The research proposes a new architecture for Cognitive Digital Twins (CDTs) that moves beyond state synchronization. The architecture consists of four layers: physical, digital-twin, cognitive, and task. This layered approach establishes a self-evolving operational loop. Physical states are synchronized into digital representations. The cognitive layer constructs task-specific models using knowledge, memory, and attention. Task-level decisions are generated considering practical constraints. Operational feedback refines the cognitive experience and updates relationships within the digital representation.
The framework characterizes two operation modes: user-request-driven cognition and self-driven cognition. Key enabling mechanisms include semantic communication, knowledge querying, task orchestration, and closed-loop synchronization. A simulation study demonstrates reliable closed-loop task feasibility with limited semantic information and improved operational efficiency through accumulated task experience.
This architecture provides a structured foundation for future CDT systems. The research addresses deployment challenges associated with the identified mechanisms. The architecture’s design facilitates continuous adaptation and improvement within the digital twin.
Source: https://arxiv.org/abs/2609.09625