In the envieronmntal vehicle trajectory prediction link of intelligent driving， to obtain the trajectory timing characteristics of the surrounding vehicles more accurately， the dropout layer is embedded into the long short-term memory （LSTM） neural networks model to enhance network generalization， and the attention mechanism is introduced into LSTM to distribute larger weight to the time series data with a marked influence on the prediction effect， so as to improve the reliability of the prediction results. Subsequently， the improved LSTM model is combined with the GRU model （LSTM-GRU） to further improve the prediction accuracy of the surrounding vehicle trajectory. On this basis， the LSTM-GRU model is trained， verified， and tested using the NGSIM public dataset. The results show that， compared with the standard LSTM and GRU model， the LSTM-GRU model integrating dropout and attention mechanism has advantages in predicting the surrounding vehicle trajectory with a longer time sequence， which improves the accuracy of trajectory prediction and reduces the root mean square error and mean absolute error between the actual and the predicted trajectory.