A Trajectory Prediction Method Based on Bayonet Importance Encoding and Bidirectional Lstm

Predicting travel trajectory of vehicles can not only provide users with efficient travel suggestions, but also help traffic managers make better decisions and planning. This paper aims to address the next location prediction problem to enable the prediction of the city-wide movement trajectory of an individual vehicle by considering where the vehicle will go next. We propose a deep learning model based on Bi-LSTM neural network with self-attention mechanism that fuses both positional encoding and bayonet importance encoding. In particular, to obtain the bayonet importance encoding, we introduce embedding word vectors in natural language processing and improve the PageRank algorithm. At the same time, we cluster all trajectories according to the trajectory sentence vector, and achieve better results by building a prediction model in each cluster. Using real bayonet vehicle trajectory data, we compare the predictive performance of the proposed model with several existing models, including LSTM, GRU and Bi-LSTM models. The result shows that our proposed model yields higher accuracy for the location prediction task