Artificial Intelligence in Wireless Communication: Paving the Road to 6G

AI
July 2, 2025

As we stand at the threshold of 6G development, artificial intelligence (AI) and machine learning (ML) are emerging as critical enablers of the next generation of wireless communications. While today’s wireless networks benefit from algorithm-driven optimizations, future 6G systems will be AI-native by design, leveraging data-driven intelligence across every layer of the communication stack.

From Weak AI to AI-Native Radios

We currently live in the age of weak AI, characterized by systems designed for specific tasks such as:

  1. Logical reasoning – e.g., AlphaGo
  2. Perception – e.g., facial recognition
  3. Knowledge representation – e.g., IBM Watson for Oncology
  4. Language processing – e.g., Apple Siri, Amazon Alexa
  5. Planning and navigation – e.g., self-driving cars

By contrast, strong AI envisions machines matching or surpassing human intelligence. While strong AI remains on the horizon, machine learning, particularly neural networks, has already started transforming wireless systems.

Neural Networks in Wireless Communication

Neural networks—a core branch of ML—are reshaping wireless technologies. Three architectures are particularly promising:

  • Recurrent Neural Networks (RNNs): Effective in modeling time sequences. RNNs are applied in RF front-end linearization, leveraging memory effects for digital pre- and post-distortion.
  • Convolutional Neural Networks (CNNs): Originally built for image recognition, CNNs are now being explored as neural receivers capable of decoding radio signals more efficiently than traditional methods.
  • Autoencoders: These unsupervised learning models compress data by filtering out redundant information. They are being used to reduce channel state information (CSI) feedback, a crucial element in maintaining wireless link quality.

AI’s Role in 6G Innovation Areas

AI and ML are not isolated elements in 6G—they permeate every major 6G research domain, including:

  • Terahertz (THz) communication
  • Integrated sensing and communication (ISAC)
  • Reconfigurable intelligent surfaces (RIS)
  • Cell-free massive MIMO
  • Full-duplex communications

AI models can optimize all of these, enhancing energy efficiency, link reliability, and spectrum utilization.

The Neural Receiver and the AI-Native Air Interface

The goal for 6G is to build an AI-native air interface by replacing conventional physical-layer signal processing blocks (e.g., channel estimation, equalization, demapping) with trained ML models. These combined tasks form a neural receiver, which learns to decode signals more effectively by adapting to environmental conditions and hardware imperfections.

ML in the RF Front-End: A Three-Phase Evolution

Phase 1: ML-Based Linearization of RF Components

Machine learning replaces traditional linearization techniques for power amplifiers and transceivers. Since a single vendor often designs these components, access to training data is straightforward, making this phase more feasible.

Phase 2: Replacing Receiver Signal Processing

Signal processing tasks like channel estimation and demodulation are consolidated into a single ML model. This simplifies receiver architecture and allows for more dynamic adaptability in real-world conditions.

Phase 3: End-to-End Learning and Optimization

End-to-end (E2E) learning aims to co-optimize transmission and reception across the network based on use case—whether that’s voice, video, web, or XR. Custom modulation schemes, learned from real-world conditions, replace conventional static constellations. This enables pilotless transmission, further improving spectral efficiency.

Challenges: Data, Interoperability, and Lifecycle Management

Despite the promise, integrating AI into wireless networks presents several challenges:

  • Data Access: Training AI models requires high-quality datasets—often proprietary or hard to obtain.
  • Robustness: AI models must operate reliably under extreme or rare conditions.
  • Lifecycle Management: Ongoing training, monitoring, and coordination between device and network are essential to ensure consistent performance.
  • Interoperability Testing: Devices from multiple vendors must function seamlessly within an AI-enhanced ecosystem.

Rohde & Schwarz and the Future of AI in 6G

Rohde & Schwarz is at the forefront of 6G AI research, actively contributing to initiatives like 6G-ANNA (6G – Access, Network of Networks, Automation & Simplification). This lighthouse project aims to deliver a full end-to-end 6G architecture that merges human, machine, and environmental interaction through advanced sensors and AI-driven simplification.

The company is also involved in defining test and measurement standards to verify AI/ML performance across the network and device lifecycle.

Watch the On-Demand Webinar: Will AI/ML Revolutionize 6G?

Topics include:

  • The current status of AI/ML research in 3GPP Release 18
  • The architecture of an AI-native air interface
  • The neural receiver: how it outperforms traditional signal processing

Conclusion: AI as the Cornerstone of 6G

The shift toward AI-native wireless systems marks a profound transformation in how networks are designed, optimized, and operated. By embedding intelligence at every layer—from the antenna to the core—6G will not just be faster and more capable; it will be smarter, more efficient, and adaptive by design.

As we move into this new era, continuous innovation, collaborative research, and robust testing will be the pillars of successful deployment.

6GaIAI in Wireless Communication

Leave a Comment