A Conflict-Aware and Dynamically Adaptive Trajectory Prediction Method for Multi-Agent Traffic

DOI

10.2991/978-94-6239-740-8_2How to use a DOI?

Keywords

Multi-Agent Interaction; Dynamic Model Switching; Safety-Aware Prediction

Abstract

In complex multi-agent traffic environments, the accuracy and safety awareness of trajectory prediction directly affect the reliability of decision-making in autonomous driving systems. Existing approaches primarily focus on fitting historical motion patterns, while lacking effective mechanisms for modeling interactive conflicts and risk-inducing behaviors, making it difficult to achieve stable prediction in high-density scenarios. To address these limitations, this paper proposes a conflict-aware and dynamically adaptable trajectory prediction framework for multi-agent traffic scenes. First, a two-dimensional time-to-collision (2D-TTC) matching mechanism is developed to jointly evaluate longitudinal and lateral TTC values, enabling the selective identification of influential interacting agents and suppressing irrelevant interaction noise at the source. Second, a dynamic model-switching strategy based on global interaction intensity is introduced to adaptively adjust network complexity according to traffic density and risk levels, thereby balancing prediction accuracy and computational efficiency. Furthermore, a conflict-aware weighted loss function is incorporated during training to enhance the model's sensitivity to high-risk interaction samples at the parameter-learning level, improving its ability to recognize potentially dangerous behaviors. Experiments conducted on the Waymo Open Dataset demonstrate that the proposed method achieves reductions in Average Displacement Error (ADE) and Final Displacement Error (FDE) compared with representative baseline models, while simultaneously improving inference efficiency. In addition, the method exhibits stronger risk perception and earlier conflict recognition under the post-hoc safety metric Alert Rate (AR). These results indicate that the proposed framework significantly enhances safety awareness and scene adaptability without compromising prediction accuracy, offering new insights and practical engineering value for safety-critical modeling and risk-aware trajectory prediction in intelligent transportation and autonomous driving systems.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Chen Wang
AU  - Haojie Li
AU  - Yuwen Zhang
AU  - Gang Ren
PY  - 2026
DA  - 2026/08/14
TI  - A Conflict-Aware and Dynamically Adaptive Trajectory Prediction Method for Multi-Agent Traffic
BT  - Proceedings of the 2026 9th International Symposium on Traffic Transportation and Civil Architecture (ISTTCA 2026)
PB  - Atlantis Press
SP  - 4
EP  - 18
SN  - 2589-4943
UR  - https://doi.org/10.2991/978-94-6239-740-8_2
DO  - 10.2991/978-94-6239-740-8_2
ID  - Wang2026
ER  -