基于状态空间模型的高效语义感知轨迹表示学习方法

doi:10.11896/jsjkx.250600130

摘要/Abstract

摘要： 车辆轨迹为各类交通服务应用提供了关键的运动信息。为了更好地利用车辆轨迹,有必要开发轨迹表示学习方法来准确且高效地提取包括运动行为和出行目的在内的出行语义,以支持精确的下游应用。然而,这一任务面临两大挑战:1)运动行为本质上是时空连续的,难以从离散轨迹点中有效提取;2)出行目的与车辆经过的区域和路段的功能相关,但这些功能无法从原始时空轨迹特征中直接获得,也难以从相关的复杂文本特征中提取。为了解决这些挑战,提出了一种高效语义感知轨迹表示学习方法ESTRL。首先,引入了基于Mamba的轨迹编码器,使用高阶移动特征参数化轨迹状态空间模型,有效且高效地建模车辆的连续运动行为。其次,提出了出行目的感知预训练机制,通过对比学习将出行目的融入学习到的轨迹嵌入中,从而无需在嵌入计算过程中引入额外开销。在真实数据集上的大量实验表明,所提方法在效率和准确性方面均优于当前先进的基线模型。

关键词: 车辆轨迹, 轨迹表示学习, 状态空间模型, 对比学习, 自监督预训练

Abstract: Vehicle trajectories provide crucial movement information for various traffic service applications.To better utilize vehicle trajectories,it is essential to develop trajectory representation learning methods that can effectively and efficiently extract tra-vel semantics,including movement behaviors and travel purposes,to support accurate downstream applications.However,this task presents two major challenges:1) movement behaviors are inherently spatio-temporally continuous,making them difficult to extract effectively from discrete trajectory points;2) travel purposes are related to the functionalities of areas and road segments traversed by vehicles,but these functionalities cannot be directly obtained from the raw spatio-temporal trajectory features,nor can they be extracted from the relevant complex textual features.To address these challenges,this paper proposes an efficient semantic-aware trajectory representation learning method called ESTRL.Firstly,a Mamba-based trajectory encoder is introduced.It uses high-order movement features to parameterize the trajectory state space model(Traj-SSM),which effectively and efficiently models continuous movement behaviors of vehicles.Secondly,a travel purpose-aware pre-training procedure is proposed.It integrates travel purposes into the learned trajectory embeddings through contrastive learning without introducing extra overhead to embedding calculation process.Extensive experiments on real-world datasets demonstrate that the proposed method outperforms state-of-the-art baseline models in both efficiency and accuracy.

Key words: Vehicle trajectories, Trajectory representation learning, State space model, Contrastive learning, Self-supervised pre-training

中图分类号:

TP311

刘奕辰, 林彦, 周泽宇, 郭晟楠, 林友芳, 万怀宇. 基于状态空间模型的高效语义感知轨迹表示学习方法[J]. 计算机科学, 2026, 53(4): 134-142. https://doi.org/10.11896/jsjkx.250600130

LIU Yichen, LIN Yan, ZHOU Zeyu, GUO Shengnan, LIN Youfang, WAN Huaiyu. Efficient Semantic-aware Trajectory Representation Learning Method via State Space Model[J]. Computer Science, 2026, 53(4): 134-142. https://doi.org/10.11896/jsjkx.250600130

