基于长短期记忆神经网络局部认知结构的最优非线性卡尔曼滤波

兰州理工大学学报 ›› 2025, Vol. 51 ›› Issue (4): 88-94.

• 自动化技术与计算机技术 • 上一篇下一篇

基于长短期记忆神经网络局部认知结构的最优非线性卡尔曼滤波

陈辉^*1, 胡荣海¹, 席磊²

1.兰州理工大学自动化与电气工程学院, 甘肃兰州 730050;
2.甘肃省科学院自动化研究所, 甘肃兰州 730000

收稿日期:2022-08-07 出版日期:2025-08-28 发布日期:2025-09-05
通讯作者: 陈辉(1978-),男,山西闻喜人,博士,教授,博导.Email:huich78@hotmail.com
基金资助:
国家自然科学基金(62163023,61873116,62366031,62363023),甘肃省基础研究创新群体项目(25JRRA058),中央引导地方科技发展资金项目(25ZYJA040),甘肃省军民融合发展专项资金,甘肃省重点人才项目(2024RCXM86)

Optimal nonlinear Kalman filter based on local cognitive structure of long-term and short-term memory neural network

CHEN Hui¹, HU Rong-hai¹, XI Lei²

1. School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou 730050, China;
2. Institute of Automation, Gansu Academy of Sciences, Lanzhou 730000, China

Received:2022-08-07 Online:2025-08-28 Published:2025-09-05

摘要/Abstract

摘要： 为了解决复杂非线性目标跟踪中状态空间模型参数不准确、滤波性能变差的问题,提出一种基于长短期记忆神经网络局部认知结构的最优非线性卡尔曼滤波.首先,在贝叶斯滤波框架下用长短期记忆神经网络认知卡尔曼增益这一组件,以数据驱动的方式逼近最优卡尔曼增益,不需要完全了解底层模型参数,可以在具有部分信息非线性状态下执行扩展卡尔曼滤波.其次,使用一种无监督离线训练算法,不需要提供真实数据,而是通过滤波器的可解释性,根据预测下一时刻观测的内部特征进行无监督损失函数计算.通过仿真验证表明,在非线性模型中的参数不准确时,提出的滤波器性能优于传统滤波器.

关键词: 扩展卡尔曼滤波, 长短期记忆神经网络, 深度学习, 数据驱动

Abstract: To solve the problems of inaccurate state space model parameters and poor filtering performance in complex nonlinear target tracking, an optimal nonlinear Kalman filter based on local cognitive structure using a long-term and short-term memory neural network is proposed. Firstly, in the framework of Bayesian filtering, long-term and short-term memory neural network is adopted to recognize Kalman gain, which approximates optimal Kalman gain in a data-driven way. It does not need to fully understand the underlying model parameters, and the extended Kalman filter can be performed in the nonlinear state with partial information. Next, an unsupervised off-line training algorithm is used, which does not need to provide real data, but calculates the unsupervised loss function according to internal characteristics of the observation at the next time through filter interpretability. Simulation results show that the proposed filter can significantly improve the performance in comparison with traditional filter when the parameters in nonlinear model are inaccurate.

Key words: extended Kalman filter, long-term and short-term memory neural network, deep learning, data driven

中图分类号:

TP273

陈辉, 胡荣海, 席磊. 基于长短期记忆神经网络局部认知结构的最优非线性卡尔曼滤波[J]. 兰州理工大学学报, 2025, 51(4): 88-94.

CHEN Hui, HU Rong-hai, XI Lei. Optimal nonlinear Kalman filter based on local cognitive structure of long-term and short-term memory neural network[J]. Journal of Lanzhou University of Technology, 2025, 51(4): 88-94.

