基于分解四元数的自适应姿态四元数卡尔曼滤波  

Adaptive attitude quaternion Kalman filter based on factored quaternions

作　　者：张德先 聂桂根[1,2] ZHANG De-xian;NIE Gui-gen(GNSS Research Center,Wuhan University,Wuhan Hubei 430079,China;Collaborative Innovation Center for Geospatial Information Technology,Wuhan Hubei 430079,China)

机构地区：[1]武汉大学卫星定位导航技术研究中心,湖北武汉430079 [2]地球空间信息技术协同创新中心,湖北武汉430079

出　　处：《控制理论与应用》2018年第3期367-374,共8页Control Theory & Applications

基　　金：国家重点基础研究发展计划项目(2013CB733205);武汉市科学技术局:应用基础研究计划项目(2015011701011639)资助~~

摘　　要：基于Clifford代数的四元数卡尔曼滤波在融合陀螺/加表/磁强计以估计姿态时,由于四元数各参数与欧拉角不是一一对应关系,无法独立估计各个欧拉角.这样即使重力观测量是可信的,受到干扰的磁场观测量也会影响整个估计结果.为了消除磁场观测量对四元数中横滚角和俯仰角分量的影响,对四元数进行分解,以重新组合重力/磁场观测量.同时,为了减少载体附近磁场和线性加速度干扰对姿态估计的影响,构造了观测噪声自适应算法和观测量自适应干扰补偿.消费级微机电系统(micro-electro-mechanical system,MEMS)传感器的实验结果表明,对比四元数卡尔曼滤波的原型,改进后的抗干扰能力明显提升.但由于自适应过程引入了两个经验参数,这使得其工作范围和抗干扰能力有待考验.The quaternion Kalman filter based on Clifford algebra cannot estimate each Euler angle separately while fusing gyro/accelerator/magnetometer observations because the parameters of quaternions are not one-to-one corresponding to the parameters of Euler angles.That makes the whole estimation results affected by disturbed magnetic field observations,even though the accelerator observations are reliable.In order to eliminate the affection to roll and pitch angle components in quaternions caused by magnetic field observations,this paper recombined the observations of gravity and magnetic field with factored quaternions.Beside,in order to reduce the impact of magnetic and linear acceleration interference to attitude estimation,this paper constructed adaptive noise covariance of observations and adaptive interference compensations of observations.Compared to the original quaternion Kalman filter,experiments based on consumptionlevel micro-electro-mechanical system(MEMS)sensors showed that the anti-interference ability of the modified one has been improved markedly.However,two empirical parameters were introduced into the adaptive procedure.Therefore,its working range remains to test.

关 键 词：卡尔曼滤波 四元数 自适应算法 姿态估计 惯性传感器 磁强计 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]