Multi-Module Fusion Vehicle Trajectory Prediction Based on Generative Adversarial Networks

Vehicle trajectory prediction, as one of the core technologies in intelligent transportation and autonomous driving, plays a significant role in reducing collision risks. However, complex spatiotemporal dependencies and dynamic interactions remain challenging technical issues that require urgent resolution. To address these challenges, this paper proposes a GAN-based multi-module fusion prediction model. By integrating multiple architectures-including a graph attention network, a Transformer encoder, and a temporal convolutional network-the model simultaneously optimizes spatial interactions, temporal dependencies, and trajectory feature modeling. Experimental results demonstrate that, compared with baseline methods such as CV, CS-LSTM, and TCN-LSTM, the proposed model significantly improves prediction accuracy by reducing the average displacement error and final displacement error by 24.38% and 14.46%, respectively, thereby fully validating its effectiveness in complex traffic scenarios.