pith. sign in

arxiv: 2607.00989 · v1 · pith:U4URYFIVnew · submitted 2026-07-01 · 💻 cs.HC · cs.AI

SenseWalk: Agent-Based Semantic Trajectory Simulation Powered by Large Language Models in Zoned Environments

Ziyue Lin , Xinhang Xie , Kangyi Wang , Siming Chen This is my paper

Pith reviewed 2026-07-02 06:05 UTC · model grok-4.3

classification 💻 cs.HC cs.AI
keywords semantic trajectoriesLLM-powered agentssimulation workflowsocial force modelzoned environmentsinteractive systemhuman movement modeling
0
0 comments X

The pith

An interactive system called SenseWalk lets LLM agents generate semantic trajectories by blending language model decisions with physical movement rules.

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

Semantic trajectory analysis seeks to understand human movement not only through paths but also through the underlying profiles, goals, and behaviors that explain why people move as they do. Collecting such rich data in real settings is expensive and often incomplete, while existing simulation tools require significant technical skill that limits their use. SenseWalk addresses this by providing an interactive platform where large language model agents handle semantic choices in zoned environments, paired with the social force model to ensure movements follow plausible physical dynamics. The system includes a user interface for setting up scenarios and reviewing results. It was tested through quantitative experiments on the workflow and a user study involving twelve participants to check its practicality.

Core claim

SenseWalk is an interactive system that supports simulating semantic trajectories by LLM-powered agents. We develop a simulation workflow that combines LLMs and the social force model to balance physical plausibility and semantic coherence. A user-friendly interface is designed to facilitate users in customizing the simulation configuration and analyzing simulation outputs. We also conduct a quantitative experiment to evaluate the effectiveness of our simulation workflow, and a user study (n=12) to assess the usefulness and efficiency of our system.

What carries the argument

The simulation workflow that combines LLMs for semantic decision making with the social force model for physical movement, operating within zoned environments.

If this is right

  • Practitioners can simulate high-quality semantic trajectory data without costly real-world collection.
  • The system reduces the need for substantial technical expertise in trajectory simulation.
  • Users can customize and analyze simulations through an accessible interface.
  • The approach supports better understanding of movement patterns by incorporating semantic information.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If effective, this method could be adapted for training AI models on human behavior in various settings like retail or public spaces.
  • The zoned environment structure might allow integration with mapping data for real-world applications.
  • Scaling the agent interactions could reveal emergent behaviors in large crowds not captured by current models.

Load-bearing premise

Combining LLMs with the social force model will successfully balance physical plausibility and semantic coherence in the generated trajectories.

What would settle it

A direct comparison showing that trajectories from the system deviate significantly from real human movement data in either physical accuracy or semantic alignment.

Figures

Figures reproduced from arXiv: 2607.00989 by Kangyi Wang, Siming Chen, Xinhang Xie, Ziyue Lin.

Figure 1
Figure 1. Figure 1: The simulation configuration includes (a) Environment [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The simulation workflow. Input consists of three parts: [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The interface of SenseWalk. It consists of four views: (A) MapSidebar supports the environment map configuration; (B) [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of the mean of physical loss between the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of semantic loss between the baseline and [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Quantitative results of user study. Participants rated the system from (a) authoring, (b) analyzing, and (c) system perspectives. [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of baseline and simulated trajectories for [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
read the original abstract

Semantic trajectory analysis has recently emerged as an approach for modeling human movement by capturing implicit patterns and behaviors through semantic information (e.g., visitors' profiles and goals) beyond raw spatial paths to better understand why people move in certain ways. However, analyzing semantic trajectories in real-world scenarios remains challenging, as collecting high-quality data is costly and often lacks rich semantic information. Meanwhile, existing simulation tools require substantial technical expertise, which makes them difficult for practitioners to adopt. To address these limitations, the paper proposes ${SenseWalk}$, an interactive system that supports simulating semantic trajectories by LLM-powered agents. We develop a simulation workflow that combines LLMs and the social force model to balance physical plausibility and semantic coherence. A user-friendly interface is designed to facilitate users in customizing the simulation configuration and analyzing simulation outputs. We also conduct a quantitative experiment to evaluate the effectiveness of our simulation workflow, and a user study (n=12) to assess the usefulness and efficiency of our system.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes SenseWalk, an interactive system for simulating semantic trajectories in zoned environments using LLM-powered agents. The core contribution is a simulation workflow that integrates large language models for semantic coherence (agent profiles and goals) with the social force model for physical plausibility. The paper describes a user-friendly interface for customizing simulations and analyzing outputs, and states that it conducted a quantitative experiment to evaluate the workflow's effectiveness together with a user study (n=12) to assess usefulness and efficiency.

