1st Reinforcement Learning Applied to Networking (RLAN) workshop

Abstract

Reinforcement Learning (RL) has achieved remarkable progress in recent years, from mastering complex games to enabling adaptive decision-making in robotics and industrial systems. However, transferring these successes into real-world domains such as computer networking presents unique challenges. This talk will reflect on key lessons learned from the broader RL community regarding sample efficiency, generalization, safety, and scalability—factors that are critical when deploying RL beyond controlled benchmarks. Promising research directions will also be discussed, such as hierarchical decision-making, offline RL, and safe exploration, and how these approaches may inform the development of intelligent and reliable networking systems. By connecting theoretical insights with practical deployment considerations, the talk will highlight opportunities for collaboration between the RL and networking communities, paving the way toward robust, real-world impact.

Speaker Bio

Marcello Restelli is a Full Professor at the Department of Electronics, Information and Bioengineering at Politecnico di Milano, where he coordinates the Real-Life Reinforcement Learning Research Lab (RL3). He is the author of more than 200 international scientific publications, primarily focused on the study and development of new reinforcement learning techniques. His research results are applied to real-world problems through numerous industrial collaborations in diverse sectors, including finance, e-commerce, Industry 4.0, and automotive. He is an ELLIS Fellow and serves as the research lead for the Artificial Intelligence Observatory of Politecnico di Milano. In 2020, he co-founded ML cube, a spin-off of Politecnico di Milano, where he is currently scientific advisor.

Abstract

Reinforcement learning (RL) is a promising technology for automating and enhancing network operations. However, practical constraints are often overlooked when applying it to real network operations. This keynote will first point out three challenges: (1) safety, (2) the combinatorial explosion of the action space, and (3) control and actuation latency from decision to application in production networks from an industrial research perspective. Then, several methods proposed by NTT’s R&D will be introduced: safety assurance through constrained RL, action space reduction using cooperative multi-agents, and delay-aware approaches for use cases such as fault recovery, VNF placement, task offloading optimization, and auto-scaling of core network functions. Finally, the talk will explain NTT's research toward autonomous network operations.

Speaker Bio

Yoichi Matsuo received his M.E.and Ph.D. in applied mathematics from Keio University in 2012 and 2015.

He is currently a researcher at the Network Service Systems Labs., NTT, Inc., Tokyo, Japan.

Since he joined NTT, he has been engaged in research on network management using AI technologies such as deep learning and reinforcement learning. Specifically, his interests lie in researching and integrating AI technologies into real-world network operations.