陀螺零偏二次模型参数的系统级辨识算法

以单轴旋转光学捷联惯性导航系统为原型,假设水平陀螺常值漂移的影响得以完全调制,方位陀螺漂移为随时间变化的二次模型,在水平阻尼工作模式下推导了系统位置误差与方位陀螺漂移之间严格的数学关系。分别设置了方位陀螺漂移仅有常值项、一次项、二次项和全系数误差的误差模型,利用递推最小二乘算法成功辨识出设定的二次模型中各个参数值。仿真结果表明,常值项首先被辨识出来,估计时间约为14 h,估计误差为6.54e-6(°)/h;一次项系数估计时间约为30 h,估计误差为2.73e-8(°)/h;二次项系数估计时间约为42 h,估计误差为1.51e-9(°)/h;全系数估计需要45 h,估计误差为7.28e-6(°)/h。辨识结果验证了该算法的正确性。实际系统中,可适当增加总的辨识估计时间,以达到更高精度的辨识结果。

System-level identification algorithm for quadratic model parameters of gyro bias

Based on the single-axis SINS,the strict mathematical relationship between the system position err and the azimuth gyro drift error was derived under the level damp mode by assuming that the level-gyro constant drift could be fully modulated and the azimuth gyro drift was changed with time and could be expressed as a quadratic model.The error models of azimuth gyro drift were set when with constant terms,one-time items,the quadratic terms and the all-coefficient model,respectively.Various parameters in the quadratic model were effectively identified by using a recursive least squares algorithm.Simulation results show that the constant coefficient can be identified firstly,the estimated time is in about 14 h,the estimated error is 6.54e-6(°)/h;then the one-time items coefficient is identified,and the estimated time is about 30 h,the estimated error is 2.73e-8(°)/h;finally the quadratic coefficient is identified,and the estimated time is about 42 h,the estimated error is 1.51e-9(°)/h.The estimated time of the all-coefficient model is about 45 h,and the estimated error is 7.28e-6(°)/h.The identification results demonstrate the correctness of the algorithm.In order to achieve better recognition results in the practical engineering system,we can increase the total identification time appropriately.