Among the many special issues dealing with wireless communications, it is refreshing to have a Special Issue on optical communication and networking. This article brings the perspectives of some of the Guest Editors, Alex Alvarado (AA), Marco Secondini (MS), Laurent Schmalen (LS), Marija Furdek (MF), and Junho Cho (JC), on several questions related to the role of optical communication in the next-generation communication infrastructure.

JSAC: Optical networks represent one of the most successful technologies that have brought the internet to its present, large-scale state. Which problems in optical communication and networking do you see to be requiring urgent solutions in order to support the future use and scale of the internet?

AA: I still think that there are quite a few areas in optical communications where a deeper understanding of the fundamentals of communication theory can help. An area that currently intrigues me is communications through free-space optics (FSO), where current solutions are based on simple incoherent (intensity-modulation, direct-detection). In such FSO links, the channel is strongly dependent on the weather conditions (atmospheric turbulence) as well as geometric spreads, pointing errors, etc. I believe that for such links to become a big success, they will have to move to coherent technologies, for which transponders will have to be re-designed with the FSO channel in mind.

MS:  Historically, optical communications have always lagged behind radio communications in terms of theoretical and system aspects. While advanced transmission techniques and codes were used in radio communications, optical communications mainly relied on IM/DD and simple codes, operating far away from the Shannon limit. Today, perspectives have changed radically. Consider, for instance, the meticulous attention given in long-distance fiber systems to optimizing constellation shaping to squeeze out the last fraction of available capacity. In this field, several fundamental problems are still open: the nature of the nonlinear fiber channel itself remains elusive; we do not know its capacity limits, nor the optimal modulation and detection strategies. Though extremely fascinating, these theoretical problems are not probably the most relevant ones from a practical perspective. In fact, scaling the capacity of optical networks in the long term requires drastic solutions, reasonably involving massive spatial parallelization and an increase in the utilized bandwidth. In this context, practical aspects and technological issues are the main factors. The ultimate goal is, very pragmatically, to scale capacity while minimizing the cost per bit: an elusive metric that requires difficult predictions on future technological developments and economic trends.

LS: Compared to mobile communication systems, optical networks are still in a relative technological infancy. There are still many fundamental problems to solve, especially related to the nonlinearities that occur during propagation. On top of that, the data rates reaching more than 1 Tbit/s require well-engineered solutions for digital signal processing and its VLSI implementation in future receivers. Furthermore, current optical communication systems are very rigid. In order to achieve a similar level of flexibility and interoperability as the mobile communication systems, we need radically new solutions from physical layer to network management.

MF: Along with the continuous need for capacity upgrade to match the increase in network traffic, solutions that offer ultra-high availability, resilience, and security are also urgently needed. On top of this, a network should operate in a programmable, flexible and automated way, ideally without requiring manual interventions.

JC: Optical communications allowed the internet and telecommunication networks to have the enormous capacity they have today. Behind this are several fundamental advantages of optical communications over other communications: (i) Lightwave offers enormous bandwidth unparalleled by radio- and microwave-frequency sources, (ii) It provides extremely stable connectivity and capacity through guided fiber channels compared to satellite or wireless communications that rely on atmospheric channels whose quality varies greatly depending on environmental factors such as weather or vegetation, (iii) Signal attenuation is several orders of magnitude smaller in optical fiber channels compared to copper channels, allowing signals to be transmitted much farther to cover the entire globe. In addition to these traditionally important advantages, with the rise of artificial intelligence (AI) and machine learning (ML) technologies that require unprecedented amounts of computation and communication under severe power constraints, the advantage in high power efficiency is now calling for the expansion of optical communications into chip-to-chip, and intra-, inter-data center communications. Urgent solutions are needed for power-efficient optical communications for the future scale of the internet.

JSAC: In the wireless community the general topic that attracts the most attention is the 6G technology, which will bring increased fusion between the digital and physical world. Which role do you see for the optical networks in the evolving wireless network infrastructure?

AA: I believe that the proliferation of ultra high speed optical pluggables operating at 800Gbps, 1.6Tbps and even 3.2Tbps will have a great impact on future networks. The combination of those pluggables in the context of for example FSO links for both terrestrial (horizontal) and space (vertical) links is also very interesting.

LS: We can already see a convergence of wireless and optical communications from a technical viewpoint. With the use of terahertz communications in 6G, this convergence will continue and we will see the emergence of free-space terahertz links that are used, e.g., for high-speed data transfer and connection of micro/nano/pico-cells. The 6G network also requires a flexible fiber infrastructure to connect, e.g., distributed base stations.

MF: 6G requires innovation across all network domains and a high degree of confluence of different technologies. An example of such efforts is combining photonic radio fixed wireless and free space optical transmission for fixed wireless connections to enable creating an edge mesh network. On the metro/core network side, high-capacity and high-availability, secure fiber links remain crucial for converged end-to-end communications relying on advanced monitoring and slice-aware control protocols.  

JSAC: Machine Learning (ML) and Artificial Intelligence (AI) are pervading all domaines, including the engineering research. What do you see that the ML/AI can do for the optical networking and, vice versa, what can optical networking do for the (distributed) AI/ML?

AA: I believe that ML algorithms will eventually become integrated in some subsystems. I do not believe that they will completely replace a full receiver chain, but I do think hybrid solutions will at some point become the norm.

MS: I think ML will play a crucial role in all areas where complex design and optimization problems are found. This includes, for instance, network design and operation, monitoring and fault localization and classification. It is perhaps harder to find specific domains where ML will find less space. In terms of digital signal processing, for instance, there are many classical algorithms that are already simple, effective, and adaptive. Many of them will likely survive as they are. As for the other part of the question, I consulted ChatGPT, which runs on a remote server located across the ocean, and received the response via optical fibers…

LS: ML/AI has become an indispensable optimization tool for many problems in the design of the algorithms in optical communications. Furthermore, ML/AI is and will be used for network monitoring, ânomaly detection, and network reconfiguration. On the other hand, large AI/ML models are currently trained on huge clusters with many servers and many GPUs that need a strong interconnection network. This network can only be realized with optical fibers and requires new solutions with extremely low latency.

MF: ML/AI is key for advanced monitoring of the optical fibre network. Many external changes, including attempts of tampering with the network infrastructure, failures or natural disasters cause intricate changes in optical signal parameters which cannot be detected by humans, but can be observed by ML/AI. ML/AI also enables advanced and autonomous resource provisioning, QoT estimation, and network adaptation policies.

JC: ML/AI requires unprecedented scale of computation and communication, even requiring construction of so-called hyperscale data centers. Much of the data transmitted within or between hyperscale data centers follows the internet protocol (IP) and must be standard-compliant to pass through Ethernet switches and optical transceiver modules. Accordingly, extensive efforts are currently being made to extend standards and implementation agreements (IAs) to accommodate the explosive growth in IP traffic originating from ML/AI, such as IEEE 802.3 or the Optical Internetworking Forum (OIF)’s various IAs. To this end, optical communications research is committed to designing optimal solutions to meet the emerging requirements for ML/AI, including for example the forward error correction and modulation solutions and proposing them to standardization bodies.