基于航迹预测的无人机短时航迹偏离检测方法

Short-time trajectory deviation detection method for UAV based on trajectory prediction

保障无人机(unmanned aerial vehicle,UAV)飞行安全已经成为推动无人驾驶航空创新应用与规模发展的关键问题。针对UAV在低空结构化航路网络运行过程中由航迹偏离导致的安全隐患,提出一种异常航迹检测方法(abnormal trajectory detection method,ATDM)。首先,建立航迹数据预处理和重构模型,构筑包含位置、速度、航向等多维属性的航迹数据。其次,以具有多维属性的航迹数据为输入,采用双向长短时记忆网络算法构建UAV短期航迹预测模型。最后,基于历史航迹点和预测航迹点间的多维度局部异常因子,将航迹偏离检测转化为航迹点密度分类问题,建立UAV航迹偏离检测方法,实现短时范围内航迹偏离状态的动态监测。结果表明,ATDM在短的预测时间范围内具有较好的精度优势和实时性。

Safeguarding unmanned aerial vehicle(UAV)flight safety has become a key issue in promoting the innovative application and scale development of unmanned aviation.To solve the safety hazards caused by trajectory deviation during the operation of UAV in low-altitude structured airway networks,a abnormal trajectory detection method(ATDM)is proposed.Firstly,a trajectory data preprocessing and reconstruction model is established to construct the trajectory data containing multidimensional attributes such as position,velocity,heading and so on.Secondly,the UAV short-term trajectory prediction model is constructed by using the trajectory data with multi-dimensional attributes as input,and the bidirectional long and short-term memory network algorithm is used.Finally,based on the multidimensional local anomaly factor between historical and prediction trajectory points,the trajectory deviation detection is transformed into a density classification problem of the trajectory points,and the UAV trajectory deviation detection method is established to realize the dynamic monitoring of the trajectory deviation status in the short-time range.Results show that ATDM has better accuracy advantage and real-time performance in short prediction time range.