Significance. If the quantitative experiment demonstrates that the LLM-social force workflow produces trajectories that are measurably both physically plausible and semantically coherent, the work could offer a practical advance in semantic trajectory simulation by lowering barriers to adoption compared with existing tools that require substantial technical expertise. The user study could further establish the system's value for practitioners. The proposal itself addresses a clear gap between costly real-world data collection and accessible simulation.

major comments (2)
  1. [Abstract] Abstract: The manuscript states that a quantitative experiment was conducted to evaluate the effectiveness of the simulation workflow that combines LLMs and the social force model, yet provides no description of the experimental design, metrics for physical plausibility versus semantic coherence, baselines, results, error bars, or statistical tests. This omission leaves the central claim without visible empirical support.
  2. [Abstract] Abstract: The user study is referenced only by sample size (n=12) with no information on tasks performed, dependent measures, quantitative or qualitative findings, or comparison conditions, preventing assessment of the claimed usefulness and efficiency of the interface.
minor comments (1)
  1. [Abstract] The abstract introduces the notion of 'zoned environments' without defining zoning or explaining how it interacts with the LLM-social force workflow; a brief clarification would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive feedback on the abstract. We agree that the abstract should be expanded to provide sufficient detail on the quantitative experiment and user study so that the central claims are empirically supported within the abstract itself. We will revise the abstract in the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The manuscript states that a quantitative experiment was conducted to evaluate the effectiveness of the simulation workflow that combines LLMs and the social force model, yet provides no description of the experimental design, metrics for physical plausibility versus semantic coherence, baselines, results, error bars, or statistical tests. This omission leaves the central claim without visible empirical support.

    Authors: We agree that the current abstract does not include these details. The full manuscript contains a dedicated evaluation section that describes the experimental design (comparing LLM-augmented agents against social-force-only and random baselines in zoned environments), the metrics (physical plausibility via collision rate and trajectory smoothness; semantic coherence via goal-completion rate and profile alignment scored by an independent LLM judge), baselines, quantitative results with error bars, and statistical tests. To address the referee's concern directly in the abstract, we will add a concise summary of the design, metrics, and key findings (including significance) while respecting length limits. revision: yes

  2. Referee: [Abstract] Abstract: The user study is referenced only by sample size (n=12) with no information on tasks performed, dependent measures, quantitative or qualitative findings, or comparison conditions, preventing assessment of the claimed usefulness and efficiency of the interface.

    Authors: We acknowledge the abstract provides only the sample size. The full manuscript reports the user-study protocol (tasks: configuring zoned environments and agent profiles, running simulations, and inspecting outputs), dependent measures (NASA-TLX, SUS, task completion time, and open-ended feedback), quantitative results (mean scores and comparisons to a baseline non-LLM interface), and qualitative themes. We will revise the abstract to include a brief summary of tasks, measures, and main findings so readers can assess the claimed usefulness and efficiency. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is a system proposal describing SenseWalk, an interactive workflow that combines LLMs (for semantic profiles/goals) with the social force model (for physical movement) to generate trajectories. No equations, parameter-fitting steps, derivations, or load-bearing self-citations appear in the abstract or described content. The central claim is evaluated via a quantitative experiment and user study (n=12), which constitute independent tests rather than reductions to fitted inputs or self-referential definitions. This matches the default expectation for non-circular system papers in cs.HC.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no specific free parameters, axioms, or invented entities can be identified from the provided text. The social force model is referenced as an existing component but its parameters are not detailed here.

pith-pipeline@v0.9.1-grok · 5705 in / 1159 out tokens · 20463 ms · 2026-07-02T06:05:26.626075+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

76 extracted references · 54 canonical work pages

  1. [1]

    2015.Semantic Trajectories: A Survey from Modeling to Application

    BasmaH.Albanna,IbrahimF.Moawad,SherinM.Moussa,andMahmoudA.Sakr. 2015.Semantic Trajectories: A Survey from Modeling to Application. Springer International Publishing, Cham, 59–76. doi:10.1007/978-3-319-16667-4_4

    work page doi:10.1007/978-3-319-16667-4_4 2015

  2. [2]

    Mohammed Almutairi, Charles Chiang, Haoze Guo, Matthew Belcher, Nandini Banerjee, Maria Milkowski, Svitlana Volkova, Daniel Nguyen, Tim Weninger, III Yankoski, Michael G, Trenton W Ford, and Diego Gómez-Zará. 2025. VIRT- LAB: An AI-Powered System for Flexible, Customizable, and Large-scale Team Simulations. InAdjunct Proceedings of the 38th Annual ACM Sym...

