Pith. sign in

REVIEW 4 major objections 5 minor 47 references

Learning Generalized and Flexible Trajectory Models from Omni-Semantic Supervision

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A trajectory's raw path, turns, roads, and regions can share one embedding space, making retrieval a flexible vector lookup.

desk verdict Genuinely new task formalization and a solid, reproducible system, but the 'generalized' label oversells it—road/region encoders are closed-vocabulary and the transfer result is topology-only. read the letter →

arxiv 2505.17437 v1 pith:LGNV2UNK submitted 2025-05-23 cs.CV cs.AI

Yuanshao Zhu , James Jianqiao Yu , Xiangyu Zhao , Xiao Han , Qidong Liu , Xuetao Wei , Yuxuan Liang This is my paper · ORCID
classification cs.CVcs.AI
keywords trajectoryretrievalmultimodallearningcontrastiveembeddingalignmentsimilaritycondition-basedqueryspatio-temporaldataminingurbanmobility
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

OmniTraj sets out to replace rigid whole-trajectory similarity search—methods like DTW that align every point pair—with a retrieval system that treats a trip as four complementary views: the raw GPS sequence, a compressed topology of key turning points, the road segments traversed, and the urban regions crossed. The paper argues that if these four views are each given a dedicated encoder and then aligned into one shared embedding space by contrastive learning, any one view or any combination of views can be used as a query, and matching trajectories are found by nearest-neighbour lookup in linear time. That matters because real queries are often conditional ('which delivery routes pass through this road segment and this district?'), which similarity measures cannot express. The central claim, supported by experiments on two large urban datasets, is that this single unified model beats specialised trajectory-similarity baselines and CLIP-style multimodal aligners at both whole-trajectory and condition-based retrieval, while remaining efficient at million-trajectory scale.

What carries the argument

The machinery is a four-encoder architecture: a patch-based transformer (borrowing the Vision Transformer design) for raw trajectories; a rotary-position-embedding (RoPE) transformer over key topological points; a lookup-table embedding plus RoPE transformer over road segment identifiers; and a position-free transformer over region identifiers. The four encoders output embeddings projected into a shared space of dimension 512, aligned by a bidirectional InfoNCE contrastive loss with temperature scaling. The load-bearing property is that this shared space makes heterogeneous inputs—whole trips, single road IDs, region IDs—mutually comparable as vectors, so retrieval (including multi-condition queries formed by concatenating modality embeddings) reduces to cosine-similarity ranking over precomputed candidates.

What would settle it

Withhold a set of road segment IDs and region IDs from training, then run condition-based queries that mention exactly those held-out IDs against the test set; if the system cannot embed the query (or returns chance-level coverage), the claim that retrieval generalizes across the full space of road and region conditions is falsified while the topology-only transfer results would still stand.

Watch

Extended reading notes

Core claim

The paper's central discovery is that modality decoupling plus shared-space alignment is sufficient for both trajectory similarity and condition-based retrieval. Each modality encoder produces a fixed-dimensional embedding; a two-layer projector maps them into a common space; and a bidirectional InfoNCE loss pulls embeddings of the same trajectory together across modalities while pushing distinct trajectories apart. Once trained, queries expressed in any modality—a raw trajectory, a set of topology points, a road segment ID, a region ID, or a concatenation of several—are embedded into the same space, and the database is searched by cosine similarity over precomputed embeddings. The paper reports that the full OmniTraj model achieves the best mean reciprocal rank and hit rate on trajectory similarity across both datasets, and the highest condition-coverage rates for road- and region-based queries, and that these results hold at a per-query cost independent of database size.

Load-bearing premise

The load-bearing assumption is that the fixed road segment and region identifiers used during training are the only ones ever queried; an unseen road or region cannot be embedded by the lookup tables, so the claimed generality for condition-based queries breaks for any identifier outside the training vocabulary.

