Answers are grounded in this paper and cite their sources.
Accurate vehicle trajectory prediction (VTP) plays a fundamental role in the advancement of intelligent transportation systems and autonomous driving. It improves situational awareness, optimizes decision-making and safety in complex traffic scenarios. Although several surveys have reviewed vehicle trajectory prediction approaches, a comprehensive review of recent deep learning-based approaches from a system and modeling perspective is still lacking. To bridge this gap, we conducted a systematic review of the literature on deep learning-based methods by critically analyzing emerging trends in model design. We introduce a system-oriented taxonomy of vehicle trajectory prediction methods which includes classical, reinforcement learning-based, collaborative, and deep learning-based methods. We evaluate their capacity to model uncertainty, inter-agent interactions, and dynamic traffic behavior. This survey emphasizes three key perspectives that drive performance in deep learning-based models including data preprocessing and feature extraction, model creation and attention mechanisms. Deep-learning-based methods are critically analyzed in design dimensions, e.g. temporal modeling strategy, interaction encoding, modeling capability, suitable deployment scenarios and their key strengths for trajectory prediction. Finally, we discuss open research challenges and propose practical recommendations for next-generation VTP in autonomous driving.
No takes yet. Share an insight, caveat, or question.
Rashid et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: