This paper proposes an intelligent interaction framework oriented towards physical space and object management. Relying on a dual-layer geometric space structure, a unique address identification system, and a standardized task parsing mechanism, the framework realizes precise restoration of physical space, flexible computation in a virtual layer, full-lifecycle object tracking, and closed-loop execution of intelligent tasks. The architecture distinguishes between an entity precision layer and a virtual overlay layer: the entity layer maintains a fixed normal baseline and updates memory synchronously after renovations; for known venues and outdoor spaces, it adapts to database retrieval or an incremental "walk-and-record" logic, retaining only visual-range information in large scenes. Concurrently, it employs a 1:1 real-time placeholder rendering mode for dynamic humans and animals, adhering strictly to a WYSIWYG, non-predictive operational rule. Complemented by a spatial memory database and an authorized storage mechanism for privacy-related interactions, the overall architecture is highly minimalist, closed-loop, and legally compliant, suitable for various scenarios including object modification, spatial statistics, and intelligent execution.
Yan (Violet) Huang (Sat,) studied this question.