参考文献

[1] LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015,521(7553):436-444.
[2] BENGIO Y.Learning deep architectures for AI[M].Boston:Now Publishers Inc,2009:1-127.
[3] HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[4] YIN W,KANN K,YU M,et al.Comparative study of CNN and RNN for natural language processing[J/OL].[2022-06-10].https://arxiv.org/abs/1702.01923.
[5] LI S,LI W,COOK C,et al.Independently recurrent neural network(IndRNN):building a longer and deeper RNN[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Salt Lake City:[s.n.],2018:5457-5466.
[6] FU R,ZHANG Z,LI L.Using LSTM and GRU neural network methods for traffic flow prediction[C]//2016 31st Youth Academic Annual Conference of Chinese Association of Automation (YAC).Orlando:IEEE,2016:324-328.
[7] CHUNG J,GULCEHRE C,CHO K H,et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[J/OL].[2022-06-10].https://arxiv.org/abs/1412.3555.
[8] SHEWALKAR A.Performance evaluation of deep neural networks applied to speech recognition:RNN,LSTM and GRU[J].Journal of Artificial Intelligence and Soft Computing Research,2019,9(4):235-245.
[9] VASWANI A,SHAZEER N,PARMAR N,et al.Attention is all you need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.Long Beach:[s.n.],2017:6000-6010.
[10] KNUDSEN E I.Fundamental components of attention[J].Annual Review of Neuroscience,2007,30(1):57-78.
[11] ZAHEER M,AHMED A,SMOLA A J.Latent LSTM allocation:joint clustering and non-linear dynamic modeling of sequence data[C]//34th International Conference on Machine Learning.NSW,Australia:[s.n.],2017:3967-3976.
[12] YAQI C U I,YOU H E,TIANTIAN T,et al.A new target tracking filter based on deep learning[J].Chinese Journal of Aeronautics,2022,35(5):11-24.
[13] BECKER P,PANDYA H,GEBHARDT G,et al.Recurrent Kalman networks:factorized inference in high-dimensional deep feature spaces[C]//36th International Conference on Machine Learning.Long Beach:[s.n.],2019:544-552.
[14] AL BITAR N,GAVRILOV A I.Neural networks aided unscented Kalman filter for integrated INS/GNSS systems[C]//2020 27th Saint Petersburg International Conference on Integrated Navigation Systems (ICINS).Petersburg:IEEE,2020:1-4.
[15] JONDHALE S R,DESHPANDE R S.Kalman filtering framework-based real time target tracking in wireless sensor networks using generalized regression neural networks[J].IEEE Sensors Journal,2018,19(1):224-233.
[16] SATORRAS V G,AKATA Z,WELLING M.Combining generative and discriminative models for hybrid inference[C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems.Vancouver:[s.n.],2019:13820-13830.
[17] KRISHNAN R G,SHALIT U,SONTAG D.Deep Kalman filters[J/OL].[2022-06-10].https://arxiv.org/abs/1511.05121.
[18] WANG R,LIU M S,ZHOU Y,et al.A deep belief networks adaptive Kalman filtering algorithm[C]//2016 7th IEEE International Conference on Software Engineering and Service Science (ICSESS).Beijing:IEEE,2016:178-181.
[19] JUNG S,SCHLANGEN I,CHARLISH A.Sequential Monte Carlo filtering with long short-term memory prediction[C]//2019 22th International Conference on Information Fusion (FUSION).Ottawa:IEEE,2019:1-7.
[20] COSKUN H,ACHILLES F,DIPIETRO R,et al.Long short-term memory Kalman filters:recurrent neural estimators for pose regularization[C]//Proceedings of the IEEE International Conference on Computer Vision.Venice:[s.n.],2017:5524-5532.
[21] REVACH G,SHLEZINGER N,VAN SLOUN R J G,et al.Kalman net:data-driven Kalman filtering[C]//2021 IEEE International Conference on Acoustics,Speech and Signal Processing (ICASSP).Toronto:IEEE,2021:3905-3909.
[22] REVACH G,SHLEZINGER N,LOCHER T,et al.Unsupervised learned Kalman filtering[J/OL].[2022-06-10].https://arxiv.org/abs/2110.09005.
[23] REVACH G,SHLEZINGER N,NI X,et al.Kalman net:neural network aided Kalman filtering for partially known dynamics[J].IEEE Transactions on Signal Processing,2022,70(1):1532-1547.

基于长短期记忆神经网络局部认知结构的最优非线性卡尔曼滤波

Optimal nonlinear Kalman filter based on local cognitive structure of long-term and short-term memory neural network

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 6

编辑推荐

Metrics

本文评价

[1]	陈辉, 牛丽丽, 付辉, 张天佑, 席磊. 基于多尺度残差和注意力机制的图像去雾算法[J]. 兰州理工大学学报, 2024, 50(2): 69-76.
[2]	杜先君, 邱小彧. 基于改进LSTM神经网络的化工过程故障诊断[J]. 兰州理工大学学报, 2023, 49(6): 72-79.
[3]	王文婷, 刘姝君, 张耀聪, 杜小泽, 许潼. 基于Seq2Seq深度学习模型的焦炉煤气发生量预测方法研究[J]. 兰州理工大学学报, 2023, 49(5): 50-58.
[4]	郭润兰, 史方青, 范雅琼, 何智. 基于卷积神经网络的壁面清洗机器人障碍物检测识别算法[J]. 兰州理工大学学报, 2022, 48(4): 83-89.
[5]	赵宏, 傅兆阳, 王乐. 基于特征融合的中文文本情感分析方法[J]. 兰州理工大学学报, 2022, 48(3): 94-102.
[6]	刘金霞. 基于深度学习的无线电信号分类[J]. 兰州理工大学学报, 2021, 47(4): 106-110.