Claudio Marche, Vlad Popescu, L Atzori, Michele Nitti
The Internet of Things (IoT) is increasingly supporting the tourism sector by enabling adaptive services that enhance the visitor experience. Among emerging applications, itinerary planning has gained significant attention, leveraging IoT data to deliver personalized and adaptive routes. However, current systems show key limitations. Personalization often depends on static forms rather than adaptive learning; validation is usually restricted to simulations, and the few real implementations mostly rely on mobile apps, tools that tourists are reluctant to download and often abandon after limited use. To overcome these challenges, this paper proposes an itinerary planning system that integrates reinforcement learning for tourist profiling with a genetic algorithm for multi-objective optimization, implemented in a real-world scenario through a cloud infrastructure and delivered via an interactive totem that serves as the access point for tourists. Results confirm both the efficiency and scalability of the approach, showing that the system can be seamlessly extended to diverse urban contexts.