    work page doi:10.1145/3746058 2025

  3. [3]

    Sotiris Angelis, Konstantinos Kotis, and Dimitris Spiliotopoulos. 2021. Semantic TrajectoryAnalyticsandRecommenderSystemsinCulturalSpaces.BigDataand Cognitive Computing5, 4 (2021). doi:10.3390/bdcc5040080

    work page doi:10.3390/bdcc5040080 2021

  4. [5]

    Agent-BasedModelinginEconomics and Finance: Past, Present, and Future.Journal of Economic Literature63, 1 (March 2025), 197–287

    RobertL.AxtellandJ.DoyneFarmer.2025. Agent-BasedModelinginEconomics and Finance: Past, Present, and Future.Journal of Economic Literature63, 1 (March 2025), 197–287. doi:10.1257/jel.20221319

    work page doi:10.1257/jel.20221319 2025

  5. [6]

    Federico Bianchi and Flaminio Squazzoni. 2015. Agent-based models in sociology.WIREs Computational Statistics7, 4 (2015), 284–306. arXiv:https://wires.onlinelibrary.wiley.com/doi/pdf/10.1002/wics.1356 doi:10. 1002/wics.1356

    work page doi:10.1002/wics.1356 2015

  6. [7]

    Ji Cao, Tongya Zheng, Qinghong Guo, Yu Wang, Junshu Dai, Shunyu Liu, Jie Yang, Jie Song, and Mingli Song. 2025. Holistic Semantic Representation for Navigational Trajectory Generation. arXiv:2501.02737 [cs.CV] https://arxiv.org/ abs/2501.02737

    work page arXiv 2025

  7. [8]

    Sergey Dorokhin, Alexander Artemov, Dmitry Likhachev, Alexey Novikov, and Evgeniy Starkov. 2020. Traffic simulation: an analytical review.IOP Conference Series: Materials Science and Engineering918, 1 (sep 2020), 012058. doi:10. 1088/1757-899X/918/1/012058

    2020

  8. [9]

    Yang Fang, Bei Luo, Ting Zhao, Dong He, Bingbing Jiang, and Qilie Liu. 2022. ST-SIGMA: Spatio-temporal semantics and interac- tion graph aggregation for multi-agent perception and trajectory forecast- ing.CAAI Transactions on Intelligence Technology7, 4 (2022), 744–

    2022

  9. [10]

    arXiv:https://ietresearch.onlinelibrary.wiley.com/doi/pdf/10.1049/cit2.12145 doi:10.1049/cit2.12145

    work page doi:10.1049/cit2.12145

  10. [11]

    Santiago Garcia-Gabilondo, Yuya Shibuya, and Yoshihide Sekimoto. 2024. En- hancing geospatial retail analysis by integrating synthetic human mobility sim- ulations.Computers, Environment and Urban Systems108 (2024), 102058. doi:10.1016/j.compenvurbsys.2023.102058

    work page doi:10.1016/j.compenvurbsys.2023.102058 2024

  11. [12]

    Anindya Ghose, Beibei Li, and Siyuan Liu. 2019. Mobile Targeting Using Customer Trajectory Patterns.Management Science65, 11 (2019), 5027–5049. arXiv:https://doi.org/10.1287/mnsc.2018.3188 doi:10.1287/mnsc.2018.3188

    work page doi:10.1287/mnsc.2018.3188 2019

  12. [13]

    Fosca Giannotti, Mirco Nanni, Dino Pedreschi, Fabio Pinelli, Chiara Renso, Salvatore Rinzivillo, and Roberto Trasarti. 2011. Unveiling the complexity of human mobility by querying and mining massive trajectory data.The VLDB Journal20, 5 (Oct. 2011), 695–719. doi:10.1007/s00778-011-0244-8

    work page doi:10.1007/s00778-011-0244-8 2011

  13. [14]

    Trajectory patternmining.InProceedingsofthe13thACMSIGKDDInternationalConference on Knowledge Discovery and Data Mining(San Jose, California, USA)(KDD ’07)

    FoscaGiannotti,MircoNanni,FabioPinelli,andDinoPedreschi.2007. Trajectory patternmining.InProceedingsofthe13thACMSIGKDDInternationalConference on Knowledge Discovery and Data Mining(San Jose, California, USA)(KDD ’07). Association for Computing Machinery, New York, NY, USA, 330–339. doi:10.1145/1281192.1281230

    work page doi:10.1145/1281192.1281230 2007

  14. [15]

