Skip to content

Papers

XLRON is a research project. The features documented on this site come from the following papers — each links to a step-by-step reproduction guide (where applicable) that walks through the exact commands used to generate every figure.

Published

XLRON: Accelerated Reinforcement Learning Environments for Optical Networks

Doherty, M., Beghelli, A. — OFC 2024.

The original announcement of XLRON, focused on the JAX implementation, GPU parallelism, and the DeepRMSA reproduction. PDF: ofc_paper.pdf.

@INPROCEEDINGS{doherty_xlron_2024,
  author    = {Doherty, Michael and Beghelli, Alejandra},
  booktitle = {2024 Optical Fiber Communications Conference and Exhibition (OFC)},
  title     = {{XLRON}: Accelerated Reinforcement Learning Environments for Optical Networks},
  year      = {2024},
  pages     = {1-3}
}

Reinforcement Learning with Graph Attention for Routing and Wavelength Assignment with Lightpath Reuse

Doherty, M., Beghelli, A. — Optical Network Design and Modelling (ONDM) 2025. Preprint: arXiv:2502.14741.

A Graph-Attention RL agent for the RWA-with-Lightpath-Reuse formulation: the agent must balance routing decisions against capacity packing on already-established lightpaths. Trained on NSFNET via the rwa_lightpath_reuse environment using XLRON's GNN actor-critic.

→ Reproduce all figures

@inproceedings{doherty_rwalr_2025,
  author    = {Doherty, Michael and Beghelli, Alejandra},
  title     = {Reinforcement Learning with Graph Attention for Routing and Wavelength Assignment with Lightpath Reuse},
  booktitle = {Optical Network Design and Modelling (ONDM)},
  year      = {2025},
  note      = {arXiv:2502.14741},
  url       = {https://arxiv.org/abs/2502.14741}
}

Reinforcement Learning for Dynamic Resource Allocation in Optical Networks: Hype or Hope?

Doherty, M., Beghelli, A. — JOCN 17(9), D1 (2025) — DOI: 10.1364/JOCN.559990 · arXiv:2406.01919

A systematic survey of ~100 RL-for-RMSA papers. Reproduces five highly-cited published RL solutions in matched simulation settings and shows that simple KSP-FF / FF-KSP heuristics with sufficient candidate paths and an appropriate path-sort criterion match or beat them. Also introduces a defragmentation-based capacity bound estimator (reconfigurable routing) and uses cut-set bounds as an upper-bound capacity reference.

→ Reproduce all figures

@article{doherty_reinforcement_2025,
  author  = {Doherty, Michael and Beghelli, Alejandra},
  title   = {Reinforcement Learning for Dynamic Resource Allocation in Optical Networks: Hype or Hope?},
  journal = {Journal of Optical Communications and Networking},
  volume  = {17},
  number  = {9},
  pages   = {D1},
  year    = {2025},
  doi     = {10.1364/JOCN.559990},
  url     = {https://opg.optica.org/abstract.cfm?URI=jocn-17-9-D1}
}

Submitted / pending publication

Comparison of Dynamic Elastic Optical Network Capacity Bound Estimation Methods

Doherty, M.*, Deng, B.*, Beghelli, A., Toni, L., Bayvel, P. — Submitted to ECOC 2026. *Equal contribution.

A systematic comparison of three capacity bound estimation methods — cut-set (CS), resource-prioritized defragmentation (RPD), and KSP heuristics — across the 118 real-world topologies of TopologyBench. Median gap over KSP at 0.1 % blocking is 5.8–9.4 % for CS (top-100 / top-10 cuts) and 17.0 % for RPD, indicating that spectrum fragmentation is a larger source of lost capacity than routing inefficiency.

→ Reproduce all figures

@unpublished{doherty_deng_capacity_bounds_2026,
  author  = {Doherty, Michael and Deng, Bohua and Beghelli, Alejandra and Toni, Laura and Bayvel, Polina},
  title   = {Comparison of Dynamic Elastic Optical Network Capacity Bound Estimation Methods},
  note    = {Submitted to ECOC 2026},
  year    = {2026}
}

In preparation (not yet published)

XLRON: A Framework for Hardware-Accelerated and Differentiable Simulation of Optical Networks

Doherty, M., Beghelli, A., Jarmolovičius, M., Deng, B., Killey, R., Bayvel, P., Toni, L. — In preparation, targeted at JOCN.

Introduces XLRON's architecture, benchmarks it against seven other open-source simulation libraries, validates the ISRS GN + DRA physical layer against the Gerard et al. 2025 C+L-band experiment to within 0.5 dB, demonstrates the differentiable-simulation capabilities (gradient-based pump optimisation, direct RSA optimisation), and describes the GUI / CLI.

→ Reproduce all figures

@unpublished{doherty_xlron_framework,
  author  = {Doherty, Michael and Beghelli, Alejandra and Jarmolovi\v{c}ius, Mindaugas and Deng, Bohua and Killey, Robert and Bayvel, Polina and Toni, Laura},
  title   = {{XLRON}: A Framework for Hardware-Accelerated and Differentiable Simulation of Optical Networks},
  note    = {In preparation},
  year    = {2026}
}

Graph Transformers and Stabilized Reinforcement Learning for Large-Scale Dynamic Routing, Modulation and Spectrum Allocation in Elastic Optical Networks

Doherty, M., Beghelli, A., Toni, L. — In preparation, targeted at JOCN.

Presents the first transformer architecture trained from scratch with RL for dynamic RMSA. Combines off-policy invalid action masking, a valid-mass log-barrier loss, per-step damping, hard gating, Pre-LayerNorm, and Wavelet-Induced Rotary Encodings (WiRE) to inject graph structure. Beats the strongest heuristic on every standard benchmark (NSFNET, COST239, USNET, JPN48 — DeepRMSA / RewardRMSA / GCN-RMSA / MaskRSA / PtrNet-RSA settings), then scales to USA100 (100 nodes) and TataInd (143 nodes) — the largest dynamic RMSA instances ever attempted with RL.

→ Reproduce all figures · Highlights

@unpublished{doherty_graph_transformer,
  author  = {Doherty, Michael and Beghelli, Alejandra and Toni, Laura},
  title   = {Graph Transformers and Stabilized Reinforcement Learning for Large-Scale Dynamic Routing, Modulation and Spectrum Allocation in Elastic Optical Networks},
  note    = {In preparation},
  year    = {2026}
}
  • TopologyBench — 119 real-world optical network topologies, all bundled with XLRON. See Topologies.
  • Gymnax — JAX gym-style API; XLRON's environment interface follows the same pattern.
  • PureJaxRL — XLRON's PPO implementation derives from PureJaxRL.
  • Stoix — multi-device sharding inspiration.
  • GNPy — vendor-agnostic optical-network planning, complementary to XLRON for multi-vendor production planning.
  • Optical-Networking-Gym — Cython-based RMSA simulation library, benchmarked against XLRON in the XLRON framework paper above.

Acknowledgements

This work was supported by the EPSRC grant EP/S022139/1 (Centre for Doctoral Training in Connected Electronic and Photonic Systems) and EPSRC Programme Grant TRANSNET (EP/R035342/1).