计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (24): 222-226.

• 信号处理 • 上一篇    下一篇

AR预测模型的IMM跟踪算法

竹  博,周  游,仵国锋,胡捍英   

  1. 解放军信息工程大学,郑州 450002
  • 出版日期:2014-12-15 发布日期:2014-12-12

AR prediction model based IMM tracking algorithm

ZHU Bo, ZHOU You, WU Guofeng, HU Hanying   

  1. PLA Information Engineering University, Zhengzhou 450002, China
  • Online:2014-12-15 Published:2014-12-12

摘要: 针对LOS/NLOS混合条件下对机动目标的鲁棒跟踪问题,提出一种基于AR预测模型的交互式多模型(Interacting Multiple Model,IMM)跟踪算法(ARIMM)。该算法利用AR预测模型对运动状态建模,针对LOS与NLOS条件下观测噪声的分布不同分别使用无迹卡尔曼滤波器(Unscented Kalman Filter,UKF)和改进的无迹卡尔曼滤波器(Robust Unscented Kalman Filter,RUKF),通过IMM方法估计出移动台的位置,利用该位置更新AR模型的参数,使AR模型与真实运动状态更加匹配,实现精确跟踪。仿真结果表明,在LOS/NLOS混合条件下,与传统的UKF和RUKF算法相比,该算法对机动目标跟踪的鲁棒性更好。

关键词: 机动目标跟踪, 交互式多模型, 自回归(AR)预测模型, 无迹卡尔曼滤波器

Abstract: In view of the problem of robust tracking of maneuvering target under LOS/NLOS condition, an IMM algorithm based on AR prediction model is proposed(ARIMM). AR prediction model is adopted to model the motion state, and UKF and RUKF are utilized separately for the reason that the state LOS and NLOS have different distribution of observation noise, and the IMM filter is used to estimate the position of BS, and the position is used to update the current parameters in AR prediction model and the AR model is made more matched with the true motion state, therefore the algorithm can perform precisely tracking. Simulation result demonstrates that the proposed algorithm performs better robustness under LOS/NLOS condition compared with the traditional UKF and RUKF.

Key words: maneuvering target tracking, Interacting Multiple Model(IMM), Auto Regressive(AR) prediction model, Unscented Kalman Filter(UKF)