MeshForce: An AI-Assisted Real-Time Volunteer Dispatch System for Mass Gathering Events — A Case Study on Mahakumbh 2025

References

[1] M. Yamin, “Managing crowds with technology: cases of Hajj and Kumbh Mela,” International Journal of Information Technology, vol. 11, no. 2, pp. 229–237, Jun. 2019. Springer

[2] H. Verma, V. Dewangan, and S. Wadhwa, “Sangam: Smart Crowd Management and Assistance Platform,” International Journal for Research in Applied Science and Engineering Technology (IJRASET), 2025. IJRASET

[3] “At the Mahakumbh, Faith Met Tragedy: Computational Analysis of Stampede Patterns Using Machine Learning and NLP,” arXiv:2502.03120, 2025. arXiv

[4] R. Roy Choudhury, A. Karmakar, and R. P. Mitra, “Drishti AI-Event Guardian: An Intelligent Real-Time Crowd Monitoring and Emergency Response System for Mass Gathering Events,” arXiv:2606.05185, Kalinga Institute of Industrial Technology, Jun. 2026. arXiv

[5] M. Falasca and C. Zobel, “An optimization model for volunteer assignments in humanitarian organizations,” Socio-Economic Planning Sciences, vol. 46, no. 4, pp. 250–260, 2012. ScienceDirect

[6] N. Matinrad and T. Andersson Granberg, “Optimal pre-dispatch task assignment of volunteers in daily emergency response,” Socio-Economic Planning Sciences, vol. 87, part B, Art. no. 101589, 2023. ScienceDirect

[7] “AI to ensure efficient crowd control during Kumbh Mela,” The Week, Jan. 2019. The Week

[8] “Maha Kumbh 2025: Organisers use AI to prevent stampedes, ensure safety of 400mn pilgrims,” WION, Jan. 2025. WION

[9] M. Ptaszynski, F. Masui, Y. Nakajima, H. Hayakawa, T. Saito, and Y. Miyamori, “Comparison of machine learning techniques for classification of information types on Twitter,” in Proc. Annual Meeting of the Association for NLP, 2019.

[10] Asinthara K., M. Jayan, and L. Jacob, “Classification of Disaster Tweets using Machine Learning and Deep Learning Techniques,” in Proc. 2022 Int. Conf. on Trends in Quantum Computing and Emerging Business Technologies (TQCEBT), IEEE, 2022.

[11] J. Pánek, L. Marek, V. Pászto, and J. Valuch, “The Crisis Map of the Czech Republic: the nationwide deployment of an Ushahidi application for disasters,” Disasters, vol. 41, no. 4, pp. 649–671, 2016.

[12] Anthropic, “Claude API Documentation,” [Online]. Available: Anthropic. [Accessed: Sep. 2026].

[13] Supabase Inc., “Supabase Realtime Documentation,” [Online]. Available: Supabase. [Accessed: Sep. 2026].

[14] U. D. Wisnesky, S. W. Kirkland, B. H. Rowe, S. Campbell, and J. M. Franc, “A Qualitative Assessment of Studies Evaluating the Classification Accuracy of Personnel Using START in Disaster Triage: A Scoping Review,” Frontiers in Public Health, vol. 10, Art. no. 676704, 2022. Frontiers

[15] “What’s an Incident? ITIL Priority Levels: Critical, High, Medium and Low,” OnPage Corporation. [Online]. Available: OnPage. [Accessed: Sep. 2026].

[16] World Health Organization, Public Health for Mass Gatherings: Key Considerations. WHO Global Capacity Alert and Response Department, Geneva, 2015, ISBN 978-92-4-156493-9.

[17] R. W. Sinnott, “Virtues of the Haversine,” Sky and Telescope, vol. 68, no. 2, p. 159, 1984.

@article{1723466,
      author = {Anwesa Ray, Dr. R. Senthil Kumar, Diya Mittal, Siya Bojewar; Naman Shrivastava, Khushi Gupta},
      title = {MeshForce: An AI-Assisted Real-Time Volunteer Dispatch System for Mass Gathering Events — A Case Study on Mahakumbh 2025},
      journal = {Iconic Research And Engineering Journals},
      year = {2026},
      volume = {10},
      number = {3},
      pages = {3376-3382},
      issn = {2456-8880},
      url = {https://www.irejournals.com/formatedpaper/1723466.pdf},
      abstract = {Mass religious gatherings such as the Mahakumbh attract tens of millions of pilgrims within a temporary, high-density event footprint, placing enormous coordination demands on the volunteer workforce responsible for medical, crowd-control, and lost-and-found response. Existing coordination at such events is largely manual, relying on radio calls and word-of-mouth routing, which delays the matching of an available, appropriately skilled volunteer to an emerging incident [1], [3]. This paper presents MeshForce, an AI-assisted, real-time volunteer dispatch system comprising a volunteer Progressive Web App (PWA), a FastAPI backend, a Supabase (PostgreSQL + Realtime) data layer, and an administrative command-center dashboard with a live map. Incoming incident reports, submitted in free-form natural language or via SMS in low-connectivity conditions, are parsed into structured metadata using a large language model, and a composite scoring function combining geodesic distance, skill overlap, language match, and volunteer exhaustion ranks and dispatches the best-available volunteers. We describe the system architecture, the dispatch scoring algorithm, and a cost-conscious mock/production LLM design that enables full-scale simulation without incurring inference cost. Simulated evaluation with 50 volunteers and 15 concurrent incidents is used to validate dispatch latency and correctness against the system's stated non-functional requirements. We discuss applicability of this architecture to other mass-gathering and disaster-response contexts and outline limitations, including the absence of a field trial.},
      keywords = {AI-Assisted Volunteer Dispatch, LLM-Based Incident Triage, Mass-Gathering Crowd Coordination, Skill-Fatigue-Aware Scoring Algorithm, SMS-Fallback Crisis Reporting.},
      month = {September},
  }