基于深度学习的航空器场面轨迹预测

Aircraft Surface Trajectory Prediction Based on Deep Learning

轨迹预测研究是安全高效控制场面滑行的重要基础,在路由规划,风险预警,航班次序,重要节点的时间安排等都能起到重要作用。利用深度学习中循环神经网络的长期记忆性特点,对航空器场面历史数据进行分析和预处理,设定网络模型参数,构建轨迹预测模型,提出了一种基于深度学习的航空器场面滑行轨迹预测方法。结合场面航空器运动状态的变化,改进长短期记忆网络的隐藏层结构,实现对航空器场面轨迹的中期预测。

Trajectory prediction research is an important basis for safe and efficient control of surface taxiing. It can play an important role in routing planning, risk warning, flight sequence, and timing of important nodes. Using the long-term memory characteristics of the recurrent neural network in deep learning, the historical aircraft scene data is analyzed and preprocessed, the network model parameters are set, and the trajectory prediction model is constructed. A deep learning-based aircraft taxi trajectory prediction method is proposed. Combining with the changes of the aircraft motion state on the scene, improve the hidden layer structure of the long and short-term memory network to realize the mid-term prediction of the aircraft scene trajectory.