A 4D reconstruction of ancient Troy with AI-assisted interpretive interaction

This paper presents a phase-based, real-time four-dimensional (4D) reconstruction of ancient Troy, enhanced by artificial intelligence (AI)-assisted interpretive interaction with selected virtual characters. Defined by nine superimposed settlement layers, Troy presents particular challenges for conventional two-dimensional (2D) documentation and static three-dimensional (3D) representations, especially when communicating long-term urban transformation to specialist and non-specialist audiences. To address this challenge, all nine principal phases were reconstructed as spatially aligned but separately managed layers within a single Unreal Engine 5 environment. Excavation records, stratigraphic documentation, published archaeological interpretations, and architectural reference drawings were converted into evidence-based 3D models in Autodesk 3ds Max and optimized for real-time visualization. The interpretive component uses retrieval-grounded generative conversational AI, implemented through Convai, for selected phase-aware non-player characters (NPCs). Each character is linked to curated, phase-specific historical knowledge and constrained by a defined speaking scope, contextual prompts, and response rules. The AI was not used to generate archaeological geometry, determine architectural form, or create unrestricted historical narratives. Instead, it supports contextual explanation, visitor orientation, educational mediation, and phase-specific readability while remaining subordinate to the archaeological evidence. The workflow also separates documented, inferred, and speculative elements so that uncertainty can be communicated without presenting every visual decision as equally certain. The project demonstrates that a phase-based real-time environment, combined with carefully bounded conversational AI and virtual reality (VR) interaction, can improve comparative understanding of a multi-layer archaeological site while maintaining archaeological traceability, technical performance, and interpretative transparency. Furthermore, the proposed workflow establishes a reproducible framework for integrating archaeological documentation, immersive visualization, and AI-assisted interpretation within a single interactive environment. By allowing users to examine each settlement phase independently while preserving their spatial correspondence, the system facilitates clearer observation of architectural continuity and transformation over time while supporting transparent scholarly communication, heritage education, future comparative research, and informed public engagement.