Editorial extensions

If this is right

  • Point-wise alignment cost disappears: whole-trajectory similarity search runs in $O(|\mathcal{D}|)$ time per query against a candidate set $\mathcal{D}$, instead of the quadratic cost of methods like DTW, because all candidate embeddings are precomputed.
  • Condition-based queries such as 'passes through road $r$ and region $R$' are expressible directly in the embedding space, without any geometric matching step at query time.
  • Coarse road- or region-based filters can shrink a million-trajectory database to a few hundred candidates while preserving downstream fine-grained topology retrieval accuracy, enabling a cheap two-stage pipeline.
  • The learned embeddings double as semantic condition signals for downstream generative models, demonstrated by condition-guided trajectory generation that visually matches the requested topology, road, and region constraints.
  • The topology modality transfers between cities with no retraining and closes much of the gap with light fine-tuning, suggesting that geometric trajectory structure is the city-agnostic core of the representation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The fixed-vocabulary lookup tables for road and region IDs put a boundary on the advertised generality: a query that names a road or region never seen in training cannot be embedded at all, and the paper's cross-city transfer experiments exercise only the topology modality, leaving road- and region-encoder transfer untested.
  • A testable extension would replace identifier lookup with geometry-aware encoders (e.g., embedding from segment coordinates or region boundaries), which would extend the observed zero-shot transfer property to road- and region-based queries.
  • The recipe—decouple modalities, align with contrastive loss, query by vector similarity—is generic: applied to other spatio-temporal objects such as vessel tracks or wildlife GPS paths with their own landmark vocabularies, the same design should yield comparable condition-based retrieval behaviour.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes OmniTraj, a trajectory retrieval framework that learns four modality-specific encoders—raw trajectories, topology, road segments, and regions—and aligns their embeddings in a shared space using a bidirectional InfoNCE loss. Retrieval is then performed by embedding a query (from any modality or combination) and ranking database trajectories by cosine similarity. The authors evaluate on Chengdu and Xi'an datasets, comparing against heuristic distance measures (DTW, EDR, Hausdorff, Fréchet) and learned trajectory embedding methods (t2vec, TrajCL, etc.) for whole-trajectory similarity, and against simple embedding/CLIP-style baselines for condition-based road/region retrieval. They also report scalability, transfer learning, and downstream trajectory-generation experiments.

Significance. If the results hold, OmniTraj would be a useful step toward flexible, condition-based trajectory retrieval, a task that is indeed under-supported by existing methods. The decoupled encoder design and the use of contrastive alignment across trajectory-derived semantics are reasonable and clearly described. The release of implementation code is a concrete strength, and the experiments cover two sizable real-world datasets. However, the central 'generalized' claim is only demonstrated for the topology modality; the road and region encoders are closed-vocabulary lookup tables, so the headline capability of condition-based retrieval over unseen roads/regions is not tested. The empirical margins over strong baselines also lack uncertainty quantification, which matters because several reported differences are small.

