基于视觉的惯性导航误差在线修正

陀螺零偏和加速度计零偏是影响惯性测量单元(IMU)积分精度的重要因素。提供一组精确的实时的零偏估计可以提高IMU的积分精度,为视觉导航提供良好的位姿预测,提高整个系统的动态性能。通过合理地建立IMU的噪声模型以及IMU和视觉的组合方程,利用一种基于李群和李代数知识的IMU预积分方法将零偏进行合理的线性化,运用Kalman滤波进行IMU零偏的在线估计。实验结果表明,通过本文的修正方法,惯性导航的平均积累误差由0.034m/s提高到0.0037m/s,精度明显提高。

The Online Correction of IMU Biases for Visual-Inertial Navigation

The IMU(Inertial Measurement Unit, here refers to gyroscopes and accelerometers) biases are the major factors affecting the accuracy of inertial navigation system. Accurate real-time estimation of IMU biases can improve the integration accuracy of IMU, better predict the position and attitude for camera and improve the dynamic performance of the navigation system. The noise model of IMU and visual-inertial combined models are properly developed. The equations of IMU biases are linearized through preintegrating IMU measurements based on Lie group and Lie algebra, and IMU biases are estimated by using the Kalman filter. The experimental results show that the accuracy of inertial navigation system is improved obviously. The average accumulation error of inertial navigation system is increased from 0.034m/s to 0.0037m/s, and the accuracy is improved obviously.