文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 工程科技 > 能源/化工 > Variable Structure Multiple Model Particle Filter for
侵权投诉

Variable Structure Multiple Model Particle Filter for

Variable Structure Multiple Model Particle Filter for

Maneuvering Radar Target Tracking

Min Zhang, Weidong Chen

Department of Electronics Engineering and Information Science, University of Science & Technology of China, Hefei, P.R.China, 230027

Abstract-This paper presents two methods for maneuvering radar target tracking, one is called variable structure multiple model particle filter (VSMMPF),which combines the variable structure multiple model (VSMM) filter wit a Sampling-Importance Resampling (SIR) particle filter (PF); The other one is called modified adaptive gird multiple model (AGMM), which is suitable when model parameter scope is unknown beforehand. PF is used to deal with problems with arbitrary nonlinearities and non-Gaussian noise. Thus the combination of VSMM and PF is able to deal with problems of nonlinearities, non-Gaussian and maneuvering. As one of the VSMM algorithms, Grid adaptation (GA) is unique one which can produce new model-set that are not predefined and model parameter can be any value within continuous model space. However, it only used to trackingmaneuvering target in 2D coordinate turn system. Our modified AGMM could track generic planar maneuvering target. Simulation results indicate that performance of the algorithm VSMMPF outperforms interacting multiple model (IMM) PF and the AGMM PF which combines AGMM with PF runs well especially when model scope is unknown in advance.

hIndex terms: Variable structure, particle filter, adaptive grid, radar target tracking

I.INTRODUCTION

The problem of tracking maneuvering radar target has attracted much attention in recent years. Nonlinearity, non-Gaussianity and maneuvering often exist in target tracking problems. The Kalman filter is the optimal solution to the Bayesian estimation problem for a given linear system with additive Gaussian noise sources of known statistics. An extended Kalman filter (EKF) or an unscented Kalman filter (UKF) is usually implemented for weak nonlinearity. Another group of popular techniques is the class of sequential Monte Carlo methods, also known as the particle filters (PF), which can be applied to problems with arbitrary nonlinearities in dynamics and measurements.

It is generally difficult to use a single model to represent the motion of a maneuvering target. Hence, multiple model (MM) based approaches are often used for maneuvering target tracking. In particular, the IMM algorithm is widely accepted as one of the most cost-effective dynamic MM methods. IMM use a fixed set of models at each time. When these algorithms are applied to solve real-world problems, it is often the case that the use of only a small number of models is not good enough. The use of more models increases the computational burden considerably with the performance not remarkable

improves. In [1], the concept of VSMM estimation is introduced and several variable structure algorithms are proposed. A series of VSMM algorithms are developed by Li and his team [2-5]. Model-set adaptation, which involves determination of multiple candidate model-sets and selection of best model-set from candidate sets, is an important approach for VSMM in recent years. In [1], the MM algorithms are formulated in a graph theoretic framework. Three representative model-set adaptation approaches, Model-group switching (MGS), likely–model set (LMS) and GA are all belong to these schemes.

Compared with MGS and LMS approaches, GA is unique one for its ability to produce new model-set, which are not predefined and model parameter can be any value within continuous mode space, in real time. In [6], for maneuvering aircraft (MA) tracking problem, two VSMM algorithms with different model-set adaptation approaches, (switching grid) SG IMM and adaptive grid (AG) IMM, has been designed. In [7], a new method for tracking a maneuvering target called IMMPF is proposed. The basic idea is to combine the IMM approach with a PF approach. In the derivation of the standard IMM filter, a merging and filtering process are defined. It adopts a regularized particle filter for this filtering step, and performs the merging step on these probability densities, represented by a Gaussian mixture.

In this paper, a modified AGMM approach used to track planar maneuvering target and a new approach called VSMMPF, which combines VSMM with PF, are proposed. In contrast to existed MM algorithms, the proposed approaches can deal with the tracking problems with nonlinear dynamics, nonlinear measurements and maneuvers. Simulation results show that the modified AGMM is better than LMS. VSMMPF outperforms IMMPF when the model parameter scope is unknown beforehand, and the AGMMPF has a well performance when scope is unknown in advance.

II.

SYSTEM SETUP

Given the following model of a stochastic hybrid system

x(k)=Fkj 1x(k 1)+Gkj 1wkj 1

z(k)=Hkjx(k)+vkj

Where (xk) is the system base state vector, z(k) is the noisy measurement vector,wandv are the mode-dependent process and measurement random noise sequences, which assumed to be mutually independent with covariance, respectively, Qkand Rk,. The random parameter j is the

(1)

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email