观测卫星不足时的GPS/INS紧组合抗差自适应滤波

常规GPS/INS紧组合抗差自适应滤波只适用于卫星数≥4的情况,且预测残差构造自适应因子要求观测值可靠。针对该局限性,对常规抗差自适应滤波算法做出两点改进:1)采用两步滤波,用第1步常规EKF滤波残差构造第二步抗差算法的粗差判别量;2)在第2步滤波用预测残差构造自适应因子时,剔除异常观测值对应的预测残差和预测残差协方差,以削弱观测异常对自适应因子的不良影响。实验结果表明,常规抗差算法在卫星数4时不适用。常规自适应滤波算法在观测值存在异常的情况下无法正确修正模型异常。改进后的抗差自适应滤波算法在组合系统观测卫星数4且观测值存在异常的情况下,仍能正确修正观测粗差和动力学模型异常,能够达到良好的导航精度。

Robust Adaptive Filtering of GPS/INS Tightly Integrated System with Observation Satellite Deficient

The conventional robust adaptive filtering of GPS/INS tightly integrated system is only available under the condition of more than 4 satellites,and construction of adaptive factor utilizing predicted residuals demands reliable observed value.For these two limitations,two improvements are proposed.Firstly,the two-step filter are utilized.Conventional EKF filtering residuals in step 1 are utilized to construct the gross error identification of robust algorithm in step 2.Secondly,the adaptive factor in step 2 is constructed using predicted residuals,among which the residuals and its covariance of abnormal values are removed to weaken the harmful effect from abnormal observation to adaptive factor.Finally,the experiment results show that conventional robust algorithm is not available under the condition of less than 4 satellites,and model abnormity cannot be correctly modified in the conventional adaptive algorithm when observation abnormity exists.Gross errors and dynamic model abnormities can be correctly modified in the improved robust adaptive algorithm when the satellite number is less than four and observed values are not reliable.Then a favorable navigation precision can be achieved.