    Newsl.10, 1 (May 2008), 40–49

    ArisGkoulalas-DivanisandVassiliosS.Verykios.2008.Aprivacy-awaretrajectory tracking query engine.SIGKDD Explor. Newsl.10, 1 (May 2008), 40–49. doi:10. 1145/1412734.1412739

    work page arXiv 2008

  15. [16]

    LetianGong,YanLin,XinyueZhang,YiwenLu,XuediHan,YichenLiu,Shengnan Guo, Youfang Lin, and Huaiyu Wan. 2024. Mobility-LLM: Learning Visiting Intentions and Travel Preference from Human Mobility Data with Large Language Models. InAdvances in Neural Information Processing Systems, A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhan...

    work page doi:10.52202/079017-1141 2024

  16. [17]

    Tesfay Gidey Hailu, Xiansheng Guo, Haonan Si, Lin Li, and Yukun Zhang. 2024. Theories and Methods for Indoor Positioning Systems: A Comparative Analysis, Challenges, and Prospective Measures.Sensors24, 21 (2024). doi:10.3390/ s24216876

    2024

  17. [18]

    Zhenxuan He, Tong Zhang, Wangshu Wang, and Jing Li

  18. [19]

    A deep pedestrian trajectory generator for complex in- door environments.Transactions in GIS28, 2 (2024), 411–

    2024

  19. [20]

    arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1111/tgis.13143 doi:10.1111/tgis.13143

    work page doi:10.1111/tgis.13143

  20. [21]

    Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber

  21. [22]

    InThe Twelfth International Conference on Learning Representations

    MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework. InThe Twelfth International Conference on Learning Representations. https: //openreview.net/forum?id=VtmBAGCN7o

  22. [23]

    Erzhen Hu, Yanhe Chen, Mingyi Li, Vrushank Phadnis, Pingmei Xu, Xun Qian, Alex Olwal, David Kim, Seongkook Heo, and Ruofei Du. 2025. DialogLab: Authoring, Simulating, and Testing Dynamic Human-AI Group Conversations. In Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, N...

    work page doi:10.1145/3746059.3747696 2025

  23. [24]

    Zeyuan Huang, Cangjun Gao, Yaxian Shan, Haoxiang Hu, Qingkun Li, Xiaoming Deng, Cuixia Ma, Yu-Kun Lai, Yong-Jin Liu, Feng Tian, Guozhong Dai, and Hongan Wang. 2025. SketchGPT: A Sketch-based Multimodal Interface for Application-Agnostic LLM Interaction. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). As...

    work page arXiv 2025

  24. [25]

    TrajLLM:A Modular LLM-Enhanced Agent-Based Framework for Realistic Human Trajectory Simulation

    ChenluJu,JiaxinLiu,ShobhitSinha,HaoXue,andFloraSalim.2025. TrajLLM:A Modular LLM-Enhanced Agent-Based Framework for Realistic Human Trajectory Simulation. InCompanion Proceedings of the ACM on Web Conference 2025. Association for Computing Machinery, New York, NY, USA, 2847–2850. doi:10. 1145/3701716.3715201

    work page arXiv 2025

  25. [26]

    Kyuwon Kim, Jaeryeong Hwang, Younseo Lee, Jeanhee Lee, Sung-Eun Kim, and Hyo-Jeong So. 2025. Your Thoughtful Opponent: Embracing Cognitive Conflict with Peer Agent. InAdjunct Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST Adjunct ’25). Association for Computing Machinery, New York, NY, USA, Article 180, 3 page...

    work page doi:10.1145/3746058.3758410 2025

  26. [27]

    doi:10.1007/s10707- 020-00430-x

    Alexandros Kontarinis, Karine Zeitouni, Claudia Marinica, Dan Vodislav, and DimitrisKotzinos.2021.Towardsasemanticindoortrajectorymodel:applicationto museumvisits.GeoInformatica25,2(April2021),311–352. doi:10.1007/s10707- 020-00430-x

    work page doi:10.1007/s10707- 2021

  27. [29]

    Francesco Lettich, Chiara Pugliese, Chiara Renso, and Fabio Pinelli. 2023. Se- mantic Enrichment of Mobility Data: A Comprehensive Methodology and the MAT-BUILDER System.IEEE Access11 (2023), 90857–90875. doi:10.1109/ ACCESS.2023.3307824

    work page arXiv 2023

  28. [30]

    AsurveyonLLM-based multi-agent systems: workflow, infrastructure, and challenges.Vicinagearth1, 1 (Oct

