Digital-twin-driven real-time inverse modulation design for elastic optical networks

Elastic optical networks (EONs) exhibit time-varying traffic and nonlinear fibre impairments, demanding inverse modulation train agility to optimize spectral efficiency. Fixed modulation schemes or lookup-table based modulation cycles cannot efficiently leverage real-time conditions; while existing inverse modulation methods adapted via pre-design are typically link-adapted and offline. This paper proposes a digital-twin-driven, real-time inverse modulation design framework for EONs, comprising GN/EGN physical layer channel modeling, neural digital twins of physical channel response, and SDN-controlled, network-level online synthesis. In particular, the proposed system adaptively selects and synthesizes modulation formats in real time to maximize achievable information rate (AIR) through BER constraint. Numerical results show AIR gains, increased transmission distance, lower blocking probability, and millisecond adaptation latency relative to fix and offline inverse modulation methods. The system facilitates AI-native network implementation toward self-optimizing optical networks.