密度的形式无关。特别是Kulhavy′算法产生一组表示局部传感器密度的简化的充分统计量(RSS),这是在全局处理机中进行简单地相加和相减而获得最优融合。我们讨论了关于贝叶斯-闭式算法的各种近似值,得到非线性量测模型的实际参数估计器,并将这种近似技术应用于纯方位跟踪问题。把分布式跟踪器的性能与基于在修正的极坐标(MPC)中实现的广义卡尔曼滤波器(EKF)的另一种算法作了比较。已经证明:从通常的EKF意义上讲,贝叶斯-闭式估计器没有发散,因此在单向和双向发送方式中都可以利用贝叶斯-闭式技术。
This paper presents a distributed parameter estimation algorithm for a general non-linear measurement model with additional Gaussian noise. We show that the Bayesian-closed estimation algorithm proposed by Kulhavy ’has nothing to do with the form of the local posterior density when extending to a multisensor case yields a linear fusion rule. In particular, the Kulhavy’s algorithm produces a set of simplified sufficient statistics (RSS) that represent the local sensor densities, which is obtained by simply adding and subtracting in the global processor to obtain the optimal fusion. We discuss various approximations of the Bayesian-closed algorithm and obtain the actual parameter estimators of the nonlinear measurement model. We apply this approximation to the purely azimuth tracking problem. The performance of the distributed tracker is compared to another algorithm based on Generalized Kalman Filter (EKF) implemented in Modified Polar Coordinate (MPC). It has been shown that the Bayesian-closed estimator has no divergence in the usual sense of EKF, so that Bayesian-closed technique can be used in both unidirectional and bidirectional transmission schemes.