WIP: Federated Learning for Routing in Swarm Based Distributed Multi-Hop Networks

TL;DR

This paper explores integrating federated learning with the B.A.T.M.A.N. routing protocol to enable UAV networks to predict topology changes and improve routing efficiency in dynamic environments.

Contribution

It proposes a novel approach combining federated learning with B.A.T.M.A.N. to enhance routing in UAV networks with dynamic topologies.

Findings

Designed an FL testbed on a network emulator.

Proposed ML model for topology prediction.

Aimed to reduce network congestion.

Abstract

Unmanned Aerial Vehicles (UAVs) are a rapidly emerging technology offering fast and cost-effective solutions for many areas, including public safety, surveillance, and wireless networks. However, due to the highly dynamic network topology of UAVs, traditional mesh networking protocols, such as the Better Approach to Mobile Ad-hoc Networking (B.A.T.M.A.N.), are unsuitable. To this end, we investigate modifying the B.A.T.M.A.N. routing protocol with a machine learning (ML) model and propose implementing this solution using federated learning (FL). This work aims to aid the routing protocol to learn to predict future network topologies and preemptively make routing decisions to minimize network congestion. We also present an FL testbed built on a network emulator for future testing of the proposed ML aided B.A.T.M.A.N. routing protocol.