摘要
为了实现机器人运动学参数标定,提出一种用惯性测量单元(IMU)实时获取其末端姿态信息的方法。然而,IMU在进行机器人动态姿态测量时,存在加速度计信号中有害加速度(除重力加速度之外的其他加速度)叠加,噪声统计特性参数不易获取,陀螺仪信号随时间发生漂移等影响测量精度的问题。针对这些问题,设计了一种自适应拓展卡尔曼滤波(EKF)姿态测量改进算法。基于EKF模型,首先构建第一级量测噪声方差阵,设定权重因子,降低有害加速度对测量结果的影响;其次在Sage-Husa自适应滤波算法中引入了渐消记忆因子的思想,实时跟踪采样数据的量测噪声,构建第二级量测噪声方差阵;最后采用姿态更新的四元数算法进行数据融合,修正陀螺仪信号漂移产生的误差。实验结果表明,相比Sage-Husa自适应滤波算法,该算法峰高时俯仰角和横滚角的平均绝对误差分别降低了50%和36.43%,峰谷时俯仰角和横滚角的平均绝对误差分别降低了14.28%和19.44%,能有效提高姿态测量精度。
In order to realize the robot kinematic parameter calibration, a method is proposed to acquire its real time terminal attitude information using inertial measurement unit(IMU). However, during the robot dynamic attitude measurement with IMU, there exist some problems that affect the accuracy of measurement, such as the superposition of harmful accelerations(other acceleration information except gravity acceleration signal) in the accelerometer signal, the difficulty in obtaining the statistical characteristic parameters of noise and the drift of gyroscope signal with time. Aiming at these problems, an improved algorithm of adaptive extended kalman filter(EKF) for attitude measurement is designed. Based on the EKF model, this algorithm firstly constructs the first level measurement noise variance matrix, sets up weighted factors and reduces the influence of harmful acceleration signals on measurement result. Secondly, the idea of fading memory factor is introduced in Sage-Husa adaptive filter algorithm to track the measurement noise of sampling data in real time and construct the second level measurement noise variance matrix. Finally, the quaternion algorithm of attitude updating is adopted to conduct data fusion and correct the error caused by gyroscope signal drift. The experiment results show that compared with the Sage-Husa adaptive filter algorithm, the improved self-adaptive EKF algorithm can effectively improve the attitude measurement accuracy;the mean absolute errors of pitch and roll at peak height are reduced by 50% and 36.43%, respectively, and the mean absolute errors of pitch and roll at peak valley are reduced by 14.28% and 19.44%, respectively.
作者
班朝
任国营
王斌锐
陈相君
Ban Zhao;Ren Guoying;Wang Binrui;Chen Xiangjun(College of Mechanical and Electrical Engineering,China Jiliang University,Hangzhou 310018,China;National Institute of Metrology,Beijing 100029,China;State Key Laboratory of Precision Measurement Technology and Instrument,Tianjin University,Tianjin 300072,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2020年第2期33-39,共7页
Chinese Journal of Scientific Instrument
基金
国家重点研发计划(2018YFF0212702)项目资助.
关键词
机器人
惯性测量单元
姿态测量
数据融合
指数渐消记忆
拓展卡尔曼滤波
robot
inertial measurement unit
attitude measurement
data fusion
exponential fading memory
extended Kalman filter
