What is this research about?
Machine learning is used in everything from image recognition and classification to natural language processing and robotic systems. But what could these data-driven, trained systems mean for future mobile networks? What if we could engineer radios that could adapt their signalling schemes to achieve the best performance for any situation, learning from one another and effectively thinking for themselves? We believe a dynamic AI/ML-defined native air interface will be a key component of 6G networking in the future.
What are the main use cases?
Private networks could really benefit from these flexible radios. A learning network in a factory could be configured to support industrial sensors at one moment then rapidly reconfigured to support robotic systems or video surveillance. These radios make smart choices – moving between spectrum bands, reconfiguring signal paths and modulation schemes as the specific application demands. With the ability to handle scenarios that engineers haven’t even considered, a flexible, automatically adapting interface could very well change the face of network optimization.
What are some of the advantages of an AI/ML-defined native air interface?
-
A learning radio could dynamically learn and set up bespoke waveforms, constellations, and pilot signals that make more efficient use of available spectrum, resulting in improved performance.
-
AI/ML-based physical layer solutions can enhance the energy efficiency of 6G networks by achieving as much as a 50% reduction in transmit power over 5G for the same bandwidth and data rate.
-
On the MAC layer, AI/ML could customize signalling and access schemes, which could adapt to the service needs automatically.
-
A learned air interface could embrace hardware non-linearities and limitations and fully adapt to any target platform.
-
AI/ML could choose among all the myriad of parameters in a radio network far more effectively than a human.
-
A learned AI/ML air interface could transform how R&D is done by fundamentally altering the way algorithms are designed.
Read our white paper and blog post
