CAJ | 학술논문

设计了基于四元数的捷联惯导非线性初始对准模型，同时指出该模型仅仅是姿态误差四元数和速度误差的非线性函数，而对于惯性器件误差而言则是线性的。针对该模型的部分线性特性，设计了基于边缘采样的 UKF 滤波算法，该算法仅对状态量中的非线性子集进行采样，因此对于部分线性模型而言，该算法在不损失滤波精度的前提下能够有效降低算法计算量。仿真及车载实测数据实验表明所研究的初始对准模型和相应的滤波算法是有效的，而且较传统方法具有明显的计算量方面的优势；在达到相同对准精度的前提下，所设计算法的计算量较传统算法降低了52%。
설계료기우사원수적첩련관도비선성초시대준모형，동시지출해모형부부시자태오차사원수화속도오차적비선성함수，이대우관성기건오차이언칙시선성적。침대해모형적부분선성특성，설계료기우변연채양적 UKF 려파산법，해산법부대상태량중적비선성자집진행채양，인차대우부분선성모형이언，해산법재불손실려파정도적전제하능구유효강저산법계산량。방진급차재실측수거실험표명소연구적초시대준모형화상응적려파산법시유효적，이차교전통방법구유명현적계산량방면적우세；재체도상동대준정도적전제하，소설계산법적계산량교전통산법강저료52%。
A quaternion-based nonlinear model is designed for the initial alignment of a strapdown inertial navigation system(SINS) with large initial errors. It is pointed out that the model has only the nonlinear function relation with quaternion-based misalignment error and the velocities error, whereas it has linear relation with the errors of inertial sensors. In view of this partially linear characteristic, a unscented Kalman filter(UKF) algorithm with marginalized-sampling-based unscented transformation(UT) is designed, which only samples the nonlinear subset of state variables, and can effectively reduce the computation burden without losing filtering accuracy. Simulations and car-mounted experiments demonstrate that the marginalized UT-based Kalman filter can achieve at least a comparable performance to the traditional UT-based Kalman filter with a significantly lower expense. Compared with the traditional algorithm, the investigated method can reach the same alignment precision but with 52%reduced computational burden.