Federated and Intelligent Digital Twins for Cultural Heritage: Frameworks, Applications, and Maturity Assessment

Cultural heritage institutions face increasing challenges related to environmental degradation, structural vulnerabilities, visitor congestion, and resource limitations. At the same time,the digitalization of cultural assets has led to fragmented and siloed datasets distributed across heterogeneous systems. While Digital Twins (DTs) have the potential to support real-time monitoring, predictive simulation, and operational optimization, their application in cultural heritage remains constrained by issues of scalability, data sovereignty, and the absence of adaptive, multi-institution intelligence. Most existing approaches rely on static or centralized digital models that are insufficient for the dynamic, collaborative, and ethically sensitive nature of heritage management. This dissertation aims to develop a comprehensive framework that advances DT technology from isolated digital replicas to adaptive, privacy-preserving, and cognitively enriched federated ecosystems. The research seeks to: (i) systematically evaluate the current state and maturity of DTs in cultural heritage; (ii) design multi-layered Federated and Cognitive Digital Twin architectures; and (iii) validate their ability to support decentralized intelligence, real-time emergency coordination, and knowledge-driven decision-making across multiple heritage institutions. The study employs a multi-stage methodology comprising: (1) a systematic review of 108 scholarly works to identify DT trends, enabling technologies, maturity levels,user-centered application and research gaps; (2) architectural design of federated frameworks, including an Intelligent FDT for distributed decision-making, and Cognitive Federated Digital Twin (CFDT) integrating semantic reasoning, reinforcement learning, and the MAPE-K loop; (3) simulation-based validation using agent-based modeling, federated learning (FedProx), and reinforcement learning with proximal policy optimization (PPO) in multi-museum evacuation scenarios; and (4) cross-framework synthesis and maturity assessment evaluating scalability, responsiveness, and cognitive adaptability. Findings from the systematic review highlight fragmented infrastructures, low DT maturity, and limited integration of AI, XR, or federated methods in heritage contexts. The proposed architectures demonstrate significant advancements: the Intelligent FDT framework improves emergency coordination with enhanced evacuation efficiency and reduced latency; and the CFDT provides stable learning, faster convergence, semantic explainability, and cross-institution knowledge transfer. Federated reinforcement learning yields higher cumulative rewards while reducing communication overhead and maintaining data sovereignty. This dissertation establishes Intelligent, Federated, and Cognitive Digital Twin as a novel paradigm for cultural heritage management one that is adaptive, collaborative, ethically aligned, and scalable. By integrating DTs with AI,and federated intelligence, the work enables a transition from static digital archives to dynamic, learning ecosystems capable of supporting predictive conservation, real-time emergency response, and long-term sustainability. The resulting frameworks offer foundational theory, validated methods, and practical pathways for deploying intelligent federated digital infrastructures across the global cultural heritage sector.