TrustOPT: a trusted online optimization strategy for reliable autonomous optical networks with field-trial demonstration

Optical power optimization is essential for improving the quality of transmission (QoT) in autonomous optical networks (AONs). However, online optical power optimization is challenging due to the potential performance degradation and even service disruption caused by exploratory adjustments. In this paper, we propose TrustOPT, a trusted online optimization strategy for reliable optical power optimization in AONs. By leveraging the Lipschitz constant (L-constant), TrustOPT defines an evolving trusted configuration space, ensuring that the QoT can remain above the required threshold even under worst-case degradation after each adjustment. Within the trusted space, advanced optimization algorithms are employed to efficiently optimize optical amplifier (OA) configurations for improved QoT. TrustOPT is validated on a field-deployed testbed with live services. Field-trial results demonstrate that TrustOPT achieves optimal optical power while maintaining a 4 dB Q-factor margin, even in a challenging scenario with only a 0.15 dB initial Q-factor margin. Compared with established optimization strategies, TrustOPT achieves superior performance in both optimization results and processes. Specifically, in terms of worst-case Q-factor during optimization, TrustOPT outperforms Bayesian optimization by over 4.73 dB, significantly enhancing the optimization reliability. Moreover, a strong agreement between the estimated and field-measured L-constants is observed, further verifying the reliability and practical effectiveness of the proposed approach. TrustOPT thus provides a robust and practical solution for achieving reliable online optical power optimization in future AONs.