“AI/ML for 6G MIMO: From fundamentals to industrialization”
Date, hour and room to be defined
Speakers:
- Luca Sanguinetti (University of Pisa and CNIT, IT)
- Adrian Garcia-Rodriguez (Ericsson Research, )
- Tanguy Kerdoncuff (Ericsson Research, FR)
- Renato Luís Garrido Cavalcante (Fraunhofer Institute for Telecommunications, DE)
- Paul Almasan (Telefónica Innovación Digital, ES)
Motivation and Context
The advent of 6G marks a pivotal moment in the evolution of mobile communications, with artificial intelligence and machine learning (AI/ML) expected to play integral roles in the design, optimization, and operation of the future air interface. As 3GPP has just initiated the first studies on 6G technologies, the community stands at a unique juncture: the lessons learned during 5GAdvanced—particularly on AI-enabled air interface procedures—provide valuable foundations, yet the scalability, reliability, and interpretability requirements of 6G MIMO bring an entirely new set of challenges.
Massive MIMO will remain a cornerstone of radio access in 6G, but its increasing dimensionality, the push for extreme energy efficiency, and the ambition to support highly dynamic and heterogeneous deployments call for solutions that go beyond classical signal processing. AI/ML promises substantial gains in areas such as reference signal design, beam management, or hardware impairment compensation. However, turning this promise into reality requires addressing key obstacles related to dataset generation, model generalization, real-time feasibility, robustness, and standardization.
This tutorial aims to provide clarity on these opportunities and challenges by combining fundamental theory, latest 3GPP insights, and hands-on industrial perspectives. Building on the presenters’ contributions to the MIMO fundamentals field and 3GPP RAN1 6G studies, we will highlight where AI/ML has demonstrated tangible potential, where open questions remain, and which research directions we believe are most critical for shaping the 6G MIMO landscape.
Given the strong momentum within 3GPP, the ongoing wave of academic activity, and the strategic importance of AI-native radio designs, this topic is exceptionally timely for the 2026 EuCNC & 6G Summit.
Structure and Content
The tutorial is structured in two complementary parts, providing a cohesive industrial–academic perspective on 6G MIMO and the emerging role of AI/ML in next-generation radio systems. The first part introduces the distinguishing features, design principles, and key challenges of 6G MIMO relative to previous cellular generations. It also offers a clear and intuitive overview of the AI/ML algorithms most relevant to 6G MIMO, laying the theoretical groundwork for the second part of the tutorial. The second part explores how AI/ML is expected to transform 6G MIMO design. It presents the latest 3GPP updates and highlights the most promising use cases, ongoing studies, and practical considerations driving the integration of AI-native solutions into the air interface.
:: Part 1 – Fundamentals
Session 1: Introduction to 6G MIMO – Luca Sanguinetti (45 min)
1.1 Fundamentals (30 min): Distinguishing features, design principles, and key challenges.
1.2 Motivation (15 min): AI/ML integration in 6G MIMO.
Session 2: AI/ML Algorithms for 6G MIMO – Renato L. G. Cavalcante (45 min)
2.1 Domain knowledge (10 min): Hybrid data-driven + model-based design, role of classical ML.
2.2 Hybrid approaches (10 min): Spectrum from model-based to data-driven; fixed-point theory as underlying structure.
2.3 Examples (25 min): Academic cases combining model-based methods with neural networks or kernel methods; highlighting measurable gains of hybrids.
:: Part 2 – Applications
Session 3: AI/ML in 6G MIMO – Adrian Garcia Rodriguez & Tanguy Kerdoncuff (45 min)
3.1 3GPP 6G standardization and RAN1 studies (10 min)
3.2 Latest 3GPP updates on AI/ML for MIMO (10 min)
3.3 Key AI/ML use cases: AI/ML receivers, channel prediction, reference signal reduction (25 min)
Session 4: AI/ML Applications – RAN Digital Twins & CSI Estimation – Paul Almasan (45 min)
4.1 CSI Estimation (15 min): Problem, existing solutions, use cases
4.2 Radio Map & Coverage Prediction (15 min): CNN-, GNN-, geometry-aware approaches, network planning
4.3 Open Challenges (10 min): Generalization, robustness, deployment, data efficiency, interpretability, physics-aware design