    XinyiLi,SaiWang,SiqiZeng,YuWu,andYiYang.2024. AsurveyonLLM-based multi-agent systems: workflow, infrastructure, and challenges.Vicinagearth1, 1 (Oct. 2024), 9. doi:10.1007/s44336-024-00009-2

    work page doi:10.1007/s44336-024-00009-2 2024

  29. [31]

    Ziyue Lin, Yi Shan, Lin Gao, Xinghua Jia, and Siming Chen. 2025. SimSpark: Interactive Simulation of Social Media Behaviors.Proc. ACM Hum.-Comput. Conference acronym ’XX, June 03–05, 2018, Woodstock, NY Trovato et al. Interact.9, 2, Article CSCW168 (May 2025), 32 pages. doi:10.1145/3711066

    work page doi:10.1145/3711066 2025

  30. [32]

    Ziyue Lin, Siqi Shen, Zichen Cheng, Cheok Lam Lai, and Siming Chen. 2025. Carbon and Silicon, Coexist or Compete? A Survey on Human-AI Interactions in Agent-based Modeling and Simulation. arXiv:2502.18145 [cs.HC] https: //arxiv.org/abs/2502.18145

    work page arXiv 2025

  31. [33]

    arXiv preprint arXiv:2506.23306 , year=

    QiLiu,CanLi,andWanjingMa.2026. GATSim:UrbanMobilitySimulationwith Generative Agents. arXiv:2506.23306 [cs.AI] https://arxiv.org/abs/2506.23306

    work page arXiv 2026

  32. [34]

    WeiwenLiu,XuHuang,XingshanZeng,XinlongHao,Shuai Yu,Dexun Li,Shuai Wang, Weinan Gan, Zhengying Liu, Yuanqing Yu, Zezhong Wang, Yuxian Wang, Wu Ning, Yutai Hou, Bin Wang, Chuhan Wu, Xinzhi Wang, Yong Liu, Yasheng Wang, Duyu Tang, Dandan Tu, Lifeng Shang, Xin Jiang, Ruiming Tang, Defu Lian, Qun Liu, and Enhong Chen. 2025. ToolACE: Winning the Points of LLM F...

    work page arXiv 2025

  33. [35]

    Gregor Lämmel and Gunnar Flötteröd. 2015. A CA Model for Bidirectional Pedestrian Streams.Procedia Computer Science52 (2015), 950–955. doi:10. 1016/j.procs.2015.05.171 The 6th International Conference on Ambient Systems, Networks and Technologies (ANT-2015), the 5th International Conference on Sustainable Energy Information Technology (SEIT-2015)

    2015

  34. [36]

    Macal and M.J

    C.M. Macal and M.J. North. 2005. Tutorial on agent-based modeling and simulation. InProceedings of the Winter Simulation Conference, 2005.14 pp.–. doi:10.1109/WSC.2005.1574234

    work page doi:10.1109/wsc.2005.1574234 2005

  35. [37]

    Macal and Michael J

    Charles M. Macal and Michael J. North. 2009. Agent-based modeling and simulation. InProceedings of the 2009 Winter Simulation Conference (WSC). 86–98. doi:10.1109/WSC.2009.5429318

    work page doi:10.1109/wsc.2009.5429318 2009

  36. [38]

    YuMei,YuanxiWang,ShiyiWang,QingyangWan,ZhuojunLi,ChunYu,Weinan Shi, and Yuanchun Shi. 2025. InterQuest: A Mixed-Initiative Framework for Dynamic User Interest Modeling in Conversational Search. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Articl...

    work page doi:10.1145/3746059.3747753 2025

  37. [39]

    Michael Mireku Kwakye. 2017. Modelling and Design of Generic Semantic Trajectory Data Warehouse. doi:10.11575/PRISM/31331

    work page doi:10.11575/prism/31331 2017

  38. [40]

    Mehdi Moussaïd, Niriaska Perozo, Simon Garnier, Dirk Helbing, and Guy Ther- aulaz. 2010. The Walking Behaviour of Pedestrian Social Groups and Its Impact on Crowd Dynamics.PLOS ONE5, 4 (04 2010), 1–7. doi:10.1371/journal.pone. 0010047

    work page doi:10.1371/journal.pone 2010

  39. [41]

    Christian Nitzsche and Stefan Simm. 2024. Agent-based modeling to estimate the impactoflockdownscenariosandeventsonapandemicexemplifiedonSARS-CoV- 2.Scientific Reports14, 1 (June 2024), 13391. doi:10.1038/s41598-024-63795-1 Publisher: Nature Publishing Group

    work page doi:10.1038/s41598-024-63795-1 2024

  40. [42]

