| 摘要: |
| 针对同时含有恒定传感器系统偏差和重尾量测噪声的非线性系统目标跟踪问题，提出一种基于交互式多模型框架的增量容积卡尔曼滤波器（Incremental Cubature Kalman Filter Based on Interactive Multiple Mode Framework,IMM-ICKF）。首先，针对带有非高斯重尾量测噪声的目标估计问题，IMM-ICKF将量测噪声建模成高斯混合分布。然后，对相邻两个时刻存在恒定传感器系统偏差的量测进行差分得到增量量测，减弱了恒定传感器系统偏差对滤波的不利影响。同时，根据增量量测噪声的高斯混合分布特性，在交互式多模框架下将增量量测噪声分解成多个并行滤波器进行处理，最后综合多个并行滤波器的递推估计结果进行融合输出。仿真结果表明，在同时含有恒定传感器系统偏差和重尾量测噪声的非线性系统环境下，与基于高斯混合的增量容积卡尔曼滤波算法相比，IMM-ICKF位置跟踪精度提升约10%，运行时间降低约92%。 |
| 关键词: 非线性系统 目标跟踪 增量容积卡尔曼滤波器 交互式多模型 |
| DOI：10.20079/j.issn.1001-893x.240831003 |
|
| 基金项目:国家自然科学基金资助项目(62263007)；中央引导地方科技发展资金项目(桂科ZY22096026)；广西重点研发项目(桂科AB23026147)；认知无线电与信息处理教育部重点实验室基金(CRKL210101)；桂林电子科技大学研究生创新项目(2024YCXB06)；广西高校数据分析与计算重点实验室开放基金；广西应用数学中心（桂林电子科技大学）开放基金(桂科AD23023002) |
|
| Interactive Multiple Model Incremental Cubature Kalman Filtering Algorithm with Heavy-tail Measurement Noise |
| CAI Ruhua琣,CHU Yuanlan琣,WU Sunyong琣,b,YU Runhua琧,ZHANG Xiaoqi琣 |
| (a.School of Mathematics and Computing Science;b.Guangxi Key Laboratory of Cryptography and Information Security;c.School of Information and Communication,Guilin University of Electronic Technology,Guilin541004,China) |
| Abstract: |
| For the problem of nonlinear system target tracking which contains constant sensor system biases and heavy-tail measurement noise,an incremental cubature Kalman filter based on interactive multiple mode framework(IMM-ICKF) is proposed.Firstly,for the target estimation problem with non-Gaussian heavy-tail measurement noise,IMM-ICKF models the measurement noise into Gaussian mixture distribution.Then,the differential measurement of two adjacent moments with constant sensor system biases is used to obtain the incremental measurement,which reduces the adverse effect of constant sensor system biases on filtering.At the same time,according to the Gaussian mixture distribution characteristics of the incremental measurement noise,the incremental measurement noise is decomposed into multiple parallel filters for processing under the interactive multiple mode framework.Finally,the recursive estimation results of the multiple parallel filters are integrated for fusion output.The simulation results show that,in a nonlinear system environment characterized by both constant sensor system biases and heavy-tailed measurement noise,the position tracking accuracy of the IMM-ICKF algorithm is improved by approximately 10% compared with that of Gaussian mixture-based incremental cubature Kalman filter algorithm,while its execution time is reduced by about 92%. |
| Key words: nonlinear system target tracking incremental cubature Kalman filter interactive multiple model |
《电讯技术》编辑部 