基于运行状态识别的无人机航迹预测

无人机航迹预测对于无人机冲突检测、任务规划及异常管控至关重要。在很多情况下难以为无人机这种复杂系统建立精确的物理模型,给基于模型的滤波方法带来一定难度。为解决上述问题,提出一种基于运行状态识别的无人机高斯过程-无味卡尔曼滤波的混合估计方法。首先,利用运行状态识别机制将无人机运行数据分为不同数据段,以确定无人机实时状态并提高预测模型的适应性;然后,根据不同的运行状态,从航迹数据中学习高斯过程递归模型,将其作为无味卡尔曼滤波器的状态转移方程,以实现更高的预测精度;最后,利用动作捕捉系统采集的真实无人机运行数据验证了所提出方法的有效性,利用均方误差检验了方法的精确度。

UAV track prediction based on operating state recognition

: UAV trajectory prediction is critical for UAV collision detection, mission planning, and anomaly management. In many cases, it is difficult to establish an accurate physical model for a complex system such as a drone, which brings difficulty to the model-based filtering method. In order to solve the above problems, this paper proposes a hybrid estimation method based on running state recognition for Gaussian process-unscented Kalman filter. The method uses the running state recognition mechanism to divide the drone operation data into different data segments to determine the real-time status of the drone and improve the adaptability of the prediction model. Then, according to different operating states, the GP recursive model is learned from the track data and used as the state transition equation of the unscented Kalman filter to achieve higher prediction accuracy. Finally, the effectiveness of the proposed method is verified by the real drone operation data collected by the motion capture system. The accuracy of the method is verified by the mean square error.