    ChristineParent,StefanoSpaccapietra,ChiaraRenso,GennadyAndrienko,Natalia Andrienko, Vania Bogorny, Maria Luisa Damiani, Aris Gkoulalas-Divanis, Jose Macedo, Nikos Pelekis, Yannis Theodoridis, and Zhixian Yan. 2013. Semantic trajectories modeling and analysis.ACM Comput. Surv.45, 4, Article 42 (Aug. 2013), 32 pages. doi:10.1145/2501654.2501656

    work page doi:10.1145/2501654.2501656 2013

  41. [43]

    Bernstein

    Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. Generative Agents: Interactive Simulacra of Human Behavior. InProceedings of the 36th Annual ACM Symposium on User Interface Software and Technology(San Francisco, CA, USA)(UIST ’23). Association for Computing Machinery, New York, NY, USA, ...

    work page doi:10.1145/3586183.3606763 2023

  42. [44]

    Bernstein

    Joon Sung Park, Lindsay Popowski, Carrie Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2022. Social Simulacra: Creating Populated Prototypes for Social Computing Systems. InProceedings of the 35th Annual ACM Symposium on User Interface Software and Technology(Bend, OR, USA) (UIST ’22). Association for Computing Machinery, New York, N...

    work page doi:10.1145/3526113.3545616 2022

  43. [45]

    2012.Problem seeking: An architectural programming primer

    William M Pena and Steven A Parshall. 2012.Problem seeking: An architectural programming primer. John Wiley & Sons

    2012

  44. [46]

    Philippopoulos, Ioannis C

    Panos I. Philippopoulos, Ioannis C. Drivas, Nikolaos D. Tselikas, Kostas N. Koutrakis, Elena Melidi, and Dimitrios Kouis. 2024. A Holistic Approach for Enhancing Museum Performance and Visitor Experience.Sensors24, 3 (2024). doi:10.3390/s24030966

    work page doi:10.3390/s24030966 2024

  45. [47]

    Aurora Polo-Rodríguez, Miguel Ángel Anguita-Molina, Ignacio Rojas-Ruiz, and Javier Medina-Quero. 2025. Multi-occupant tracking with radar and wearable devices for enhanced accuracy in indoor environments.Engineering Applications ofArtificialIntelligence154(2025),110872. doi:10.1016/j.engappai.2025.110872

    work page doi:10.1016/j.engappai.2025.110872 2025

  46. [48]

    Markazi, and Mo- hammadreza Bemanian

    Morteza Rahbar, Mohammadjavad Mahdavinejad, Amir H.D. Markazi, and Mo- hammadreza Bemanian. 2022. Architectural layout design through deep learning and agent-based modeling: A hybrid approach.Journal of Building Engineering 47 (2022), 103822. doi:10.1016/j.jobe.2021.103822

    work page doi:10.1016/j.jobe.2021.103822 2022

  47. [49]

    Amir Rasouli and Iuliia Kotseruba. 2022. PedFormer: Pedestrian Behavior Prediction via Cross-Modal Attention Modulation and Gated Multitask Learning. arXiv:2210.07886 [cs.CV] https://arxiv.org/abs/2210.07886

    work page arXiv 2022

  48. [50]

    2022.Simulation

    Sheldon M Ross. 2022.Simulation. academic press

    2022

  49. [51]

    Michael J Seitz, Felix Dietrich, Gerta Köster, and Hans-Joachim Bungartz. 2016. The superposition principle: A conceptual perspective on pedestrian stream simulations.Collective Dynamics1 (2016), 1–19

    2016

  50. [52]

    Donghoon Shin, Daniel Lee, Gary Hsieh, and Gromit Yeuk-Yin Chan. 2025. PosterMate: Audience-driven Collaborative Persona Agents for Poster Design. In Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Article 201, 20 pages. doi:10.1145/3746059.3747769

    work page doi:10.1145/3746059.3747769 2025

  51. [53]

    Stefano Spaccapietra, Christine Parent, Maria Luisa Damiani, Jose Antonio de Macedo, Fabio Porto, and Christelle Vangenot. 2008. A conceptual view on trajectories.Data & Knowledge Engineering65, 1 (2008), 126–146. doi:10.1016/ j.datak.2007.10.008 Including Special Section: Privacy Aspects of Data Mining Workshop (2006) - Five invited and extended papers

    2008

  52. [54]

