Link Adaptation Techniques Based on Inner Receiver Statistics and Machine Learning

Abstract

Future mobile networks will provide a wide variety of new applications and use cases. In this context, a significant demand for high data rate and reliability will be driven to support all new emerging services. Many of these services have different requirements, which can lead to divergent decisions being taken according to the same indicator. Using the modulation and coding parameters defined by the Modulation and Coding Scheme (MCS) index as an example, the system option must result in a number of errors lower than a limit condition when the requirement is reliability. On the other hand, if the service demands a high data rate, the MCS index must be as close as possible to the point of highest spectral efficiency that still reaches a less severe maximum target of error rate. But in either case, the link status indicator needs a very accurate estimation. The elementary indicator is the measure that represents the interference level and noise compared to the sent signal. Since the calculation of this elementary indicator still requires improvements in the current methods, a strategy presented in this work implements a complementary indi cator that generates not only more precise information but also considers the imper fections of all processes involved in the signal reception. In the pursuit of state-of-the art, it is clear that the implementation particularities in signal processing algorithms, hardware characteristics, quantization, and many other factors present in the reception process must be included. Moreover, the decision-making process is also a complex task as it involves man aging multiple users being added and leaving the system at any given time. In this scenario, there is an increasing heterogeneity of services and different demands to be accommodated. With as much information as possible, the process called Link Adap tation (LA) can be appropriately implemented. For the best efficiency of link adaptation, this work is based on these three topics, presenting strategies for each one of them: a proposal for improving the estimation of the elementary indicator (signal-to-noise ratio), the creation of a complementary indi cator and more systemic considerations for the control itself. In addition, the research looking for methods to increase precision and efficiency brought the employment of machine learning techniques as a primary strategy, which has shown remarkable results in many diverse areas.

Keyords: Link Adaptation, Adaptive Coding and Modulation, 6G, Machine Learning, Reliability