GHOST IN THE ROUTER: FIVE ROUTING TECHNIQUES TO ELIMINATE CONGESTION IN MESH NETWORKS

Abstract

Wireless mesh networks (WMNs) promise resilient, self-organising connectivity, but their multi-hop shared medium makes them acutely susceptible to congestion: as traffic converges on gateway-adjacent nodes, throughput collapses, latency spikes, and packet loss cascades. Conventional shortest-hop routing exacerbates this by funnelling flows onto a few central relays. This study evaluates five practical routing techniques—congestion-aware metric routing, load-balanced multipath forwarding, gateway-aware load distribution, queue-length-based backpressure routing, and adaptive airtime-fair path selection—for their effectiveness in eliminating congestion in mesh networks. A discrete-event simulation testbed modelling a 50-node IEEE 802.11s-style mesh was subjected to escalating offered loads and hotspot traffic patterns across 900 simulation runs. The five techniques were compared against a shortest-hop (hop-count) baseline on aggregate throughput, end-to-end latency, packet-delivery ratio, and fairness. Results show that no single technique dominates across all conditions, but a combined congestion-aware multipath approach with backpressure improved aggregate throughput by 41%, reduced 95th-percentile latency by 58%, and raised packet-delivery ratio from 0.79 to 0.96 under heavy load relative to the baseline. Airtime-fair selection produced the largest fairness gains, while gateway-aware distribution most effectively relieved gateway hotspots. The findings demonstrate that congestion in mesh networks is best addressed by combining a congestion-sensitive metric with multipath spreading and local backpressure rather than by any isolated tweak. The study contributes a reproducible comparative evaluation and a layered guidance for practitioners deploying congestion-resilient mesh routing.