LLM-TripPlanner: A Large-Language-Model-Based Agent for Personalized Trip Planning

Abstract

Personalized trip planning is crucial for improving travel experiences, as it provides tailored recommendations for destinations, modes of transport, and routes that align closely with individual preferences. In this context, Large Language Models (LLMs) have also been getting involved with their contextual understanding and reasoning abilities. Additionally, the availability of urban big data, including street-view imagery and real-time traffic information, offers opportunities for data-driven and context-aware trip planning. By integrating semantic segmentation techniques with interpretations derived from LLMs, we developed a city-scale urban database that enables fine-grained characterization of travel routes. This study proposes an agent framework for end-to-end personalized trip planning, integrating LLMs with domain knowledge and external tools to generate realistic and user-specific travel plans. A retrieval-augmented generation (RAG) approach incorporates route information, real-time traffic data, and user-specific preferences to enhance LLM performance in route planning. Real-world experiments show that our travel agent meets user-specific needs, completing the entire personalized trip planning pipeline—from preference interpretation to final route selection. This framework significantly advances existing methods by providing a more comprehensive and efficient solution for personalized travel planning.

Links

DOI: 10.1109/ITSC60802.2025.11423528