ATLAS - Where Travel Meets AI: An AI-Powered Travel Planning and Decision-Support Platform with Live Data Grounding

Abstract

Travel planning is a fragmented, multi-domain decision process: a traveller must separately research destinations, transport, accommodation, food, activities and weather before assembling a feasible schedule. This paper presents ATLAS (“Where Travel Meets AI”), an implemented AI-powered travel planning and decision-support platform that pairs a large language model (LLM) planner with deterministic grounding against live geospatial and environmental data. Given a destination, an inclusive date range, a boarding location, a budget and traveller preferences, ATLAS generates a day-by-day itinerary with the open-weight gpt-oss-120b model served through Groq, while route, weather, food and accommodation agents run concurrently and record the provenance of each data source. Model output is validated rather than trusted: a schema enforces exactly one itinerary day per requested calendar date, generated place names are resolved through OpenStreetMap’s Nominatim geocoder, and road distances and travel times are computed with OSRM from the traveller’s real boarding location, with unknown legs reported as unavailable instead of zero. The system is implemented as a React and TypeScript single-page application backed by FastAPI and PostgreSQL, and includes a live weather outlook, real restaurant and attraction listings, a simulated booking flow that issues a PDF e-ticket by e-mail, and a community Lost & Found module. We describe the architecture and grounding method, report a verification suite of 477 automated tests and preliminary functional observations, and state the system’s current limitations explicitly.

Introduction

ATLAS (“Where Travel Meets AI”) is an AI-powered travel planning and decision-support platform designed to simplify trip planning by integrating destination discovery, transportation, accommodation, restaurants, weather, and activities into a single personalized itinerary. Traditional travel platforms often provide these services separately, requiring users to manually compare information and organize their schedules.

The system uses a large language model (LLM) to generate itineraries based on users’ travel dates, budget, departure location, number of travellers, and personal preferences. It combines this planning capability with specialized agents for route calculation, weather forecasting, food discovery, and accommodation estimation. The platform also includes a conversational assistant, interactive itinerary and route viewer, simulated booking workflow, and personal dashboard.

ATLAS addresses a key limitation of AI-generated travel plans: inaccurate or unverified information. To improve reliability, it validates destinations using OpenStreetMap, calculates road routes using OSRM, and retrieves weather information from an external service. It also labels information as either live data or estimated data, ensuring users can distinguish verified information from approximations.

The system is implemented using React and TypeScript for the frontend, FastAPI for the backend, PostgreSQL for data storage, and multiple external APIs for travel-related services. Its modular architecture enables different components to operate concurrently, while structured validation ensures that generated itineraries follow the required format and include the requested travel dates.

The main contribution of ATLAS is the integration of AI-based itinerary generation with deterministic validation, specialized services, and transparent data provenance. Rather than relying entirely on AI-generated responses, the system combines AI flexibility with external verification to produce more consistent and practical travel plans.

Conclusion

This paper presented ATLAS, an implemented AI-powered travel planning and decision-support platform that combines an LLM planner with deterministic grounding against live geographic and environmental data. Its core contribution is a practical discipline for using a generative model in a factual domain: the model drafts the plan, while schemas, geocoding and road routing verify dates, places and travel times, and every value carries a record of whether it is live or estimated. The evaluation shows that this discipline catches errors that would otherwise reach the traveller unnoticed, and the limitations section states plainly what remains, chief among them agent-to-planner coordination and generation of longer itineraries.

References

[1] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, ?. Kaiser, and I. Polosukhin, “Attention is all you need,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2017. [2] T. Brown et al., “Language models are few-shot learners,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2020. [3] J. Wei et al., “Chain-of-thought prompting elicits reasoning in large language models,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2022. [4] S. Yao, J. Zhao, D. Yu, N. Du, I. Shafran, K. R. Narasimhan, and Y. Cao, “ReAct: Synergizing reasoning and acting in language models,” in Proc. Int. Conf. Learn. Represent. (ICLR), 2023. [5] Q. Wu et al., “AutoGen: Enabling next-gen LLM applications via multi-agent conversation,” Microsoft Research, Tech. Rep. MSR-TR-2023-33, 2023. [6] J. S. Park, J. C. O’Brien, C. J. Cai, M. R. Morris, P. Liang, and M. S. Bernstein, “Generative agents: Interactive simulacra of human behavior,” in Proc. ACM Symp. User Interface Softw. Technol. (UIST), 2023. [7] Y. Gao, Y. Xiong, X. Gao, K. Jia, J. Pan, Y. Bi, Y. Dai, J. Sun, and H. Wang, “Retrieval-augmented generation for large language models: A survey,” arXiv:2312.10997, 2023. [8] OpenAI, “Introducing gpt-oss,” Aug. 2025. [Online]. Available: https://openai.com/index/introducing-gpt-oss/ [9] F. Ricci, L. Rokach, and B. Shapira, Eds., Recommender Systems Handbook, 2nd ed. New York, NY, USA: Springer, 2015. [10] A. Bentaleb, Y. El Bouzekri, A. Ait Lahcen, and M. Boulmalf, “Context aware recommender systems for tourism: A concise review,” in Proc. IEEE 5th Int. Congr. Inf. Sci. Technol. (CiST), 2018. [11] J. Borràs, A. Moreno, and A. Valls, “Intelligent tourism recommender systems: A survey,” Expert Syst. Appl., vol. 41, no. 16, pp. 7370–7389, 2014. [12] M. Haklay and P. Weber, “OpenStreetMap: User-generated street maps,” IEEE Pervasive Comput., vol. 7, no. 4, pp. 12–18, 2008. [13] D. Luxen and C. Vetter, “Real-time routing with OpenStreetMap data,” in Proc. 19th ACM SIGSPATIAL Int. Conf. Adv. Geographic Inf. Syst., 2011, pp. 513–516. [14] OpenStreetMap Foundation, “Nominatim: Open-source geocoding with OpenStreetMap data.” [Online]. Available: https://nominatim.org/ [15] OpenWeather Ltd., “5 day / 3 hour forecast API.” [Online]. Available: https://openweathermap.org/forecast5 [16] S. Ramírez, “FastAPI.” [Online]. Available: https://fastapi.tiangolo.com/ [17] Groq, Inc., “GroqCloud documentation.” [Online]. Available: https://console.groq.com/docs

Copyright

Copyright © 2026 Rahul Kawariya, Aditya Doijad, Vaishnavi Kosarkar, Pratik Rathore , Ashima Sahu. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.