Waveform management approach with machine learning for 6g systems

Abstract

5th Generation (5G) systems are designed with a more flexible structure compared to previous generations with an increasing variety of applications and services. Thus, new flexibility dimensions are observed in 5G technologies. Furthermore, emergence of these flexibility dimensions is triggered a need for advanced management paradigms for 5G and beyond. It is expected that application richness, flexibility dimensions, and the related management paradigms will show an increase with 6th Generation (6G) systems. It is possible that different flexibilities related to the waveform design can be introduced in 6G while a uniform method is used in 5G and previous generations. One of these flexibilities can be the ability to make selection through a waveform set for a new capability to meet different application and user requirements with the waveform selection. In this paper, waveform selection approaches are proposed based on machine learning (ML) with single-stage and multi-stage networks for the waveform management in the same coverage area under the assumption that multiple waveforms can be used in 6G. Hence, the problem of deciding on the best waveform for a coverage area considering different requirements and environmental conditions is studied. To provide environmental awareness, a new synthetic dataset is formed with an example simulation setup. Moreover, a feature control algorithm is proposed to limit side effects of the waveform selection approaches.