    Shayan Talaei, Meijin Li, Kanu Grover, James Kent Hippler, Diyi Yang, and Amin Saberi. 2025. StorySage: Conversational Autobiography Writing Powered by a Multi-Agent Framework. InProceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST ’25). Association for Computing Machinery, New York, NY, USA, Article 209, 26 pages. ...

    work page doi:10.1145/3746059 2025

  53. [55]

    JiaweiWang,RenheJiang,ChuangYang,ZengqingWu,MakotoOnizuka,Ryosuke Shibasaki, Noboru Koshizuka, and Chuan Xiao. 2024. Large Language Models as Urban Residents: An LLM Agent Framework for Personal Mobility Generation. In Advances in Neural Information Processing Systems, A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang (Eds....

    work page doi:10.52202/079017-3957 2024

  54. [56]

    Asurveyonlargelanguagemodelbasedautonomousagents

    Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, Wayne Xin Zhao, Zhewei Wei, andJirongWen.2024. Asurveyonlargelanguagemodelbasedautonomousagents. Frontiers of Computer Science18, 6 (March 2024), 186345. doi:10.1007/s11704- 024-40231-1

    work page doi:10.1007/s11704- 2024

  55. [57]

    Ruhan Wang, Shuowen Li, Peiran Zhang, Danqi Huang, Yijie Guo, and Haipeng Mi. 2025. PuppetLine: An Interactive System for Embodied Storytelling with LLM-driven Swarm Robots. InAdjunct Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology (UIST Adjunct ’25). Association for Computing Machinery, New York, NY, USA, Article 48...

    work page doi:10.1145/3746058.3759016 2025

  56. [58]

    2011.An Integrated Pedestrian Behavior Model Based on Extended Decision Field Theory and Social Force Model

    Hui Xi, Seungho Lee, and Young-Jun Son. 2011.An Integrated Pedestrian Behavior Model Based on Extended Decision Field Theory and Social Force Model. Springer London, London, 69–95. doi:10.1007/978-0-85729-883-6_4

    work page doi:10.1007/978-0-85729-883-6_4 2011

  57. [59]

    Claudia Yamu, Akkelies van Nes, and Chiara Garau. 2021. Bill Hillier’s Legacy: Space Syntax—A Synopsis of Basic Concepts, Measures, and Empirical Applica- tion.Sustainability13, 6 (2021). doi:10.3390/su13063394

    work page doi:10.3390/su13063394 2021

  58. [60]

    Yuwei Yan, Qingbin Zeng, Zhiheng Zheng, Jingzhe Yuan, Jie Feng, Jun Zhang, Fengli Xu, and Yong Li. 2024. OpenCity: A Scalable Platform to Simulate Urban Activities with Massive LLM Agents. arXiv:2410.21286 [cs.MA] https: //arxiv.org/abs/2410.21286

    work page arXiv 2024

  59. [61]

    SeMiTri:aframeworkforsemanticannotationofheterogeneous trajectories

    Zhixian Yan, Dipanjan Chakraborty, Christine Parent, Stefano Spaccapietra, and KarlAberer.2011. SeMiTri:aframeworkforsemanticannotationofheterogeneous trajectories. InProceedings of the 14th International Conference on Extending Database Technology(Uppsala, Sweden)(EDBT/ICDT ’11). Association for Computing Machinery, New York, NY, USA, 259–270. doi:10.114...

    work page doi:10.1145/1951365 2011

  60. [62]

    Shanwen Yang, Tianrui Li, Xun Gong, Bo Peng, and Jie Hu. 2020. A review on crowd simulation and modeling.Graphical Models111 (2020), 101081. doi:10.1016/j.gmod.2020.101081

    work page doi:10.1016/j.gmod.2020.101081 2020

  61. [63]

    Josh Jia-Ching Ying, Wang-Chien Lee, Tz-Chiao Weng, and Vincent S. Tseng

  62. [64]

    InProceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems(Chicago, Illinois)(GIS ’11)

    Semantic trajectory mining for location prediction. InProceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems(Chicago, Illinois)(GIS ’11). Association for Computing Machinery, New York, NY, USA, 34–43. doi:10.1145/2093973.2093980

    work page doi:10.1145/2093973.2093980

  63. [65]

    JoshJia-ChingYing,EricHsueh-ChanLu,Wang-ChienLee,Tz-ChiaoWeng,and Vincent S. Tseng. 2010. Mining user similarity from semantic trajectories. InPro- ceedingsofthe2ndACMSIGSPATIALInternationalWorkshoponLocationBased Social Networks(San Jose, California)(LBSN ’10). Association for Computing Machinery, New York, NY, USA, 19–26. doi:10.1145/1867699.1867703

    work page doi:10.1145/1867699.1867703 2010

  64. [66]