major comments (4)
  1. [§3.2.3–3.2.4, Eq. (6), §4.5] The road encoder (Eq. 6) and the region encoder (§3.2.4) are learned lookup tables over fixed identifier vocabularies (7,597 Chengdu road IDs, 6,018 Xi'an road IDs, and 256 grid regions per city, per Appendix B.1). Any road or region absent from training has no embedding, so it cannot be used as a query condition or encoded as part of a trajectory. The 'random replacement' augmentation in §3.2.3 operates within the same closed vocabulary and does not create open-set generalization. The transfer experiment in §4.5 explicitly uses only the topology modality, as the text states that the model 'encodes topological modalities without adding any additional city-related information.' Consequently, the paper's title-level and abstract-level claim of a 'generalized' omni-semantic framework is supported only for topology; the distinctive road/region condition-based capability remains untested for any road or region not seen during training. This is a load-bearing gap between the architecture and a core stated capability and should be addressed by either providing an open-vocabulary mechanism for road/region encoders or substantially qualifying the generalization claims.
  2. [Tables 2 and 3] Tables 2 and 3 report point estimates without standard deviations, confidence intervals, or significance tests. For example, the Chengdu MRR gain over TrajCL is 0.909 vs. 0.846, and the Chengdu road CR@1 gain over CLIP is 0.989 vs. 0.969; on Xi'an, the road CR@1 gap is 0.987 vs. 0.935. Some comparisons are much closer (e.g., Chengdu HR@10 0.989 vs. 0.974), and no information is given about run-to-run variation. Since several headline improvements are small and the baselines are evaluated under the same protocol, the authors should report multiple random seeds with standard deviations and, where appropriate, paired significance tests. Without this, the claim of consistent superiority over TrajCL and CLIP is not yet established.
  3. [Table 2 and Appendix B.2] There is a direct inconsistency about what the headlined 'OmniTraj' model in Table 2 actually encodes. Appendix B.2 states that 'OmniTraj: The OmniTraj framework utilizes only topology modality for optimal trajectory retrieval,' yet Table 2 presents the row 'OmniTraj' as a distinct and best-performing model, exceeding the explicit modality-combination variants such as OmniTraj (reg+road+top) (Chengdu MRR 0.909 vs. 0.843). Section 4.2's narrative attributes the gains to multi-modality alignment. The reader cannot tell whether the best row is a topology-only model, a full fusion model, or something else, and this ambiguity undermines the interpretation of the main retrieval comparison and the claim that fusing all modalities yields the best performance.
  4. [§4.3, Appendix B.2] The condition-based retrieval evaluation in Table 3 compares OmniTraj only against simple, self-constructed baselines (Embedding, Linear, CLIP variants). These baselines are reasonable first references, but they are not state-of-the-art condition-based trajectory retrieval systems, and the paper itself notes that 'no existing work directly addresses this task.' The strong claim in §4.3 that OmniTraj 'significantly outperforms' baselines is therefore limited by the absence of stronger recent alternatives, such as road-network-aware retrieval methods or trajectory foundation models that support partial or condition-based queries. Adding at least one such comparison, or clearly positioning the result as a first baseline for a new task, would make the evaluation more persuasive.
minor comments (5)
  1. [Eq. (6)] The notation for the road embedding matrix is inconsistent: the text says W_road ∈ R^{r×d} and then '|r| denotes the total number of unique road segments,' but |r| is not defined as a set size. Use |R| or R_road for the vocabulary size.
  2. [§4.5] The word 'Consequently' is misspelled as 'onsequently' in the last paragraph of Section 4.5.
  3. [Table 5] The column header '#Trajetory' in Table 5 contains a typo; it should read '#Trajectory'.
  4. [§3.4] The time complexity statement in §3.4 says self-attention is O(N^2 × d), but Section 3.2.1 claims the patch-based design reduces complexity to O((L/P)^2). Clarify what N denotes in each place so the two statements do not appear contradictory.
  5. [§4.1.3] It is stated that trajectory-similarity baselines 'were applied using topology modality,' but t2vec and TrajCL are commonly designed for raw GPS point sequences. Please clarify what input exactly was provided to each baseline and whether this choice gives any advantage to OmniTraj's topology encoder.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: held-out retrieval benchmarks ground the claims; the fixed-vocabulary road/region encoders create a generalization limitation, not a circular derivation.

full rationale

OmniTraj is an empirical contrastive-learning system rather than a derivation from first principles, and no step reduces its reported results to its own inputs. The four modality encoders produce embeddings (Eqs. 3, 5, 7, 8), the InfoNCE objective (Eq. 9) aligns them, and retrieval performance is then measured against held-out query/trajectory pairs using external metrics (MR, MRR, HR@k, CR@k) in Tables 2 and 3. No parameter is fitted to the test retrieval labels and then renamed as a prediction, and the similarity/condition-based retrieval results are not implied by the training loss by construction. The fixed-vocabulary lookup tables for road segments (Eq. 6) and regions (Section 3.2.4) are a genuine limitation: roads or regions absent from training cannot be encoded, and Section 4.5's transfer experiment exercises only the topology modality. However, that is a scope/generalization gap, not a circular argument, because the reported in-city retrieval numbers are still evaluated on held-out data. Self-citations such as ControlTraj [43] and UniTraj [44] appear only in related work or downstream generation and are not load-bearing for the central retrieval claim. No uniqueness theorem, ansatz, or known result is smuggled in via self-citation. The paper is self-contained against external benchmarks, so the appropriate circularity score is 0.

