KDN-Driven zero-shot learning for intelligent self-healing in 6G small cell networks

6G cellular networks represent a significant advancement in wireless communication by offering higher data rates, lower latency, and improved reliability. Ultra-dense small cell deployment is vital for providing enhanced capabilities in 6G, which in turn facilitates the development of data-intensive applications. Despite the benefits, dense small-cell deployments can also significantly increase anomaly rates. The complexity arising from the dynamic, dense, and data-intensive environment of 6G networks presents a significant challenge for anomaly detection and resolution. To address this issue, this paper proposes a Knowledge-Defined Networking (KDN)-enabled intelligent self-healing approach for 6G small cell networks, based on zero-shot learning. Our system continuously monitors network metrics and collects data. It utilises a semantic zero-shot learning model to detect anomalies, including new and previously unseen ones, without requiring retraining. When an anomaly is detected, the system analyses it using historical data and predefined rules to find the root cause. Once the root cause is identified, the system executes self-healing actions to resolve the issue in a closed-loop manner. The proposed system operates in an AI-native and zero-touch fashion, aligning with key 6G goals. It is evaluated in a simulation environment configured with realistic 6G parameters, including mmWave frequency (28 GHz), massive MIMO, and energy-aware small cells. The performance results underline that the proposed scheme achieves lower packet loss and reduced latency compared to conventional healing approaches. These results confirm that the architecture supports scalable, autonomous, and real-time fault management for future 6G infrastructures.