水下目标跟踪粒子滤波算法性能分析  

Research on Tracking Performance of Particle Filter for Tracking Underwater Targets

作　　者：张程振 丁元明[1,2] 杨阳 ZHANG Cheng-zhen;DING Yuan-ming;YANG Yang(College of Information Engineering,Dalian University,Dalian 116622,China;Communication and Network Key Laboratory,Dalian University,Dalian 116622,China)

机构地区：[1]大连大学信息工程学院,辽宁大连116622 [2]大连大学通信与网络重点实验室,辽宁大连116622

出　　处：《火力与指挥控制》2022年第2期18-24,共7页Fire Control & Command Control

基　　金：国家自然科学基金资助项目(61540024)。

摘　　要：粒子滤波(PF)、扩展粒子滤波(EPF)和无迹粒子滤波(UPF)在非线性状态估计方面表现出更好的性能。分析了PF、EPF、UPF 3种算法,UPF的估计状态依赖于测量值,对历史模型信息的敏感度较低,并避免了雅可比矩阵的计算,比EPF更容易进行模型设计;EPF在计算上虽然高效,但跟踪效果不稳定。以AUV水下三维目标跟踪为背景,在高斯噪声(GN)和散粒噪声(SN)环境下,仿真比较分析了PF、EPF、UPF 3种算法的性能。仿真实验结果表明,在高斯测量噪声中,PF和UPF获得了精确的估计,EPF误差较大;在散粒测量噪声中,EPF估计值更加不稳定;当测量噪声不变时,增大目标的状态噪声,EPF的跟踪性能优于PF和UPF。Particle filter(PF),extended particle filter(EPF)and traceless particle filter(UPF)show better performance in nonlinear state estimation.First,three algorithms,PF,EPF and UPF,are theoretically analyzed.The estimated state of UPF depends on the measured value,and it is less sensitive to the information of historical model.Moreover,it avoids the calculation of Jacobian matrix,making it easier to design the model than EPF.Although EPF is computationally efficient,the tracking effect is not stable.With AUV underwater 3D target tracking as the background,PF,EPF and UPF algorithms are compared and analyzed by simulation under the environment of Gaussian noise(GN)and granular noise(SN).The simulation results show that PF and UPF can get accurate estimation in gaussian measurement noise,and the error of EPF is larger.The EPF estimation value is more unstable in the granualarly measured noise.When the measurement noise remains unchanged,the tracking performance of EPF is better than PF and UPF when the state noise of the target is increased.

关 键 词：AUV 目标跟踪 粒子滤波 扩展粒子滤波 无迹粒子滤波 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程] TP391.9[自动化与计算机技术—控制科学与工程]