AI-Assisted Link Adaptation: Predicting Hybrid Automatic Repeat Request Success

Hybrid Automatic Repeat Request (HARQ) is a cornerstone technology for ensuring reliable data transmission in wireless communication systems. At its heart lies HARQ feedback, a mechanism that bridges the gap between sender and receiver, which ensures data integrity by enabling the receiver to inform the sender about the success or failure of a transmission. This feedback is used by an outer-loop link adaptation algorithm to maintain a target block error rate (BLER). In this paper, we present a novel artificial-intelligence (AI)-assisted link adaptation algorithm that can predict the likelihood of success before the transmission. Leveraging carefully selected features extracted from a real 5G network baseband logs, our approach dynamically adjusts link adaptation parameters to ensure reliable reception under real-time hardware constraints. Our approach is also suitable for online learning-avoiding the problems that arise due to time varying wireless channels faced by other ML approaches. Our results demonstrate significant improvements, achieving up to an 96% reduction in retransmissions and a 19.4% throughput increase compared to traditional techniques using data from an operational 5G site.