Lessons learned in radio channel measurements and modeling: Promising opportunities for deep learning for physical layer wireless communication networks and sensing
* Register (or log in) to the AI4G Neural Network to add this session to your agenda or watch the replay
This talk discusses radio propagation channel characteristics and modeling approaches from below 6 GHz to the millimeter wave and THz spectrum bands, with a focus on AI/ML opportunities. The talk highlights advances in modeling the spatial and temporal nature of radio channels, including the impact of antenna patterns and received signal envelopes of received signals, thereby offering insights into phenomena that can be used in learning models for artificial intelligence (AI) to predict signal behavior in real-world channels. Special emphasis is given to the relatively new Two-Wave With Diffuse Power (TWDP) distribution that encompasses Raleigh and Rician fading as special cases, and repeatable measured phenomenon of diffraction effects when a receiver encounters an object that physically begins to block a radio path. New approaches that learn the directions of arrivals and the extent of multipath from simple narrowband envelope measurements, and how to optimize and learn channel behaviors from ray tracing are also presented, thus offering insights into promising ML/AI approaches to physical layer prediction and learning.