Using Transition Learning to Enhance Mobile-Controlled Handoff In Decentralized Future Networks

TL;DR

This paper proposes a decentralized approach where mobile devices use transition learning algorithms to manage handoffs independently, addressing challenges in future networks lacking centralized control.

Contribution

It introduces a novel client-side learning framework for mobility management in decentralized networks, diverging from traditional network-centric models.

Findings

Demonstrates improved handoff reliability in decentralized scenarios

Shows that client-side learning reduces dependency on centralized data

Validates approach through simulation results

Abstract

Traditionally, resource management and capacity allocation has been controlled network-side in cellular deployment. As autonomicity has been added to network design, machine learning technologies have largely followed this paradigm, benefiting from the higher compute capacity and informational context available at the network core. However, when these network services are disaggregated or decentralized, models that rely on assumed levels of network or information availability may no longer function reliably. This paper presents an inverted view of the resource management paradigm; one in which the client device executes a learning algorithm and manages its own mobility under a scenario where the networks and their corresponding data underneath are not being centrally managed.