Aircraft Taxiing Trajectory Prediction Based on Multi-embedding Features
To meet the requirements of airport surfaces intelligent management,this paper innovatively in-troduces the node2vec algorithm and ROUGE-N evaluation system from the NLP field into aircraft taxiing trajectory prediction research.By establishing a topological structure map of the airport surface,a new multi-feature embedding method for taxiing trajectories is created and framework is developed for long-term trajectory prediction.Taking Shenzhen Airport as an example,we proposed an attention mechanism integrated Bi-LSTM-A prediction model and compared its performance with traditional models like RNN,LSTM,and Bi-LSTM.In view of multi-feature embedding method,it is demonstrated that the Bi-LSTM-A model surpasses the baseline models by an average of 10.11%in precision,14.76%in recall rate,and 12.78%in F1 score.This indicates that the proposed predictive technique significantly en-hances the accuracy of long-term taxiing trajectory predictions and can effectively estimate the operational status of the airport based on flight schedules,thereby improving the intelligence and efficiency of airport ground operations.
airport surfacetaxiing trajectory predictionnatural language processingfeature extractionattention mechanism
万方数据
维普