Assumptions & free parameters 10 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard contrastive-learning machinery, coarse spatial discretization of regions, and a fixed road vocabulary. No new theoretical entities are introduced; the main burden is empirical validation.

free parameters (10)
  • embedding dimension = 256
    Output dimension of all encoders; chosen in Table 4, not justified by data.
  • projection dimension = 512
    Projection head output dimension in Table 4; hand-picked.
  • number of transformer blocks = 6
    Encoder depth in Table 4; authors report performance plateaus beyond 6, but 6 is a manual choice.
  • attention heads = 8
    Number of heads per transformer block in Table 4.
  • InfoNCE temperature tau
    Temperature in Eq. 9 is not reported, so its value is a hidden manual choice.
  • fixed trajectory length L and patch size P = L=200 through interpolation; P not explicitly stated
    Section 3.2.1 resamples all trajectories to fixed length 200; patch size directly controls token count and is not reported.
  • max sequence lengths = 200 / 128 / 128 / 64
    Per-encoder maximum sequence lengths in Table 4; manually chosen.
  • region grid size = 16x16 = 256 cells
    Section B.1 divides each city into 256 grid cells; this coarse quantization defines the entire region modality.
  • pooling strategy = cls / bos / cls / cls
    Different pooling strategies per encoder in Table 4; chosen by hand.
  • learning rate = 2e-4
    Adam optimizer initial learning rate reported in Appendix B.
assumptions (5)
  • standard math InfoNCE contrastive loss aligns representations of different modalities of the same trajectory
    Used in Eqs. 9 and 10; a standard self-supervised objective, but its sufficiency for cross-modal retrieval is assumed rather than proven.
  • domain assumption Fixed-length resampling with cubic spline interpolation preserves retrieval-relevant trajectory information
    Section 3.2.1 normalizes and resamples all trajectories to length L=200, assuming no loss of discriminative signal.
  • domain assumption Regions are adequately represented by 16x16 grid identifiers
    Section B.1 divides each city into 256 grid cells, a coarse quantization of spatial semantics.
  • domain assumption Road segments are map-matched and their identifiers form a fixed vocabulary
    Section 3.2.3 uses a learnable lookup table over unique road IDs; assumes map matching is available and the vocabulary covers all queries.
  • standard math Transformer self-attention can capture local and long-range dependencies in trajectory patches
    Standard architecture assumption; not proven for trajectory data beyond empirical results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Generalized and Flexible Trajectory Models from Omni-Semantic Supervision." pith.science (2026). https://pith.science/paper/LGNV2UNK

@misc{pith2026250517437,
  author       = {Pith},
  title        = {Pith review of: Learning Generalized and Flexible Trajectory Models from Omni-Semantic Supervision},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LGNV2UNK}},
  note         = {Machine review of arXiv:2505.17437}
}
read the original abstract

The widespread adoption of mobile devices and data collection technologies has led to an exponential increase in trajectory data, presenting significant challenges in spatio-temporal data mining, particularly for efficient and accurate trajectory retrieval. However, existing methods for trajectory retrieval face notable limitations, including inefficiencies in large-scale data, lack of support for condition-based queries, and reliance on trajectory similarity measures. To address the above challenges, we propose OmniTraj, a generalized and flexible omni-semantic trajectory retrieval framework that integrates four complementary modalities or semantics -- raw trajectories, topology, road segments, and regions -- into a unified system. Unlike traditional approaches that are limited to computing and processing trajectories as a single modality, OmniTraj designs dedicated encoders for each modality, which are embedded and fused into a shared representation space. This design enables OmniTraj to support accurate and flexible queries based on any individual modality or combination thereof, overcoming the rigidity of traditional similarity-based methods. Extensive experiments on two real-world datasets demonstrate the effectiveness of OmniTraj in handling large-scale data, providing flexible, multi-modality queries, and supporting downstream tasks and applications.

Figures

Figures reproduced from arXiv: 2505.17437 by the authors.

