Research on Multi-objective Trajectory Prediction Algorithm Based on Driving Intent Classification

The trajectory prediction of the multi-object vehicle ahead plays a huge role in improving the vehicle's driving safety and improving the planning and traffic efficiency of the vehicle. However, due to the uncertainty of the driving intention of the multi-object vehicle and the uncertainty of the vehicle dynamics, its trajectory prediction faces a huge challenge. First, the fuzzy C-means (FCM) method is used for multi-object trajectories. After offline training is carried out using the relevant information of the vehicle trajectory, the driving intention can be automatically fuzzy classification. Secondly, the long short-term memory (LSTM) method uses the history and current trajectory information of multi-object vehicles ahead to predict the future trajectory under different driving intentions. Then, according to the fuzzy classification results of the driving intention, the predicted trajectories are merged, and the trajectory prediction of the multi-objective vehicle ahead in the next 1s is realized through an iterative method. Finally, use real vehicle data for experimental verification. The results show that 92.1% of the lateral distance prediction error in the tenth step is less than 0.13m. The maximum distance prediction error is 0.54m. In the longitudinal distance, the prediction error of 94.5% is less than 0.6m. The maximum distance prediction error is 1.0m.