Communication-control collaborative atomic reconfiguration mechanism for Industrial Internet of Agents

To address the challenge of dynamically and atomically adjusting the cooperative relationship and communication topology among agents in the Internet of Agents (IoA) based on high-level intentions, this paper proposes an intention-driven collaborative control architecture that supports the atomic reconfiguration of "communication-control" collaboration. Firstly, a formal model of "cooperative state" is proposed, which encapsulates the control logic within agents and their external communication topology as a dynamically evolvable whole. Secondly, to meet the strict atomicity, real-time, and continuity industrial constraints during the "cooperative state" transition, an atomic cooperative state transition (ACST) mechanism based on double buffering is designed. This mechanism combines background reconfiguration with atomic switching to achieve microsecond-level, interruption-free synchronous evolution of the internal behavior of agents and their external network connections. Particularly, at the communication adaptation level, the architecture builds a protocol-driven and data cache pool within the controller that supports dynamic addition and deletion, and through an "incremental preheating, incremental delay" resource cleaning strategy, it ensures data consistency and timing determinacy during the multi-protocol network topology reconfiguration process. Physical platform experimental results show that the proposed architecture atomically switches the cooperative state within an extremely short delay (< 17.4µs) based on the cooperative intention triggered by visual sensors, and ultimately achieves a high-precision (synchronization error < 4.1ms) dynamic cooperative sorting task, demonstrating the outstanding performance of this architecture in on-demand construction of high-fidelity agent cooperative networks.