Digital twin technology provides a new theoretical path for solving the difficulties of heterogeneous terminal interconnection and low control efficiency in the Internet of Things. Starting from the endogenous logic of empowering IoT terminals with digital twins, this article analyzes the mirror mapping and closed-loop interaction mechanism between physical entities and virtual models, achieving a paradigm leap from data synchronization to semantic collaboration. On this basis, a digital twin collaboration architecture for heterogeneous terminal linkage was constructed, and core issues such as unified twin model construction, virtual real dynamic association maintenance, twin inter collaboration triggering, and cross protocol communication decoupling were discussed. Then focus on the key technologies of intelligent management and control based on digital twins, including real-time data stream processing and edge computing integration, terminal state autonomous awareness and exception diagnosis, policy adaptive generation under the centralized-decentralized hybrid architecture, and the management and control evolution path from rule-driven to target-driven. The research provides a systematic theoretical framework for intelligent linkage and efficient control of large-scale heterogeneous IoT terminals.