By using vehicle-to-vehicle (V2V) communication technology to interconnect trains while maintain a shorten distance under the premise of safety condition, is the development direction for improving the efficiency of high-speed trains. Trajectory prediction of train ahead is an important mean to further reduce the tracking distance. In this paper, based on Gaussian mixture model (GMM) and long short-term memory (LSTM) Recurrent Neural Network (RNN), we propose a personalized trajectory prediction method model for high-speed trains. The main idea is to achieve accurate and personalized trajectory prediction by recognizing the driving style of the train ahead to realize a shorter distance tracking control. Firstly, based on the GMM, three different driving styles are identified by combining the characteristic data of tracking trains, and the characteristic importance of driving styles are analyzed by MIC. Secondly, based on different driving styles, a novelty personalized trajectory prediction algorithm is worked out by modified LSTM-RNN models. Finally, experiments are carried out using the real data of the on-board equipment and ground control equipment. The results indicate that, compared with the traditional trajectory prediction methods, the proposed personalized trajectory prediction method shows significant advantages.
Trajectory Prediction of High-Speed Train Based on GMM-LSTM
24.09.2023
2887076 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Pedstrian Trajectory Prediction Based on LSTM
IEEE | 2023
|Speed trajectory optimization for a high-speed train
TIBKAT | 2020
|Research on Fault Prediction of High-Speed Train Auxiliary Power Supply System Based on LSTM
Springer Verlag | 2022
|Space-time attention LSTM vehicle trajectory prediction method based on position-speed
Europäisches Patentamt | 2023
|