摘要
针对半球谐振陀螺(HRG)漂移传统分步标定补偿方法存在的补偿精度低与耗时长问题,提出一种基于动量算法优化的反向传播(BP)神经网络HRG漂移补偿方法。根据HRG误差模型分析了分步标定补偿方法的局限性,构建了基于BP神经网络的HRG漂移补偿模型,并引入动量算法,提升BP神经网络训练效率,利用三只自研的HRG进行了实验验证。实验结果表明:所提方法能够有效提升陀螺精度,同时简化标定和补偿流程,提高陀螺漂移补偿工作效率,相比现有分步标定补偿法,陀螺精度提升36.1%,标定补偿效率提升32.1%。
To address the issues of low compensation accuracy and time-consuming processes in traditional stepwise calibration and compensation method for hemispherical resonator gyro(HRG)drift,a momentum algorithm-optimized backpropagation(BP)neural network-based HRG drift compensation method is proposed.By analyzing the limitations of traditional stepwise calibration method based on the HRG error model,a BP neural network-based HRG drift compensation model is constructed.The momentum algorithm is introduced to enhance the training efficiency of the BP neural network.Experimental validation is conducted using three self-developed HRGs.Results demonstrate that the proposed method can effectively improve gyro accuracy and simplify the calibration and compensation process to increase the efficiency of gyro drift compensation work.Compared with existing stepwise calibration method,the gyroscope accuracy is improved by 36.1%,and the calibration-compensation efficiency is increased by 32.1%.
作者
罗巍
魏博深
陈刚
唐明浩
戴劼峰
LUO Wei;WEI Boshen;CHEN Gang;TANG Minghao;DAI Jiefeng(Tianjin Navigation Instrument Research Institute,Tianjin 300131,China)
出处
《中国惯性技术学报》
北大核心
2025年第5期502-509,共8页
Journal of Chinese Inertial Technology
基金
国家重点研发计划项目(2019YFB1108204)。
关键词
半球谐振陀螺
BP神经网络
陀螺漂移补偿
hemispherical resonator gyroscope
BP neural network
gyro drift compensation
作者简介
罗巍(1973-),男,研究员,硕士生导师,从事惯性导航研究。