    Ryo Yonetani, Jun Baba, and Yasutaka Furukawa. 2024. RetailOpt: Opt-In, Easy-to-Deploy Trajectory Estimation from Smartphone Motion Data and Retail Facility Information. InProceedings of the 2024 ACM International Symposium on Wearable Computers(Melbourne VIC, Australia)(ISWC ’24). Association for Computing Machinery, New York, NY, USA, 125–132. doi:10.11...

    work page doi:10.1145/3675095 2024

  65. [67]

    Hanzhong Zhang, Jibin Yin, Mulin Jiang, and Cong Su. 2025. Can Agents Spontaneously Form a Society? Introducing a Novel Architecture for Generative Multi-Agents to Elicit Social Emergence. InAdjunct Proceedings of the 38th AnnualACMSymposiumonUserInterfaceSoftwareandTechnology(UISTAdjunct ’25). Association for Computing Machinery, New York, NY, USA, Artic...

    work page doi:10.1145/3746058.3758414 2025

  66. [68]

    Zhishuai Zhang, Jiyang Gao, Junhua Mao, Yukai Liu, Dragomir Anguelov, and Congcong Li. 2020. STINet: Spatio-Temporal-Interactive Network for Pedestrian Detection and Trajectory Prediction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). SenseWalk: Agent-Based Semantic Trajectory Simulation Powered by Large Langu...

    2020

  67. [69]

    Yi Zuo, Junhao Jiang, and Katsutoshi Yada. 2025. Application of hybrid gate recurrent unit for in-store trajectory prediction based on indoor location system. Scientific Reports15, 1 (Jan. 2025), 1055. doi:10.1038/s41598-024-84599-3 Publisher: Nature Publishing Group

    work page doi:10.1038/s41598-024-84599-3 2025

  68. [70]

    Marta Šveb Dragija, Martijn van Zomeren, and Nina Hansen. 2024. De- signing museum experiences for eudaimonic or hedonic well-being: insights from interviews with museum visitors.Museum Management and Curator- ship0, 0 (2024), 1–18. arXiv:https://doi.org/10.1080/09647775.2024.2408239 doi:10.1080/09647775.2024.2408239 A Preliminary Study Table 1 shows the ...

    work page doi:10.1080/09647775.2024.2408239 2024

  69. [71]

    A JSON-formatted exhibition map

  70. [72]

    far left→far right→far left again

    A user's navigation request Your goal: Within the user's available time, plan a visit order that allows them to see as many booths as possible, based on the map data. ## 1. Coordinate System - The map is a 2D plane. ## 2. JSON Field Description - Booth name (unique identifier) -`pos`: [x, y, w, h]→booth position and size -`des`: booth description -`Countr...

    2018

  71. [73]

    PLAN: Decide which POIs/events this visitor should go to, in what order, and for how long

  72. [74]

    You must output a single JSON document capturing both the high-level visit plan and the detailed physical trajectory

    SIMULATE: Generate realistic physical movement trajectories through the venue space. You must output a single JSON document capturing both the high-level visit plan and the detailed physical trajectory. ## COORDINATE SYSTEM - Origin (0,0) is top-left. X increases rightward, Y increases downward. - Walls are hard boundaries - the agent CANNOT pass through ...

  73. [75]

    All walking segments must obey this

    WALKING SPEED is exactly 1 unit / 3 seconds (5 units / 15 seconds). All walking segments must obey this

  74. [76]

    Never change both X and Y in a single segment

    AXIS-ALIGNED MOTION ONLY: Between any two consecutive walking points, at most ONE coordinate (X or Y) may change. Never change both X and Y in a single segment

  75. [77]

    b) At every TURN (corner): where the direction changes (e.g., from moving along X to moving along Y)

    WHEN TO OUTPUT A POINT (walking): a) At the START of the trajectory (initial position). b) At every TURN (corner): where the direction changes (e.g., from moving along X to moving along Y). Output the corner point BEFORE the direction changes. c) When current_destination CHANGES (the agent starts heading to a new place). Conference acronym ’XX, June 03–05...

    2018

  76. [78]

    browsing

    BROWSING / ATTENDING (Time Jump): - One point on ARRIVAL (action="browsing" or "attending"). - Next point is DEPARTURE after the stay (action="walking", with new current_destination). Time jumps forward by stay duration (>=15 min for stores, full duration for events). ## OUTPUT FORMAT Return ONLY valid JSON: {{ "agent_summary": {{ ... }}, "visit_plan": [ ...

    2025