ML-Optimized Beam-based Radio Coverage Processing in IEEE 802.11 WLAN Networks

Abstract

Dynamic Radio Resource Management (RRM) is a major building block of Wireless LAN Controllers (WLC) function in WLAN networks. In a dense and frequently changing WLANs, it maximizes Wireless Devices (WD) opportunity to transmit and guarantees conformance to the design Service Level Agreement (SLA). To achieve this performance, a WLC processes and applies a network-wide optimized radio plan based on data from access points (AP) and upper-layer application services. This coverage processing requires a "realistic" modelization approach of the radio environment and a quick adaptation to frequent changes. In this paper, we build on our Beam-based approach to radio coverage modelization. We propose a new Machine Learning Regression (MLR)-based optimization and compare it to our NURBS-based solution performance, as an alternative. We show that both solutions have very comparable processing times. Nevertheless, our MLR-based solution represents a more significant prediction accuracy enhancement than its alternative.