| 引用本文: | 申屠晗,薛安克,骆吉安.多步历史估计信息反馈多模型融合方法[J].控制理论与应用,2015,32(1):11~17.[点击复制] |
| SHEN Tu-han,XUE An-ke,LUO Ji-an.Feedback multiple-stage historical estimating information multiple-model fusion method[J].Control Theory and Technology,2015,32(1):11~17.[点击复制] |
|
| 多步历史估计信息反馈多模型融合方法 |
| Feedback multiple-stage historical estimating information multiple-model fusion method |
| 摘要点击 2819 全文点击 1234 投稿时间：2014-01-09 修订日期：2014-09-25 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2014.40019 |
| 2015,32(1):11-17 |
| 中文关键词 反馈融合 多模型融合 历史估计信息 粒子滤波 |
| 英文关键词 feedback fusion multiple-model fusion historical estimating information particle filter |
| 基金项目 国家自然科学基金重大仪器专项项目(61427808), 国家自然科学基金项目(61174024), 浙江省信号处理重点实验室 开放基金项目(ZJKL--4--SP--OP2014--01)资助. |
|
| 中文摘要 |
| 针对强机动和大观测误差下的目标跟踪问题, 传统低阶多模型融合方法存在估计 精度较低、鲁棒性较差的缺点; 高阶多模型融合方法面临计算量增大和保证实时性之间 的矛盾. 为此本文针对一类多步稳健机动目标跟踪问题提出一种基于历史估计信息反馈的多模型融合框架, 首先累积和反馈历史估计信息, 然后结合当前量测计算多阶模型序列 的似然函数, 最后得到贝叶斯后验融合结果. 同时结合粒子滤波构建了易于工程实现的粒子滤波历史反馈多模型融合算法(PF--HFMM). 仿真表明, 与传统粒子滤波多模型算法 相比, 本法显著提高了估计精度和鲁棒性. |
| 英文摘要 |
| When facing the target tracking problem with high maneuver as well as large observation error, traditional low order multiple-model fusion approach exposes the defects of degrading in estimating accuracy and robustness; high order multiple-model fusion approach confronts the dilemma between the increasing calculating assumption and the insurance of good real-time performance. To this end, we propose a multiple-model fusion scheme based on feeding back the historical estimating results to the problem of tracking a class of multiple-step robust maneuvering targets. First, we accumulate and feed back the historical estimating information, then, compute the likelihoods of the multiple-step model sequences through combining the feedback information and the current observation, at last, obtain the Bayesian posterior fusion results. At the meanwhile, a particle filter historical feedback multiple-model (PF–HFMM) is constructed for the real application. The simulations show that, the proposed algorithm provides better results in fusion accuracy and robustness comparing to the traditional particle filter multiple model algorithm. |
|
|
|
|
|