An Adaptive ML Link Layer Behavior System for mmWave Networks

Millimeter wave (mmWave) communication technology fulfills the requests of networking applications seeking higher bandwidth and lower latency from wireless networks. In spite of its advantages, mmWave is very susceptible to its surrounding environments such as physical barriers, which can lead to link impairments. To address this, we aim to explore machine learning (ML) techniques to learn the link layer characteristics of mmWave behavior under various environmental conditions, helping it better adapt to these obstructions. This paper evaluates the performance of foundational machine learning models using link layer parameters of mmWave link to predict its throughput which then feeds into a multipath scheduler system aimed at adapting to these links during data transmission.