Figure 1
Figure 1. Comparison of trajectory retrieval methods. Our [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the proposed OmniTraj framework, which contains four individual modal encoders. Each encoder [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Model structure and parameters and setting. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Use OmniTraj as a semantic representation to guide [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: The main structure of each encoder in the OmniTraj framework, the trajectory encoder is shown in the main content. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 8
Figure 8. Figure 8: Coverage changes with the length of the retrieved [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 27 canonical work pages

  1. [1]

    Helmut Alt. 2009. The computational geometry of comparing shapes.Efficient Algorithms: Essays Dedicated to Kurt Mehlhorn on the Occasion of His 60th Birthday (2009), 235–248

    work page 2009

  2. [2]

    Helmut Alt and Michael Godau. 1995. Computing the Fréchet distance between two polygonal curves.International Journal of Computational Geometry & Appli- cations5, 01n02 (1995), 75–91

    1995

  3. [3]

    Hanlin Cao, Haina Tang, Yulei Wu, Fei Wang, and Yongjun Xu. 2021. On accurate computation of trajectory similarity via single image super-resolution. In2021 International Joint Conference on Neural Networks (IJCNN). IEEE, 1–9

    2021

  4. [4]

    Yanchuan Chang, Jianzhong Qi, Yuxuan Liang, and Egemen Tanin. 2023. Con- trastive trajectory similarity learning with dual-feature attention. In2023 IEEE 39th International conference on data engineering (ICDE). IEEE, 2933–2945

    2023

  5. [5]

    Yanchuan Chang, Egemen Tanin, Gao Cong, Christian S Jensen, and Jianzhong Qi

    work page

  6. [6]

    Lei Chen, M Tamer Özsu, and Vincent Oria. 2005. Robust and fast similarity search for moving object trajectories. InProceedings of the 2005 ACM SIGMOD international conference on Management of data. 491–502

    2005

  7. [7]

    Lisi Chen, Shuo Shang, Christian S Jensen, Bin Yao, and Panos Kalnis. 2020. Paral- lel semantic trajectory similarity join. In2020 IEEE 36th International Conference on Data Engineering (ICDE). IEEE, 997–1008

    work page 2020

  8. [8]

    Wei Chen, Yuxuan Liang, Yuanshao Zhu, Yanchuan Chang, Kang Luo, Haomin Wen, Lei Li, Yanwei Yu, Qingsong Wen, Chao Chen, et al. 2024. Deep Learning for Trajectory Data Management and Mining: A Survey and Beyond.arXiv preprint arXiv:2403.14151(2024)

    arXiv 2024

Show all 47 references
  1. [9]

    Jian Dai, Bin Yang, Chenjuan Guo, and Zhiming Ding. 2015. Personalized route recommendation using big trajectory data. In2015 IEEE 31st international confer- ence on data engineering. IEEE, 543–554

    2015

  2. [10]

    Rodrigo Augusto de Oliveira e Silva, Ge Cui, Seyyed Mohammadreza Rahimi, and Xin Wang. 2022. Personalized route recommendation through historical travel behavior analysis.GeoInformatica26, 3 (2022), 505–540

    2022

  3. [11]

    Jinliang Deng, Xiusi Chen, Renhe Jiang, Xuan Song, and Ivor W Tsang. 2022. A multi-view multi-task learning framework for multi-variate time series forecast- ing.IEEE Transactions on Knowledge and Data Engineering35, 8 (2022)

    2022

  4. [12]

    Jinliang Deng, Feiyang Ye, Du Yin, Xuan Song, Ivor Tsang, and Hui Xiong. 2024. Parsimony or capability? decomposition delivers both in long-term time series forecasting.Advances in Neural Information Processing Systems37 (2024)

    2024

  5. [13]

    Ke Deng, Kexin Xie, Kevin Zheng, and Xiaofang Zhou. 2011. Trajectory indexing and retrieval.Computing with spatial trajectories(2011), 35–60

    2011

  6. [14]

    Alexey Dosovitskiy. 2020. An image is worth 16x16 words: Transformers for image recognition at scale.arXiv preprint arXiv:2010.11929(2020)

    2020 arXiv

  7. [15]

    Ziquan Fang, Yuntao Du, Lu Chen, Yujia Hu, Yunjun Gao, and Gang Chen. 2021. E2dtc: An end to end deep trajectory clustering framework via self-training. In 2021 IEEE 37th International Conference on Data Engineering. IEEE, 696–707

    2021

  8. [16]

