视觉技术辅助的无人机自主着陆组合导航研究

为提高无人机自主着陆过程中导航系统的自主性与精确性，设计了一种视觉辅助惯导组合导航方法。该方法以惯导误差方程为过程方程，以着陆过程中单目摄像机2个时刻所得地面特征点投影之间的双视图几何约束为量测方程，构建了非线性滤波器；利用SR-UKF方法实现了惯导误差估计，提高算法效率的同时有效地避免了UKF中由于矩阵开方运算导致的滤波失效；最后根据估计结果校正了惯导导航数据。仿真结果表明:该方法能够提高导航系统精度，使误差降低到惯导系统的8%左右。

Vision/INS integrated navigation for UAV autonomous landing

A vision/ inertial navigation system (INS) integrated navigation method was proposed to promote the accuracy and autonomy of navigation system for unmanned air vehicle(UAV) autonomous landing. Regarding the INS error equation as process model, and two-view geometry between projective points of identical feature in different instants as measure model, a nonlinear filter was built for INS error estimation. To expedite computation and avoid invalidity of unscented Kalman filter (UKF) algorithm, the square-root-UKF(SR-UKF) was used to estimate the INS error, and the navigation data was compensated by estimating result. Simulation results show that the proposed method is effective to improve navigation system accuracy, and able to reduce the error to 8% of INS.