参考文献

[1]WU H,CHEN Z,SUN W,et al.Modeling trajectories with recurrent neural networks[C]//Proceedings of the Twenty-sixth International Joint Conference on Artificial Intelligence.2017:3083-3090.
[2]YAN B,ZHAO G,SONG L,et al.PreCLN:pretrained-basedcontrastive learning network for vehicle trajectory prediction[J].World Wide Web,2023,26(4):1853-1875.
[3]LIN Y,WAN H,HU J,et al.Origin-destination travel time oracle for map-based services[J].Proceedings of the ACM on Ma-nagement of Data,2023,1(3):1-27.
[4]YUAN H,LI G,BAO Z,et al.Effective traveltime estimation:when historical trajectories over road networks matter[C]//Proceedings of the 2020 International Conference on Management of Data.ACM,2020:2135-2149.
[5]HAN X,CHENG R,MA C,et al.DeepTEA:effective and efficient online time-dependent trajectory outlier detection[J].Proceedings of the VLDB Endowment,2022,15(7):1493-1505.
[6]LIU Y,ZHAO K,CONG G,et al.Online anomalous trajectory detection with deep generative sequence modeling[C]//Procee-dings of the 36th IEEE International Conference on Data Engineering.IEEE,2020:949-960.
[7]HU D,CHEN L,FANG H,et al.Spatio-temporal trajectorysimilarity measures:a comprehensive survey and quantitative study[J].IEEE Transactions on Knowledge and Data Enginee-ring,2024,36(5):2191-2212.
[8]YAO D,HU H,DU L,et al.TrajGAT:a graph-based long-term dependency modeling approach for trajectory similarity computation[C]//Kdd ’22:the 28th ACM Sigkdd Conference on Knowledge Discovery and Data Mining.ACM,2022:2275-2285.
[9]YAO D,ZHANG C,ZHU Z,et al.Trajectory clustering via deep representation learning[C]//2017 International Joint Confe-rence on Neural Networks.IEEE,2017:3880-3887.
[10]CHUNG J,GÜLÇEHRE Ç,CHO K,et al.Empirical evaluation of gated recurrent neural networks on sequence modeling[J].arXiv:1412.3555,2014.
[11]HOCHREITER S,SCHMIDHUBER J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[12]VASWANI A,SHAZEER N,PARMAR N,et al.Attention isall you need[C]//Proceedings of the 31st InternationalConfe-rence on Neural Information Processing Systems.2017:5998-6008.
[13]LIANG Y,OUYANG K,WANG Y,et al.TrajFormer:efficient trajectory classification with transformers[C]//Proceedings of the 31st ACM International Conference on Information & Knowledge Management.ACM,2022:1229-1237.
[14]LIANG Y,OUYANG K,YAN H,et al.Modeling trajectorieswith neural ordinary differential equations[C]//Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence.ijcai.org,2021:1498-1504.
[15]LIN Y,WAN H,GUO S,et al.Pre-training general trajectoryembeddings with maximum multi-view entropy coding[J].IEEE Transactions on Knowledge and Data Engineering,2024,36(12):9037-9050.
[16]CHEN T Q,RUBANOVA Y,BETTENCOURT J,et al.Neural ordinary differential equations[C]//Advances in Neural Information Processing Systems 31:Annual Conference on Neural Information Processing Systems 2018.2018:6572-6583.
[17]KIDGER P,MORRILL J,FOSTER J,et al.Neural controlleddifferential equations for irregular time series[C]//Advances in Neural Information Processing Systems 33:Annual Conference on Neural Information Processing Systems 2020.2020.
[18]BROWN T B,MANN B,RYDER N,et al.Language models are few-shot learners[C]//Advances in Neural Information Processing Systems 33:Annual Conference on Neural Information Processing Systems 2020.2020.
[19]DEVLIN J,CHANG M W,LEE K,et al.BERT:pre-training of deep bidirectional transformers for language understanding[C]//Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies.ACM,2019:4171-4186.
[20]DU Z,QIAN Y,LIU X,et al.GLM:general language model pre-training with autoregressive blank infilling[C]//Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics.ACM,2022:320-335.
[21]ZHOU Z,LIN Y,WEN H,et al.PLM4Traj:cognizing move-ment patterns and travel purposes from trajectories with pre-trained language models[J].arXiv:2405.12459,2024.
[22]FENG J,LI Y,ZHANG C,et al.DeepMove:predicting humanmobility with attentional recurrent networks[C]//Proceedings of the 2018 World Wide Web Conference on World Wide Web.ACM,2018:1459-1468.
[23]KONG D,WU F.HST-LSTM:a hierarchical spatial-temporallong-short term memory network for location prediction[C]//Proceedings of the Twenty-seventh International Joint Confe-rence on Artificial Intelligence.ijcai.org,2018:2341-2347.
[24]MIAO C,LUO Z,ZENG F,et al.Predicting human mobility via attentive convolutional network[C]//Wsdm ’20:the Thirteenth ACM International Conference on Web Search and Data Mi-ning.ACM,2020:438-446.
