基于改进的磁场检验和主方向算法的行人航向估计

针对以惯导为核心的行人定位算法存在航向角失真的问题，提出了一种融合惯性测量单元(IMU)和磁力计的行人导航算法，利用步幅航向的变化量将行人的运动分为直线行走和曲线行走两部分，当检测到行人直线行走时，提出一种实时建立主方向的自适应航向角漂移修正算法，对航行角误差进行补偿；当行人曲线行走时，辅助地磁场信息，使用磁场强度、地磁倾角、惯性导航系统(inertial navigation system，INS)与磁力计解算的航向角之差这三个特征值构建广义似然比检验量，对磁场进行判断获得稳定状态磁场信息，通过卡尔曼滤波对航向角误差进行补偿。将传感器置于足部进行实验，实验结果表明,两种算法可以实现优势互补，有效地抑制地磁信息的干扰，提高导航精度，算法结果相对于无航向角修正的零速修正算法、零角速度更新算法和单独使用磁力计进行融合的算法，定位误差分别降低了91.02%、82.93%和61.39%，室外和室内终点定位误差分别为4.545 4 m和0.543 6 m，占总路程的0.69%和0.17%。

Pedestrian heading estimation based on improved magnetic inspection and principal direction algorithm

Aiming at the issue of heading distortion in the pedestrian positioning algorithm that relies on inertial navigation, the pedestrian navigation algorithm integrating inertial measurement unit(IMU) and magnetometer is proposed. The pedestrian movement is divided into straight walking and curve walking based on the alteration in stride heading. When the pedestrian is detected to be walking along a straight path, the adaptive heuristic drift elimination algorithm for real-time establishment of the principal direction is proposed to compensate the navigation angle error. When the pedestrian walks along a curve path, the geomagnetic field information is assisted. The generalized likelihood ratio test is constructed by using the three eigenvalues of the magnetic field intensity, the geomagnetic inclination angle, and the difference between the heading angle calculated by inertial navigation system(INS) and the magnetometer. The magnetic field is inspected to obtain quasi-static magnetic field information, and the heading angle error is compensated by Kalman filter. The sensor is placed on the foot for experiments. The experimental results show that the two algorithms can achieve complementary advantages, effectively suppress the interference of geomagnetic information, and improve the navigation accuracy. Compared with the zero velocity update algorithm, the zero angular rate update algorithm, and the fusion algorithm using the magnetometer alone, the positioning errors have reduced 91.02%, 82.93% and 61.39%, respectively. The endpoint positioning errors in outdoor and indoor experiments are 4.545 4 m and 0.543 6 m, accounting for 0.69% and 0.17% of the total distance.