MULTI-AGENT TRAVEL PLANNER

Abstract

This paper proposes an intelligent travel planning system based on a graph-driven multi-agent architecture implemented using LangGraph. The model is designed to automate itinerary generation by decomposing the planning process into specialized agents, including destination identification, place and accommodation recommendation, budget estimation, and itinerary optimization. The system utilizes structured datasets comprising tourist locations and Airbnb listings to extract key features such as ratings, cost components, and geographic attributes. A rule-based filtering algorithm combined with parallel recommendation strategies is employed to select optimal destinations and stays, while a cost aggregation model ensures adherence to user-defined budget constraints. The architecture is formally represented as a directed graph, enabling efficient coordination between agents through a shared state mechanism. Experimental analysis shows that the system produces consistent and feasible travel plans with clear cost distribution, where accommodation constitutes the largest share (~59%), followed by food (~30%) and activities (~10.9%). The results highlight the system’s low computational overhead, fast execution, and interpretability. The proposed framework demonstrates a scalable and extensible solution for intelligent travel planning, with potential for future inte