摘要
针对粒子滤波算法在重采样环节出现粒子贫乏导致算法精度不高的问题,通常采用在状态估计过程中增加粒子数量,但这种方法会降低算法实时性,提出了基于图形处理单元(GPU)的骨干粒子群算法优化粒子滤波算法。首先利用骨干粒子群算法优化粒子滤波重采样,解决了粒子贫化的缺点。利用骨干粒子群算法中粒子群体之间相互独立运行的特点,在GPU上并行实现骨干粒子群优化的粒子滤波算法,解决粒子滤波算法在重采样过程中因数据关联而无法充分并行计算的问题。最后,将其应用到变桨距系统的故障检测中,提高故障检测的准确度和实时性。实验结果表明:该方法相较于随机重采样的粒子滤波算法误差降低了31.2%,实时性提高了82.7%。
To solve the problem of particle filtering(PF)algorithm occur poor particle quantity in resampling,usually uses increasing number of particles in the process of state estimation,but this method will reduce the real-time performance of the algorithm,a backbone particle swarm optimization(BBPSO)PF algorithm based on GPU is proposed.Firstly,the backbone particle swarm optimization algorithm is used to optimize particle filter resampling.Secondly,using the characteristics of independent operation between particle groups in backbone particle swarm optimization,the particle filtering algorithm of backbone particle swarm optimization is implemented in parallel on GPU,so as to solve the problem that the particle filtering algorithm cannot fully parallel compute due to data association in the process of resamping.Finally,it is applied to the fault detection of variable pitch system to improve the accuracy and real-time of fault detection.Experimental results show that the error of the proposed method is reduced by 31.2%and the real-time performance is improved by 82.7%.
作者
曹洁
胡文东
王进花
余萍
赵伟吉
CAO Jie;HU Wendong;WANG Jinhua;YU Ping;ZHAO Weiji(College of Computer and Communication,Lanzhou University of Technology,Lanzhou 730050,China;College of Electrical Engineering and Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Gansu Manufacturing Information Engineering Research Center,Lanzhou 730050,China)
出处
《传感器与微系统》
CSCD
北大核心
2021年第5期157-160,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61763028)
甘肃省自然科学基金资助项目(1506RJZA105)。
关键词
重采样
并行计算
粒子滤波
骨干粒子群优化算法
实时性
resampling
parallel computing
particle filtering(PF)
bare bones particle swarm optimization algorithm(BBPSO)
real-time
作者简介
曹洁(1966-),女,教授,博士研究生导师,研究领域为智能信息处理,信息融合,智能交通;胡文东(1992-),男,硕士研究生,研究方向为并行与分布式处理,E-mail:2457741164@qq.com;王进花(1978-),女,副教授,硕士研究生导师,研究领域为多源信息融合,故障诊断。
