부산대학교 도서관

To handle two shortcomings of existing methods, (i) nearly all models rely on the high-definition (HD) maps, yet the map information is not always available in real traffic scenes since map-building is expensive and time-consuming and (ii) existing models usually improve prediction accuracy at the expense of reducing computing efficiency, yet the efficiency is crucial for various real applications, this paper proposes an efficient trajectory prediction model that is not dependent on traffic maps. The core idea of our model is to encode single-agent's spatial-temporal information in the first stage and explore multi-agents' spatial-temporal interactions in the second stage. By comprehensively utilizing attention mechanism, long short-term memory, graph convolutional network, and temporal transformer to decouple spatial and temporal features on both the single agent and the multiple agents level, our model is able to learn rich dynamic and interaction information of all agents. Our model achieves the highest performance when comparing with existing map-free methods and exceeds most map-based state-of-the-art methods on the Argoverse and nuScenes datasets. In addition, our model also exhibits a faster inference speed than many state-of-the-art models.