基于空间语义分析的轨迹预测技术综述  

Review of Trajectory Prediction Techniques Based on Spatial Semantic Analysis

作　　者：杨博渊 张力航 李成 乔少杰 李任杰 李庆 黄明文 罗娜 YANG Boyuan;ZHANG Lihang;LI Cheng;QIAO Shaojie;LI Renjie;LI Qing;HUANG Mingwen;LUO Na(School of Software Engineering,Chengdu University of Information Technology,Chengdu 610225,China;Chengdu Public Security Information Technology Research Institute,Chengdu Municipal Public Security Bureau,Chengdu 610017,China;Key Laboratory of Cyberspace Security,Ministry of Education of China,Zhengzhou 450001,China;Henan Key Laboratory of Cyberspace Situation Awareness,Zhengzhou 450001,China)

机构地区：[1]成都信息工程大学软件工程学院,四川成都610225 [2]成都市公安局科技信息化处,四川成都610017 [3]网络空间安全教育部重点实验室,河南郑州450001 [4]河南省网络空间态势感知重点实验室,河南郑州450001

出　　处：《无线电工程》2025年第3期611-620,共10页Radio Engineering

基　　金：国家自然科学基金(62272066);四川省科技计划(2023YFG0027,2024YFFK0413);成都市技术创新研发项目(2024-YF05-01217-SN);成都市技术创新研发项目重点项目(2024-YF08-00029-GX);网络空间安全教育部重点实验室及河南省网络空间态势感知重点实验室开放基金课题(KLCS20240106);成都市区域科技创新合作项目(2025-YF11-00031-HZ,2025-YF11-00050-HZ,2023-YF11-00020-HZ);大学生创新创业训练计划项目(202410621195,202410621183)。

摘　　要：随着自动驾驶、智能导航等领域的快速发展,对时空轨迹预测的准确性和鲁棒性的要求不断提高。传统轨迹预测方法主要依赖运动历史数据,忽略了环境中的语义信息,在复杂场景下往往难以取得理想的预测效果。对轨迹预测领域相关研究进行综述,特别是基于空间语义分析的轨迹预测研究进展。重点探讨了视觉语言模型(Vision Language Model,VLM)和大语言模型(Large Language Model,LLM)在轨迹预测方面的应用,介绍了多种基于空间语义分析的轨迹预测模型。通过实验结果分析发现,VLM和LLM能够显著提升轨迹预测的准确率。基于空间语义分析的轨迹预测方法未来将考虑多模态融合、提升模型架构、提高推理速度等方向,以进一步提升大规模轨迹预测的性能。With the rapid development of autonomous driving,intelligent navigation and other fields,the requirements of accuracy and robustness of spatio-temporal trajectory prediction are continuously increasing.Traditional trajectory prediction methods mainly rely on motion history data,thereby ignoring the semantic information in the context,which often makes it difficult to achieve ideal prediction results in complex scenarios.The relevant research in the field of trajectory prediction is reviewed,especially the research progress of trajectory prediction techniques based on spatial semantic analysis.The application of Visual Language Model(VLM)and Large Language Model(LLM)in trajectory prediction is mainly discussed,and a variety of trajectory prediction models based on spatial semantic analysis are introduced.Experimental results show that the VLM and LLM can significantly improve the accuracy of trajectory prediction.In the future,the trajectory prediction methods based on spatial semantic analysis will take into consideration the directions of multi-modality fusion,model architecture improvement and increasing reasoning speed to further improve the performance of large-scale trajectory prediction.

关 键 词：自动驾驶 轨迹预测 空间语义 视觉语言模型 大语言模型 多模态 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]