基于视觉/惯导的无人机组合导航算法研究

Research on Vision Inertial Based Integrated Navigation Technology of UAVs

摘要:

目前视觉惯性组合导航系统多采用优化紧/松耦合以及滤波紧/松耦合算法，应用误差状态卡尔曼滤波能够将较低频率的视觉位姿信息提升到与惯性信息同步的频率。提出一种基于自适应卡尔曼滤波的视觉惯导组合导航算法，首先考虑到系统建模与传感器测量误差，采用自适应渐消卡尔曼滤波进行导航解算，通过实时计算遗忘因子，以调节历史数据的权重，可抑制建模误差，提高组合导航系统性能，然后针对视觉SLAM解算过程造成的视觉位姿信息滞后于惯导信息的问题，提出一种延时补偿方法。仿真实验表明，采用延时补偿的自适应渐消卡尔曼滤波算法能够有效抑制建模误差，并降低视觉位姿信息滞后带来的影响，提高无人机组合导航的解算精度，姿态、速度、位置解算精度分别达到5°、0.5m/s、0.4m以内。

Abstract:

At present, most of the vision inertial integrated navigation systems use optimization tight/loose coupling method based on optimize or filter. The application of error state Kalman filter can raise the frequency from visual pose to synchronized with IMU. A vision integrated navigation algorithm based on Adaptive Kalman filter is proposed. Firstly, considering the system modeling and sensor measurement error, adaptive fading Kalman filter is used to solve the navigation problem. By calculating the forgetting factor in real time, the weight of historical data can be adjusted, the modeling error can be suppressed, and the performance of integrated navigation system can be improved. In view of the problem that the visual pose information lags behind the inertial information caused by the visual slam solution process, a delay compensation method is proposed to solve the problem. The simulation results show that the adaptive fading Kalman filter algorithm with time delay compensation can effectively suppress the modeling error, reduce the influence of visual pose information lag, and improve the accuracy of integrated navigation of UAVs, the accuracy of attitude, velocity and position is within 5 °, 0.5m/s and 0.4m respectively.