    Ziquan Fang, Yuntao Du, Xinjun Zhu, Danlei Hu, Lu Chen, Yunjun Gao, and Christian S Jensen. 2022. Spatio-temporal trajectory similarity learning in road networks. InProceedings of the 28th ACM SIGKDD conference on knowledge dis- covery and data mining. 347–356

    2022

  9. [17]

    Chongming Gao, Zhong Zhang, Chen Huang, Hongzhi Yin, Qinli Yang, and Jun- ming Shao. 2020. Semantic trajectory representation and retrieval via hierarchical embedding.Information Sciences538 (2020), 176–192

    2020

  10. [18]

    Chenjuan Guo, Bin Yang, Jilin Hu, Christian S Jensen, and Lu Chen. 2020. Context- aware, preference-based vehicle routing.The VLDB Journal29 (2020), 1149–1170

    2020

  11. [19]

    Peng Han, Jin Wang, Di Yao, Shuo Shang, and Xiangliang Zhang. 2021. A graph- based approach for trajectory similarity computation in spatial networks. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. 556–564

    2021

  12. [20]

    Xiucheng Li, Kaiqi Zhao, Gao Cong, Christian S Jensen, and Wei Wei. 2018. Deep representation learning for trajectory similarity computation. In2018 IEEE 34th international conference on data engineering (ICDE). IEEE, 617–628

    2018

  13. [21]

    Yan Lin, Huaiyu Wan, Shengnan Guo, Jilin Hu, Christian S Jensen, and Youfang Lin. 2023. Pre-training general trajectory embeddings with maximum multi-view entropy coding.IEEE Transactions on Knowledge and Data Engineering36, 12 (2023), 9037–9050

    2023

  14. [22]

    Yan Lin, Tonglong Wei, Zeyu Zhou, Haomin Wen, Jilin Hu, Shengnan Guo, Youfang Lin, and Huaiyu Wan. 2024. TrajFM: A Vehicle Trajectory Foundation Model for Region and Task Transferability.arXiv:2408.15251(2024)

    2024 arXiv

  15. [23]

    Yan Lin, Zeyu Zhou, Yicheng Liu, Haochen Lv, Haomin Wen, Tianyi Li, Yushuai Li, Christian S Jensen, Shengnan Guo, Youfang Lin, et al. 2024. UniTE: A Survey and Unified Pipeline for Pre-training ST Trajectory Embeddings.arXiv e-prints (2024), arXiv–2407

    2024

  16. [24]

    Huaishao Luo, Lei Ji, Ming Zhong, Yang Chen, Wen Lei, Nan Duan, and Tianrui Li. 2022. Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning.Neurocomputing508 (2022), 293–304

    2022

  17. [25]

    Meinard Müller. 2007. Dynamic time warping.Information retrieval for music and motion(2007), 69–84

    2007

  18. [26]

    Aaron van den Oord, Yazhe Li, and Oriol Vinyals. 2018. Representation learning with contrastive predictive coding.arXiv preprint arXiv:1807.03748(2018)

    2018 arXiv

  19. [27]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. InInternational conference on machine learnin...

    2021

  20. [28]

    Han Su, Shuncheng Liu, Bolong Zheng, Xiaofang Zhou, and Kai Zheng. 2020. A survey of trajectory distance measures and performance evaluation.The VLDB Journal29 (2020), 3–32

    2020

  21. [29]

    Han Su, Kai Zheng, Haozhou Wang, Jiamin Huang, and Xiaofang Zhou. 2013. Calibrating trajectory data for similarity-based analysis. InProceedings of the 2013 ACM SIGMOD international conference on management of data. 833–844

    2013

  22. [30]

    Jianlin Su, Murtadha Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. 2024. Roformer: Enhanced transformer with rotary position embedding. Neurocomputing568 (2024), 127063

    2024

  23. [31]

    David Alexander Tedjopurnomo, Xiucheng Li, Zhifeng Bao, Gao Cong, Farhana Choudhury, and A Kai Qin. 2021. Similar trajectory search with spatio-temporal deep representation learning.ACM Transactions on Intelligent Systems and Tech- nology (TIST)12, 6 (2021), 1–26

    2021

  24. [32]

