AI-defined Native Air Interface
- (Harvard University - Joyce Yang)
- Overview
An AI/ML-defined native air interface is a key component of 6G networking that uses artificial intelligence (AI) and machine learning (ML) to make radios more dynamic and intelligent. This interface allows radios to learn from their environments and each other, and to adapt their signaling schemes to achieve the best performance in any situation.
To create an AI-native air interface, researchers replace blocks in the physical layer's signal processing chain with trained ML models. For example, tasks like channel estimation, equalization, and demapping can be combined into a single trained ML model called a neural receiver.
Some advantages of an AI/ML-defined native air interface include:
- Improved performance: Learning radios can create custom waveforms, constellations, and pilot signals.
- Cognitive networks: ML/AI can be used in aspects like virtualized network function placement, quality of service, and spectrum sharing.
- Network optimization: A flexible AI-native interface can change how networks are optimized, allowing networks to learn what to do instead of engineers designing every aspect of the system.
[More to come ...]