[25]CHEN W,LI S,HUANG C,et al.Mutual distillation learningnetwork for trajectory-user linking[C]//Proceedings of the Thirty-first International Joint Conference on Artificial Intelligence.ijcai.org,2022:1973-1979.
[26]SANG Y,XIE Z,CHEN W,et al.TULRN:trajectory user linking on road networks[J].World Wide Web,2023,26(4):1949-1965.
[27]YAO D,CONG G,ZHANG C,et al.Computing trajectory similarity in linear time:a generic seed-guided neural metric learning approach[C]//35th IEEE International Conference on Data Engineering.IEEE,2019:1358-1369.
[28]ZHOU S,HAN P,YAO D,et al.Spatial-temporal fusion graph framework for trajectory similarity computation[J].World Wide Web,2023,26(4):1501-1523.
[29]HINTON G E,SALAKHUTDINOV R R.Reducing the dimensionality of data with neural networks[J].Science,2006,313(5786):504-507.
[30]LI X,ZHAO K,CONG G,et al.Deep representation learning for trajectory similarity computation[C]//34th IEEE International Conference on Data Engineering.IEEE Computer Society,2018:617-628.
[31]FU T Y,LEE W C.Trembr:exploring road networks for trajectory representation learning[J].ACM Transactions on Intelligent Systems and Technology,2020,11(1):10:1-10:25.
[32]ZHOU F,DAI Y,GAO Q,et al.Self-supervised human mobility learning for next location prediction and trajectory classification[J].Knowledge-Based Systems,2021,228:107214.
[33]OORD A VAN DEN,LI Y,VINYALS O.Representation learning with contrastive predictive coding[J].arXiv:1807.03748,2018.
[34]JIANG J,PAN D,REN H,et al.Self-supervised trajectory representation learning with temporal regularities and travel semantics[C]//39th IEEE International Conference on Data Engineering.IEEE,2023:843-855.
[35]CHEN T,KORNBLITH S,NOROUZI M,et al.A simpleframework for contrastive learning of visual representations[C]//Proceedings of the 37th International Conference on Machine Learning.PMLR,2020:1597-1607.
[36]LIU X,WANG Z,LI Y,et al.Self-supervised learning via maximum entropy coding[C]//Advances in Neural Information Processing Systems 35:Annual Conference on Neural Information Processing Systems 2022.2022.
[37]DAO T,GU A.Transformers are SSMs:Generalized Models and Efficient Algorithms Through Structured State Space Duality[C]//Proceedings of the41th International Conference on Machine Learning.PMLR,2024:10041-10071.
[38]TANCIK M,SRINIVASAN P P,MILDENHALL B,et al.Fourier features let networks learn high frequency functions in low dimensional domains[C]//Advances in Neural Information Processing Systems 33:Annual Conference on Neural Information Processing Systems 2020.2020.
[39]GU A,DAO T.Mamba:linear-time sequence modeling with selective state spaces[J].arXiv:2312.00752,2023.
[40]MEERT W,VERBEKE M.HMM with non-emitting states for Map Matching[C]//European Conference on Data Analysis(ECDA).2018.
[41]RADFORD A,KIM J W,HALLACY C,et al.Learning transferable visual models from natural language supervision[C]//Proceedings of the 38th International Conference on Machine Learning.PMLR,2021:8748-8763.
[42]LIN Y,WAN H,GUO S,et al.Pre-training context and time aware location embeddings from spatial-temporal trajectories for user nextlocation prediction[C]//Proceedings of the AAAI Conference on Artificial Intelligence.AAAI,2021:4241-4248.
[43]CHEN Y,LI X,CONG G,et al.Robust road network representation learning:when traffic patterns meet traveling semantics[C]//Cikm’21:the 30th ACM International Conference on Information and Knowledge Management.ACM,2021:211-220.
[44]CHANG Y,QI J,LIANG Y,et al.Contrastive trajectory similarity learning with dual-feature attention[C]//2023 IEEE 39th International Conference on Data Engineering(ICDE).IEEE,2023:2933-2945.
[45]YANG S B,HU J,GUO C,et al.LightPath:lightweight andscalable path representation learning[C]//Proceedings of the 29th ACM Sigkdd Conference on Knowledge Discovery and Data Mining.ACM,2023:2999-3010.
[46]FANG Z,DU Y,ZHU X,et al.Spatio-temporal trajectory similarity learning in road networks[C]//Kdd ’22:the 28th ACM Sigkdd Conference on Knowledge Discovery and Data Mining.ACM,2022:347-356.
[47]PASZKE A,GROSS S,MASSA F,et al.PyTorch:an imperative style,high-performance deep learning library[C]//Advances in Neural Information Processing Systems 32:Annual Conference on Neural Information Processing Systems 2019.2019:8024-8035.

