ResilM-IBN: multi-agent framework for self-healing network management

To address the challenges of inaccurate intent parsing, uncontrollable execution chains, and the lack of self-healing mechanisms in intent-based networking under multi-service environments, ResilM-IBN, a self-healing multi-agent management framework for software-defined network, aiming to enhance network intelligence and robustness, was proposed. Through structured communication and multi-agent collaboration, the framework ensures semantic consistency and task traceability throughout the intent execution process. A serial closed-loop execution process following the “intent-verification-execution-re-verification” workflow was constructed to improve controllability and reliability during intent realization. Furthermore, a collaborative self-healing mechanism was designed, where cross-agent anomaly detection and automated recovery were triggered by a dual-condition logic combining user-intent constraints and runtime feedback. Experimental results obtained from a Mininet-Ryu prototype demonstrate that ResilM-IBN can achieve stable and reliable intent execution and self-healing recovery in complex scenarios. The proposed framework exhibits strong feasibility, stability, and portability, providing a practical reference for the engineering implementation of intent-based networking.