基于格拉斯曼流形的状态跟踪方法研究  

State-Tracking Method based on Grass-Mann Manifold

作　　者：赵成 苏岭东 马祥林 ZHAO Cheng;SU Ling-dong;MA Xiang-lin(Hangzhou BONUAP Intelligent Technology Co.,Ltd.,Hangzhou Zhejiang 310012,China;Power Dispatch and the Control Center,State Grid Xuzhou Power Supply Company,Xuzhou Jiangsu 221000,China;Changzhou Zhike Automation Technology Co.,Ltd.,Changzhou Jiangsu 213001,China)

机构地区：[1]杭州邦友安派智能科技有限公司,浙江杭州310012 [2]国网徐州供电公司,江苏徐州221000 [3]常州致科自动化科技有限公司,江苏常州213001

出　　处：《通信技术》2020年第3期606-610,共5页Communications Technology

摘　　要：传统的格拉斯曼流形状态估计是将状态空间模型置于格拉斯曼流行上,在后验跟踪模型的基础上递推估计。但是,该方法将观测模型和噪声都限制在格拉斯曼流形上会产生较大误差。针对上述问题,提出一种基于格拉斯曼流形的粒子滤波算法(Grass-Mann Manifolds-Paticle Filter,GM-PF),利用粒子滤波算法估计流形上的隐马尔科夫过程。仿真实例表明,该方法显著提高了流形上的隐马尔可夫过程的估计精度,且当噪声逐渐增大时,粒子滤波算法表现出良好的抗干扰能力和跟踪性能。The traditional Grass-Mann manifold state estimation is to put the state space model on the Grass-Mann epidemic,and implement recursive estimation based on the posterior tracking model.However,this method restricts the observation model and noise to the Grass-Mann manifold,which will produce larger errors.Aiming at the above problem,a particle filter algorithm(Grass-Mann Manifolds-Particle Filter,GM-PF)based on Grass-Mann manifold is proposed.It uses a particle filter algorithm to estimate the hidden Markov process on the manifold.Simulation results indicate that this method significantly improves the estimation accuracy of the hidden Markov process on the manifold,and when the noise gradually increases,the particle filter algorithm shows good anti-interference ability and tracking performance.

关 键 词：格拉斯曼流形 粒子滤波 隐马尔科夫模型 随机过程 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]