Abstract
The global tourism landscape is shifting toward more personalized, immersive, and experience-driven travel. Standardized travel packages are no longer sufficient-particularly for short urban trips lasting one to three days. Planning such trips is inherently complex, involving multiple objectives and constraints. Current online platforms still face major challenges in meeting diverse user needs, integrating heterogeneous data sources, and respecting practical limitations such as attraction operating hours or user budgets. To address these gaps, this study proposes a prototype Multi-Agent System (MAS) framework designed for personalized urban short-trip planning. The approach leverages the principle of division of labor and dynamic collaboration among specialized agents to seamlessly combine user preferences, tourism resources, and realworld constraints. The framework consists of three key agents: the User Agent, which models user preferences and limitations; the Resource Agent, which manages both static and dynamic information about points of interest (POIs); and the Planning & Coordination Agent, which constructs feasible itineraries. Inter-agent communication is coordinated through a simplified Contract Net Protocol (CNP). Experimental validation shows that the Planning Agent effectively generates itineraries that comply with time windows, budgets, and other constraints. The system also records the decision-making process, offering a degree of transparency that helps enhance user trust. Overall, this research lays a practical foundation for the next generation of intelligent tourism planning systems capable of delivering adaptive and trustworthy travel experiences.