A learning-based dynamic routing and spectrum allocation algorithm in satellite elastic optical networks

A learning-based dynamic routing and spectrum allocation algorithm (L-DRSA) is proposed for the satellite elastic optical network. In the routing selection stage, L-DRSA selects the routing path with the highest long term benefits based on the deep reinforcement learning algorithm. In the spectrum allocation stage, the spectrum is allocated greedily with the aim of minimizing the spectrum fragmentation of the network. Finally, several simulation experiments were conducted to evaluate the performance of the L-DRSA and the results demonstrate that our proposed algorithm outperforms than the baseline algorithms.