An Operational Architecture for Self-Evolving Cognitive Digital Twins

Summary

This paper proposes an operational architecture for Cognitive Digital Twins (CDTs), which extend digital twins from state synchronization toward task-oriented, knowledge-driven operation. The architecture has four layers: physical, digital-twin, cognitive, and task. Together they form a closed operational loop in which physical states are synchronized into digital representations, while the cognitive layer builds task-specific models using knowledge, memory, and attention. The task layer then produces decisions subject to practical constraints, and operational feedback updates cognitive experience as well as relationships and annotations in the digital representation. The paper distinguishes user-request-driven cognition from self-driven cognition. It also identifies semantic communication, knowledge querying, task orchestration, and closed-loop synchronization as key enabling mechanisms and deployment challenges. A lightweight simulation suggests reliable task feasibility under limited semantic information and improved operational efficiency as task experience accumulates. The framework is presented as a structured basis for designing future CDT systems, rather than as a finished implementation or a new foundation model.