J4 ›› 2014, Vol. 41 ›› Issue (6): 37-44.doi: 10.3969/j.issn.1001-2400.2014.06.007

• 研究论文 • 上一篇    下一篇

量化量测条件下的交互多模型箱粒子滤波

赵雪刚;宋骊平;姬红兵   

  1. (西安电子科技大学 电子工程学院,陕西 西安  710071)
  • 收稿日期:2013-07-25 出版日期:2014-12-20 发布日期:2015-01-19
  • 通讯作者: 赵雪刚
  • 作者简介:赵雪刚(1987-),男,西安电子科技大学硕士研究生,E-mail: xgzhao1987@163.com.
  • 基金资助:

    国家自然科学基金资助项目(61372003);中央高校基本科研业务费专项资金资助项目(JB140221)

Interacting multiple model box particle filter with quantitative measurements

ZHAO Xuegang;SONG Liping;JI Hongbing   

  1.  (School of Electronic Engineering, Xidian Univ., Xi'an  710071, China)
  • Received:2013-07-25 Online:2014-12-20 Published:2015-01-19
  • Contact: ZHAO Xuegang

摘要:

在分布式多传感器网络中,为了节省通信带宽,需要将传感器得到的点量测量化成区间量测,而传统的滤波算法均不能直接处理这种量化量测.箱粒子滤波作为一种“广义粒子滤波”算法,用箱粒子和误差界限模型来取代传统的点粒子和误差统计模型,是新近出现的处理区间量测的有力工具.相比粒子滤波,箱粒子滤波还具有所需粒子数少、算法复杂度低、运行速度快等优点.因此,为了处理量化量测条件下的机动目标跟踪问题,提出了交互多模型箱粒子滤波算法.仿真对比实验表明:在量化量测条件下,交互多模型箱粒子滤波算法和交互多模型粒子滤波算法都能够准确地估计机动目标状态,但交互多模型箱粒子滤波所需粒子数更少、计算效率更高.

关键词: 交互多模型, 粒子滤波, 箱粒子滤波, 量化量测

Abstract:

In the distributed multi-sensor networks, in order to save the communication bandwidth,  to quantize the point observations obtained by sensors into the interval measurements is required. However, the traditional filtering algorithm can not directly deal with the quantitative measurements. The box particle filter (Box-PF) as a "generalized particle filter" algorithm uses the box particles and the bounded error model to replace the traditional point particles and the error statistical model. Therefore, it is a powerful tool for processing interval measurements. Key advantages of the Box-PF against the standard particle filter (PF) are a smaller particle number,  reduced computational complexity and a fast running speed. Therefore, to cope with the maneuvering target tracking with the quantitative measurements, this paper presents an interacting multiple model box particle filter (IMMBPF) algorithm. Simulation results show that under the condition of quantitative measurements IMMBPF and IMMPF are both able to accurately estimate the states of the maneuvering target. The IMMBPF, however, needs fewer particles, and computes more efficiently.

Key words: interacting multiple model, particle filter, box particle filter, quantitative measurements

中图分类号: 

  • TN953