摘要
为准确预测滑行时间,提出一种基于机场场面运行态势演变的图卷积神经网络预测方法。首先,根据机场场面航空器时空分布情况,从路段流量、路段密度、路段速度等多角度构建交通态势指标体系;其次,利用主成分分析法对指标进行降维处理并利用K-means算法实现对机场场面路段的态势等级划分,绘制机场场面时空分布热力图;最后,利用图卷积神经网络(GCN)结合门控循环单元(GRU)来获取场面路段特征数据的时空特征,将GRU作为解码器预测输出滑行时间。以深圳宝安国际机场AirTOP仿真数据为例,对所提出的方法进行了分析和验证,并获得了符合预期的预测结果。实验结果表明,该方法在预测滑行时间方面具有有效性。
To accurately predict taxiing time,this study proposes a graph convolutional neural network prediction method based on the evolution of airport surface operation situation.Firstly,based on the spatiotemporal distribution of aircraft on the airport surface,a traffic situation indicator system is constructed from multiple perspectives such as road flow,road density,and road speed.Secondly,the principal component analysis method is used to reduce the dimensionality of the indicators and the K-means algorithm is used to achieve the classification of the situation level of the airport surface road sections,and to draw a spatiotemporal distribution heatmap of the airport surface.Finally,using Graph Convolutional Neural Network(GCN)combined with Gated Recurrent Unit(GRU)to obtain the spatiotemporal features of the road segment feature data,the GRU is used as the decoder to predict the output sliding time.This study takes the simulation data of AirTOP at Shenzhen Bao'an International Airport as an example to analyze and verify the proposed method,and obtains expected prediction results.The experimental results indicate that this method is effective in predicting taxi time.
作者
彭瑛
侯婧娉
宛照坤
孙钰
PENG Ying;HOU Jing-ping;WAN Zhao-kun;SUN Yu(Nanjing University of Aeronautics and Astronautics,Nanjing 211000,China;National Key Laboratory of Air Traffic Management System,Nanjing 211000,China)
出处
《航空计算技术》
2024年第4期1-6,共6页
Aeronautical Computing Technique
基金
国家重点研发计划项目资助(2022YFB2602401)
民航安全能力建设基金项目资助。
关键词
机场场面
K-MEANS聚类
主成分分析法
图卷积神经网络
滑行时间预测
airport scene
K-means clustering
principal component analysis method
graph convolutional neural network
taxi time prediction
