STRAP: Spatial-Temporal Risk-Attentive Vehicle Trajectory Prediction for Autonomous Driving

2025-07-11

AI Summary

Existing vehicle trajectory prediction methods often neglect the safety risks arising from surrounding vehicles’ uncertainty and aggressive maneuvers. To address this, we propose a spatio-temporal risk-aware prediction framework that pioneers the integration of interpretable risk potential field modeling and a risk-attentive mechanism into the prediction pipeline. Specifically, we construct a dynamic risk potential field to explicitly characterize local traffic risk, design a risk-weighted loss function to prioritize learning in high-risk scenarios, and adopt a spatio-temporal encoder–risk-modulated decoder architecture to enable risk-aware feature fusion. Evaluated on the NGSIM and HighD datasets, our method achieves average prediction errors 4.8% and 31.2% lower than state-of-the-art approaches, respectively—with particularly pronounced improvements in critical high-risk situations such as short inter-vehicle distances. The framework significantly enhances both prediction accuracy and safety robustness.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Accurate vehicle trajectory prediction is essential for ensuring safety and efficiency in fully autonomous driving systems. While existing methods primarily focus on modeling observed motion patterns and interactions with other vehicles, they often neglect the potential risks posed by the uncertain or aggressive behaviors of surrounding vehicles. In this paper, we propose a novel spatial-temporal risk-attentive trajectory prediction framework that incorporates a risk potential field to assess perceived risks arising from behaviors of nearby vehicles. The framework leverages a spatial-temporal encoder and a risk-attentive feature fusion decoder to embed the risk potential field into the extracted spatial-temporal feature representations for trajectory prediction. A risk-scaled loss function is further designed to improve the prediction accuracy of high-risk scenarios, such as short relative spacing. Experiments on the widely used NGSIM and HighD datasets demonstrate that our method reduces average prediction errors by 4.8% and 31.2% respectively compared to state-of-the-art approaches, especially in high-risk scenarios. The proposed framework provides interpretable, risk-aware predictions, contributing to more robust decision-making for autonomous driving systems.

Problem

• Predicting vehicle trajectories accurately for autonomous driving safety
• Addressing uncertain or aggressive behaviors of surrounding vehicles
• Improving prediction accuracy in high-risk scenarios like short spacing

Innovation

• Incorporates risk potential field for risk assessment
• Uses spatial-temporal encoder and risk-attentive decoder
• Employs risk-scaled loss for high-risk scenarios