In the following years, we will see the advent of many new applications and use-cases such as the metaverse, the adoption of XR/VR, holographic telepresence, the Internet of the Senses, the consolidation of the Internet of Things, with autonomous robots, fully automated industries and manufacturing plants, as well as smart infrastructures and environments, to mention just a few. To satisfy their strict and high requirements —in terms of throughput, latency, reliability, connectivity, and power consumption— wireless networks —and their radio interface in particular— are becoming exceedingly complex, with a plethora of advanced communication features, protocols and parameters, usually involving nonlinear dependencies between them. To deal with such complexity, the use of Artificial Intelligence and Machine Learning (AI/ML) techniques—and their ability to deal with complexity in general—is the necessary performance enabler for next-generation wireless networks.

In this project, we aim to build a new, clean-slate AI/ML-Driven Radio (MLDR) interface. This new MLDR interface will learn to communicate by selecting and configuring the set of communication protocols and functionalities that better suit every particular use-case and scenario, thus satisfying the aforementioned hard performance requirements and efficiently using the available spectrum resources. While the project proposal is groundbreaking in terms of focus and goals, we will follow a standard research approach to reach the stated objectives, i.e., we will move from use-cases, concepts/specifications and design, to implementation, evaluation and analysis. The consortium includes four partners, all working at the intersection of wireless networks and AI/ML areas, with complementary expertise. During the MLDR design and evaluation process, we will generate new knowledge in the form of new ideas, theories, practical solutions, ML algorithms, and disruptive communication functions. We expect the results from this project will guide the design of future AI/ML-driven wireless communications and networks, becoming a reference to follow and compare with.

To achieve the overall project objective of developing the MLDR concept introduced above, we target the four specific objectives described below. We also introduce the consortium background and the previous results of each partner, as well as the considered TRLs at the end.

  • Objective 1: Define use-cases (scenarios, technologies, requirements, and KPIs)
  • Objective 2: Design a native AI/ML-Driven Radio (MLDR) Interface
  • Objective 3: From protocols to ML models: re-thinking communication functionalities 
  • Objective 4: Performance evaluation / proof-of-concept

Publications

  • Wilhelmi, Francesc, Szymon Szott, Katarzyna Kosek-Szott, and Boris Bellalta. "Machine Learning & Wi-Fi: Unveiling the Path Towards AI/ML-Native IEEE 802.11 Networks." arXiv preprint arXiv:2405.11504 (2024) - Accepted in IEEE Communications Magazine. [Link] [Objective 1]
  • Bellalta, Boris, Katarzyna Kosek-Szott, Szymon Szott, and Francesc Wilhelmi. "Towards an AI/ML-defined Radio for Wi-Fi: Overview, Challenges, and Roadmap." arXiv preprint arXiv:2405.12675 (2024). [Link] [Objectives 1 & 2] - Paper published as part of the IEEE International Network Generations Roadmap (INGR) - AI/ML Working Group.

Consortium

   

 

Funding Agencies:

MLDR is a chist-era (2022 call) project: https://www.chistera.eu/projects/mldr

Home

 

Logo AKAANR logoLogo NCNState Research Agency

Contacts

Boris Bellalta (UPF, Project coordinator)

Richard Combes (Supélec)

Nandana Rajatheva (UOulu)

Szymon Szott (AGH)

 

UPF address

Department of Engineering and Information and Communication Technologies

Tallers area (Poblenou campus)
Roc Boronat, 138
08018 Barcelona