Nodes Number Estimation based on ML for Multi-operator Unlicensed Band Sharing to Extend Indoor Connectivity

Abstract

Due to ever-increasing data and resource-hungry applications, the needs of new spectrum by mobile networks keep increasing. Unlicensed spectrum is still expected to play a crucial part in meeting the capacity demand for future mobile networks. But if this will be a reality, fair coexistence attained via practical and efficient channel access procedures would be necessary. In designing such channel access schemes, awareness of the number of nodes contending for the channel resource can be strategic. This paper investigates a node number estimation approach using machine learning (ML) techniques. When multiple nodes access the same unlicensed channel, varying idle-time can be associated to a statistical distribution. In this paper, a statistical distribution of the Idle-time slots over the channel are used to characterise and analyse the channel contention based on the number of nodes. Three ML model based approaches are evaluated and the results confirm that the proposed solution’s viability but also reveal the best performing ML technique for the task of node number estimations.