    Sheng Wang, Zhifeng Bao, J Shane Culpepper, and Gao Cong. 2021. A survey on trajectory data management, analytics, and learning.ACM Computing Surveys (CSUR)54, 2 (2021), 1–36

    2021

  25. [33]

    Yong Wang, Kaiyu Li, Guoliang Li, and Nan Tang. 2022. Road-Aware Indexing for Trajectory Range Queries.IEEE Transactions on Knowledge and Data Engineering 35, 8 (2022), 8476–8489

    2022

  26. [34]

    Peng Xu, Xiatian Zhu, and David A Clifton. 2023. Multimodal learning with transformers: A survey.IEEE Transactions on Pattern Analysis and Machine Intelligence45, 10 (2023), 12113–12132

    2023

  27. [35]

    Ronghui Xu, Hanyin Cheng, Chenjuan Guo, Hongfan Gao, Jilin Hu, Sean Bin Yang, and Bin Yang. 2024. MM-Path: Multi-modal, Multi-granularity Path Repre- sentation Learning–Extended Version.arXiv preprint arXiv:2411.18428(2024)

    2024 arXiv

  28. [36]

    Di Yao, Gao Cong, Chao Zhang, and Jingping Bi. 2019. Computing trajectory similarity in linear time: A generic seed-guided neural metric learning approach. In2019 IEEE 35th international conference on data engineering (ICDE). IEEE, 1358– 1369

    2019

  29. [37]

    Di Yao, Haonan Hu, Lun Du, Gao Cong, Shi Han, and Jingping Bi. 2022. TrajGAT: A graph-based long-term dependency modeling approach for trajectory similarity computation. InProceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining. 2275–2285

    2022

  30. [38]

    Yuan Yuan, Zhaojian Li, and Bin Zhao. 2025. A Survey of Multimodal Learning: Methods, Applications, and Future.Comput. Surveys(2025)

    2025

  31. [39]

    Bolong Zheng, Nicholas Jing Yuan, Kai Zheng, Xing Xie, Shazia Sadiq, and Xiao- fang Zhou. 2015. Approximate keyword search in semantic trajectory database. In2015 IEEE 31st International conference on data engineering. IEEE, 975–986

    2015

  32. [40]

    Kai Zheng, Yan Zhao, Defu Lian, Bolong Zheng, Guanfeng Liu, and Xiaofang Zhou

  33. [41]

    Silin Zhou, Jing Li, Hao Wang, Shuo Shang, and Peng Han. 2023. GRLSTM: trajectory similarity computation with graph-based residual LSTM. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 37. 4972–4980

    2023

  34. [42]

    Silin Zhou, Shuo Shang, Lisi Chen, Peng Han, and Christian S Jensen. 2024. Grid and Road Expressions Are Complementary for Trajectory Representation Learning.arXiv preprint arXiv:2411.14768(2024)

    2024 arXiv

  35. [43]

    Yuanshao Zhu, James Jianqiao Yu, Xiangyu Zhao, Qidong Liu, Yongchao Ye, Wei Chen, Zijian Zhang, Xuetao Wei, and Yuxuan Liang. 2024. Controltraj: Controllable trajectory generation with topology-constrained diffusion model. InProceedings of the 30th ACM SIGKDD Conference on Kno...

    2024

  36. [44]

    Yuanshao Zhu, James Jianqiao Yu, Xiangyu Zhao, Xuetao Wei, and Yuxuan Liang

  37. [47]

    Avg. ”, “Avg. Road

    Unitraj: Learning a universal trajectory foundation model from billion-scale worldwide traces.CoRR(2024). ACM, SIGKDD, 2025 Yuanshao Zhu et al. A Details of Framework Structure of Encoders.As outlined in Section 3 and depicted in Figure 6, OmniTraj adopts a modular, decoupled ...

    2024

  38. [2019]

    Reference-based framework for spatio-temporal trajectory compression and query processing.IEEE Transactions on Knowledge and Data Engineering32, 11 (2019), 2227–2240

    2019

  39. [2024]

    Trajectory Similarity Measurement: An Efficiency Perspective.Proceedings of the VLDB Endowment17, 9 (2024), 2293–2306

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.