相关文章 15

[1]	程梓萌, 杨馨悦, 艾浩军, 王中元. 基于双重语义对比学习的无监督红外图像生成方法 Unsupervised Infrared Image Generation Method Based on Dual Semantic Contrastive Learning 计算机科学, 2026, 53(4): 260-268. https://doi.org/10.11896/jsjkx.250700172
[2]	于程程, 姜永发, 陈方疏, 王家辉, 孟宪凯. 融合多视角习题表征与遗忘机制的深度知识追踪 Multi-view Exercise Representation and Forgetting Mechanism for Deep KnowledgeTracing 计算机科学, 2026, 53(3): 107-114. https://doi.org/10.11896/jsjkx.250700092
[3]	李雯莉, 冯小年, 钱铁云. 基于大型语言模型增广的少样本持续毒性检测 Few-shot Continuous Toxicity Detection Based on Large Language Model Augmentation 计算机科学, 2026, 53(3): 321-330. https://doi.org/10.11896/jsjkx.250600010
[4]	宋建华, 何佳伟, 张龑. 基于对比学习的双通道源代码漏洞检测模型 Dual-channel Source Code Vulnerability Detection Model Based on Contrastive Learning 计算机科学, 2026, 53(3): 424-432. https://doi.org/10.11896/jsjkx.250200124
[5]	李春英, 汤志康, 庄郅玮, 李文博, 郭炎熙, 张晓薇. DCL-FKT:融合双重对比学习与遗忘机制的个性化知识追踪模型 DCL-FKT:Personalized Knowledge Tracing via Dual Contrastive Learning and ForgettingMechanism 计算机科学, 2026, 53(2): 99-106. https://doi.org/10.11896/jsjkx.250600002
[6]	王新喻, 宋小民, 郑慧明, 彭德中, 陈杰. 融合对比学习的掩码图自编码器 Contrastive Learning-based Masked Graph Autoencoder 计算机科学, 2026, 53(2): 145-151. https://doi.org/10.11896/jsjkx.250100155
[7]	魏金生, 周苏, 卢官明, 丁佳伟. 基于用户静动态兴趣和去噪隐式负反馈的新闻推荐算法 News Recommendation Algorithm Based on User Static and Dynamic Interests and DenoisedImplicit Negative Feedback 计算机科学, 2026, 53(2): 152-160. https://doi.org/10.11896/jsjkx.241200177
[8]	陈晓岚, 毛舜, 李伟生, 林荣华, 汤庸. 基于双层级对比学习的健壮知识追踪模型 Robust Knowledge Tracing Model Based on Two-level Contrastive Learning 计算机科学, 2026, 53(2): 31-38. https://doi.org/10.11896/jsjkx.250700196
[9]	杨明, 贺超波, 杨佳琦. 基于方向感知孪生网络的知识概念先序关系预测方法 Direction-aware Siamese Network for Knowledge Concept Prerequisite Relation Prediction 计算机科学, 2026, 53(2): 39-47. https://doi.org/10.11896/jsjkx.250600005
[10]	黄超, 程春玲, 王有康. 基于伪标签不确定性估计的无源域自适应方法 Source-free Domain Adaptation Method Based on Pseudo Label Uncertainty Estimation 计算机科学, 2025, 52(9): 212-219. https://doi.org/10.11896/jsjkx.240700159
[11]	张士举, 郭朝阳, 吴承亮, 吴凌俊, 杨丰玉. 基于关键语义驱动和对比学习的文本聚类方法 Text Clustering Approach Based on Key Semantic Driven and Contrastive Learning 计算机科学, 2025, 52(8): 171-179. https://doi.org/10.11896/jsjkx.240700008
[12]	丁政泽, 聂仁灿, 李锦涛, 苏华平, 徐航. MTFuse:基于Mamba和Transformer的红外与可见光图像融合网络 MTFuse:An Infrared and Visible Image Fusion Network Based on Mamba and Transformer 计算机科学, 2025, 52(8): 188-194. https://doi.org/10.11896/jsjkx.240600106
[13]	张涛涛, 谢钧, 乔平娟. 基于渐进式自训练开集域适应的辐射源个体识别 Specific Emitter Identification Based on Progressive Self-training Open Set Domain Adaptation 计算机科学, 2025, 52(7): 279-286. https://doi.org/10.11896/jsjkx.240600073
[14]	叶佳乐, 普园媛, 赵征鹏, 冯珏, 周联敏, 谷金晶. 混合对比学习和多视角CLIP的多模态图文情感分析 Multi-view CLIP and Hybrid Contrastive Learning for Multimodal Image-Text Sentiment Analysis 计算机科学, 2025, 52(6A): 240700060-7. https://doi.org/10.11896/jsjkx.240700060
[15]	付书凡, 王中卿, 姜晓彤. 融合情感词典和图对比学习的中文零样本立场检测 Zero-shot Stance Detection in Chinese by Fusion of Emotion Lexicon and Graph ContrastiveLearning 计算机科学, 2025, 52(6A): 240500051-7. https://doi.org/10.11896/jsjkx.240500051

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed