Lane Change Trajectory Prediction Based on Chinese Highway Ramp Scenarios

Accurately predicting the future trajectories of autonomous vehicles is crucial for achieving safe and efficient merging onto the main road in highway ramp scenarios. However, the dense and interactive environment of highway ramp merging areas presents significant challenges for trajectory prediction. This paper proposes a novel lane-changing trajectory prediction model, combining Long Short-Term Memory (LSTM) and Graph Attention Network (GAN), specifically designed for Chinese highway ramp scenarios. The model utilizes LSTM encoders to extract historical trajectory features of the autonomous vehicle and its immediate surrounding, encompassing eleven vehicles. Additionally, a GAN is employed to capture the complex interaction among the vehicles. The LSTM decoder generates accurate predictions of the future trajectory of the autonomous vehicle. Experimental evaluations are conducted using the CKQ3 dataset from the Ubiquitous Traffic Eyes open-source dataset to validate the effectiveness of the proposed model. The results demonstrate that the model achieves superior accuracy in predicting the future trajectories of lane-changing vehicles in Chinese highway ramp scenarios.