GSGFormer: Generative Social Graph Transformer for Multimodal Pedestrian Trajectory Prediction
This addresses the problem of accurate and diverse trajectory forecasting for self-driving cars and robots, though it appears incremental as it builds on existing graph and transformer methods.
The paper tackled pedestrian trajectory prediction by proposing GSGFormer, a generative model that incorporates interactions and multimodal behaviors, achieving state-of-the-art performance on multiple datasets with competitive results even in data-limited scenarios.
Pedestrian trajectory prediction, vital for selfdriving cars and socially-aware robots, is complicated due to intricate interactions between pedestrians, their environment, and other Vulnerable Road Users. This paper presents GSGFormer, an innovative generative model adept at predicting pedestrian trajectories by considering these complex interactions and offering a plethora of potential modal behaviors. We incorporate a heterogeneous graph neural network to capture interactions between pedestrians, semantic maps, and potential destinations. The Transformer module extracts temporal features, while our novel CVAE-Residual-GMM module promotes diverse behavioral modality generation. Through evaluations on multiple public datasets, GSGFormer not only outperforms leading methods with ample data but also remains competitive when data is limited.