An MDP Model for Optimal Handover Decisions in mmWave Cellular Networks

TL;DR

This paper introduces an MDP-based framework for optimal handover decisions in mmWave cellular networks, effectively managing intermittent links and dynamic conditions to improve capacity and reduce handoffs.

Contribution

It presents a novel MDP model for cell selection in mmWave networks, incorporating channel load and link quality, solved via Value Iteration Algorithm for optimal association decisions.

Findings

Significant capacity improvement over traditional methods

Reduction in total handoffs and signaling overhead

Enhanced quality of service for users

Abstract

The new frontier in cellular networks is harnessing the enormous spectrum available at millimeter wave (mmWave) frequencies above 28 GHz. The challenging radio propagation characteristics at these frequencies, and the use of highly directional beamforming, lead to intermittent links between the base station (BS) and the user equipment (UE). In this paper, we revisit the problem of cell selection to maintain an acceptable level of service, despite the underlying intermittent link connectivity typical of mmWave links. We propose a Markov Decision Process (MDP) framework to study the properties and performance of our proposed cell selection strategy, which jointly considers several factors such as dynamic channel load and link quality. We use the Value Iteration Algorithm (VIA) to solve the MDP, and obtain the optimal set of associations. We address the multi user problem through a distributed iterative approach, in which each UE characterizes the evolution of the system based on stationary channel distribution and cell selection statistics of other UEs. Through simulation results, we show that our proposed technique makes judicious handoff choices, thereby providing a significant improvement in the overall network capacity. Further, our technique reduces the total number of handoffs, thus lowering the signaling overhead, while providing a higher quality of service to the UEs.