Optical Transport Networks (OTN) has emerged as a key for transferring data through fiber optic cables, but its deployment is not an easy task. The researchers from the Polytechnic University of Catalonia and the telecom company Huawei have tried to rebuild the AI-driven technique to make OTNs run more efficiently than earlier and to solve some common problems experienced.
The researchers’ new approach commonly combines two machine learning techniques. One is reinforcement learning, means it creates a virtual agent that can learn through trial and errors of the system to reform managed resources. And another one is deep learning, means it adds a layer of sophistication to reinforcement-based approach by using a neural network. At the end of this research, it was estimated that this new work leveraging machine learning greatly increased the efficiency of optical telecommunication networks.
Still, the most advanced deep reinforcement learning algorithm can optimize resource allocation in OTN, but they fail when run into any unexpected circumstances. The researchers worked to conquer this common problem.
After learning through 5000 rounds of simulation, the deep reinforcement learning agents direct traffic with 30 percent greater efficiency than the existing up-to-date algorithm. This new approach is set to learn more about the network. The deep reinforcement learning agent can quickly learn to optimize the network without any prior knowledge. It results in optimization approaches, which outperform expert algorithms.
It was estimated that the researchers have planned to employ deep reinforcement strategies with graph networks to transform scientific and industrial fields, which include computer networks, chemistry, and logistics.