摘要
传统的永磁定位方法为Levenberg-Marquardt(LM)优化算法。针对算法存在运算时间长、快速运动下精度差的问题,提出了一种融合注意力机制的Inception网络的定位方法。首先,将一维磁传感器数据转为二维矩阵数据,以便采用卷积神经网络进行处理;其次,利用Inception网络对六维位姿进行预测,为提升定位精度,在网络中融合了轻量级的卷积块注意力机制;最后将所提方法和LM算法的运算时间及快速运动下位置精度作对比分析。研究结果表明:所提算法在静态下预测的位姿误差为(0.86±0.40)mm和(0.81±0.36)°,运算速度比LM算法提升200倍;快速定位下位置误差在1 mm左右,比LM算法下降50%以上。因此所提方法可以实现快速运动下的高精度磁定位。
The traditional permanent magnet positioning method is Levenberg-Marquardt(LM)optimization algorithm.Aiming at the problems of long operation time and poor accuracy in fast motion of the algorithm,an Inception network location method integrating attention mechanism was proposed.Firstly,the one-dimensional magnetic sensor data was converted into two-dimensional matrix data for processing by convolution neural network;secondly,the six-dimensional pose was predicted using the Inception network,in order to improve the positioning accuracy,the lightweight convolution block attention mechanism was integrated in the network;finally,the operation time and position accuracy of this method and LM algorithm were compared and analysed.The research results show that the predicted pose error of the algorithm in static state is the sum,and the operation speed is 200 times faster than LM algorithm;the position error under fast positioning is about 1 mm,which is more than 50%lower than LM algorithm.Therefore,this proposed method can achieve high-precision magnetic positioning in fast motion.
作者
王姝姝
戴厚德
林志榕
黄巧园
Wang Shushu;Dai Houde;Lin Zhirong;Huang Qiaoyuan(School of Advanced Manufacturing,Fuzhou University,Quanzhou Fujian 362251,China;Quanzhou Institute of Equipment Manufacturing,Haixi Institutes,Chinese Academy of Sciences,Quanzhou Fujian 362216,China;School of Mechanical and Automotive Engineering,Xiamen University of Technology,Xiamen Fujian 361024,China)
出处
《电气自动化》
2023年第4期105-107,111,共4页
Electrical Automation
基金
国家自然科学基金(61973293)
福建省科技计划项目(2021Y0048)。
