ABSTRACT: As data-intensive applications and services continue to proliferate, the existing optical infrastructure faces challenges in meeting growing bandwidth requirements. So far, communication technologies have been able to meet bandwidth demands. Nevertheless, researchers have anticipated a potential Capacity Crunch associated with these networks this decade. It refers to the fact that the transmission capacity limit in optical fibers is close to being reached soon. It is then urgent to evolve the current network architectures to satisfy the relentless growth in bandwidth demands.

Two complementary courses of action can face this threat. The first approach involves increasing network densification by deploying efficiently extra network resources. Two solutions have been proposed: Space Division Multiplexing (SDM) and spectrum expansion. SDM involves using multicore fibers that allow the transmission of multiple data streams through different spatial dimensions. Spectrum expansion entails leveraging advanced modulation schemes and increasing frequency bands in the optical spectrum. Both approaches aim to enhance the capacity of optical networks. The second approach focuses on improving resource management to utilize the existing network infrastructure efficiently. Traditional optical network management suffers from manual configurations, static operation, longer provisioning times, and higher operational costs. Network automation, dynamic resource allocation, and elastic optical networks (EON) have been proposed to address these issues. These solutions involve software-defined network platforms, intelligent automation, on-demand resource allocation, and adaptive bandwidth utilization.

Integrating artificial intelligence (AI) in optical network architectures with the previously entioned proposal offers the opportunity to develop intelligent optical networks. The deployment of such networks can lead to the next generation of high-speed and efficient optical communication systems that leverage digital intelligence and automation.
This research project aims to contribute to designing efficient smart optical networks by addressing
fundamental problems in optical network planning, focusing on the main features of smart optical networks: Reconfiguration capabilities, Programmability and Automation. In terms of reconfiguration capabilities, we will plan multiband elastic optical networks solving main network planning problems such as routing, resource dimensioning, balancing the bands’ usage, spectrum defragmentation protocols, and providing fault tolerance to any set of failure scenarios meanwhile the network remains connected.

Regarding programmability and automation, we will focus on integrating autonomous agents in current and future optical networks based on Artificial Intelligence most promising techniques, and the development of novel protocols harvesting interpretable machine learning techniques. In addition, we create an artificial intelligence framework for a fast and straightforward deployment of deep reinforcement learning or any other decision-making agent for autonomous resource management decisions. Remark that every solution aims to minimize the overall network capacity while guaranteeing a minimum quality of service to the connections.

Finally, we will include the planning for key future interesting applications held on optical architectures such as quantum internet services, such as Quantum Key Distribution, Quantum Clock Synchronization, Entangled Quantum Networks and Distributed Quantum Computing.

The outcomes of this project have significant economic, technical, and scientific implications since any improvement may change the optical network infrastructure and potentially pave the way for future advancements.