高斯混合模型优化的四元数约束CKF姿态估计

针对四元数非线性滤波在姿态估计中噪声为非理想高斯分布时随机模型的失配现象，提出了一种基于高斯混合模型的四元数约束容积卡尔曼滤波（GM-QCKF）。讨论了四元数在容积卡尔曼滤波中的加权均值方法，并利用高斯混合模型对高斯分量更新与求和从而精化随机模型，最后通过两步投影理论做归一化限制，有效地改进了四元数约束容积卡尔曼滤波的稳定性。通过仿真测试，表明GM-QCKF可明显提高姿态估计精度；跑车试验验证了GM-QCKF算法在实际应用中精度和稳定性的优势。

Quaternion Constrained Cubature Kalman Filter Attitude Estimation Based on Gaussian Mixture Model

A quaternion cubature Kalman filter improved by the Gaussian mixture model (GM-QCKF) is proposed to solve the model mismatch problem of a quaternion nonlinear filter with non-gaussian noise distribution in attitude estimation. The weighted mean algorithm of quaternion based CKF is discussed; every gaussian component is updated and summed by the gaussian mixture model to refine the stochastic model; and the normalization is obtained by the two-step projection theory. These improvements make quaternion nonlinear updating algorithm more stable. The simulation comparison experiments show that the GM-QCKF can improve the precision of the attitude estimation. And the car trial experiment is conducted. The results indicate that compared with the traditional algorithm the GM-QCKF is superior in accuracy and stability.