基于改进卷积网络的终端区4D航迹预测与冲突检测

随着不断扩大的旅客运输量和航线网络规模，采用飞行计划结合空中交通管制的空中管理办法已经不能与当前民航需求和空中交通流量相匹配，直接影响到航班正常率和运行安全。为解决这一问题，国际民航组织(International Civil Aviation Organization, ICAO)提出了基于航迹运行(trajectory based operation, TBO)的下一代空中交通管理运行理念，中国民航也提出了智慧民航的建设方案和目标。其中4D航迹是TBO运行的核心组成部分，也是中国建设智慧民航的重要技术指标，其可以对航空器的运行进行精确地管理和控制。因此，提高4D航迹预测的准确性成为了目前急需解决的核心问题。面向航空器的飞行任务实施阶段，从4D航迹预测和冲突检测两个问题进行了研究。在航迹预测方面，采用了基于卷积神经网络-双向门控循环单元(convolutional neural networks-bidirectional gated recurrent unit, CNN-BiGRU)的模型对航迹进行高精度预测；在冲突检测方面，引入了航迹距离检测函数以检验预测模型生成的两条航迹是否...

Improved Convolutional Network Based 4D Trajectory Prediction and Conflict Detection in Terminal Areas

With the continuous expansion of passenger volume and the scale of flight network, the method of air traffic management using flight plans combined with air traffic control has been found to be insufficient to meet the current demands of civil aviation and the volume of air traffic, which directly affects flight regularity and operational safety. To address this issue, the International Civil Aviation Organization (ICAO) has proposed the next-generation concept of air traffic management based on Trajectory Based Operation (TBO), and the Civil Aviation Administration of China(CAAC) has also put forward its construction plan and goals for intelligent aviation. The 4D trajectory is a core component of TBO and an important technical indicator for China''s smart aviation construction, which can precisely manage and control the operation of aircraft. Therefore, it has become an urgent core issue to improve the accuracy of 4D trajectory prediction. In this paper, research was conducted focusing on the flight task implementation stage of aircraft. Both 4D trajectory prediction and conflict detection were addressed. In terms of trajectory prediction, a CNN-BiGRU-based model was used for high-precision trajectory prediction; as in conflict detection, a trajectory distance detection function was introduced to check whether there is a conflict between two trajectories generated by the prediction model. The experiments were carried out using the real historical ADS-B trajectory data of a busy terminal area and were compared with the LSTM model and the GRU model on the same set of data. The simulation experiment showed that the CNN-BiGRU model was superior to the comparative models in terms of accuracy and evaluation metrics, and the detection result showed no conflict within the next 800 seconds. The method proposed in this paper provides an effective tool for the management of air traffic and has a significant value in the application.