MeshRoute

1. Abstract

Meshtastic (v2.6/2.7) uses managed flooding with SNR-based suppression for broadcasts and next-hop routing for direct messages. These are clever optimizations over naive flooding — but both still scale as O(n) per message, requiring a hop limit (default 3–7) that caps network range.

This paper presents System 5, a routing protocol that achieves O(hops) cost for all message types. Simulated against Meshtastic's actual managed flooding with realistic hop limits (3, 5, 7) across 21 scenarios (20–1500 nodes, 6 routers each), System 5 delivers 92–100% fewer transmissions in small-to-medium networks while achieving equal or higher delivery rates.

The most significant finding: Meshtastic's hop limit doesn't just cap range — it prevents delivery. At 500 nodes, managed flooding with hop limit 7 delivers only 51% of messages. At 1000 nodes, only 6%. System 5 delivers 43–76% in the same scenarios. At 1000+ nodes, System 5 uses more total TX than managed flooding — but delivers 7x more messages, making the per-delivered-message cost still lower. The hop limit is not a safety net; it is the primary scaling barrier.

2. What Meshtastic Does Today

Meshtastic's routing (v2.6/2.7) is already substantially better than naive flooding:

• Managed Flooding — Before rebroadcasting, nodes listen briefly. If they hear another rebroadcast, they suppress. SNR-based priority gives distant nodes shorter contention windows. ROUTER/ROUTER_LATE roles always rebroadcast. This saves ~40–50% of transmissions vs. naive flooding.
• Next-Hop Routing (v2.6+) — For direct messages only. First DM floods; the system learns which relay succeeded. Subsequent DMs use only that one cached relay. Falls back to flooding if the next-hop fails.
• Congestion Scaling (v2.6+) — Networks with 40+ nodes stretch broadcast intervals: Interval × (1 + (Nodes-40) × 0.075).

The remaining limitation: Both managed flooding and next-hop still scale as O(n) per message. The hop limit (3–7) is necessary because each hop multiplies transmissions proportional to network size. But this creates a double penalty: not only does each message waste bandwidth — messages that need more hops than the limit simply never arrive. At 100 nodes with hop limit 3, delivery drops to 92%. At 500 nodes, to 14%. The hop limit doesn't just cap range — it makes large networks fundamentally unreliable.

3. Approaches Compared

Four routing strategies are simulated and compared on identical network topologies:

4. System 5 — The Solution

System 5 combines six proven concepts, each borrowed from a different domain:

• OSPF Areas → Geo-Clusters — Nodes self-organize by geohash prefix. Full topology shared within clusters, only summarized routes between. Scales from 10 to 50,000+ nodes.
• B.A.T.M.A.N. OGMs → Quality Metric — Periodic originator messages measure link quality by counting reception rates. No complex calculation — just counting arrivals.
• Weighted ECMP → Load Balancing — Traffic distributed proportionally across 2–3 routes. Good paths get more traffic, but never all. No single bottleneck.
• Back-Pressure → Congestion Control — Overloaded nodes report queue pressure via piggyback (2 bytes, 0.8% overhead). Traffic naturally avoids congested paths.
• Ant Colony Optimization → Self-Optimization — Successful deliveries strengthen routes, timeouts weaken them. Unused paths fade. The network learns through experience.
• DNS → Node Discovery — Hierarchical location cache. Ask locally first, then cluster, then region. Scoped flooding only as last resort. Answers are cached.

The proportional distribution creates a negative feedback loop: overloaded paths lose weight → traffic shifts → path recovers → weight increases. The network self-balances without central control.

5. Mathematical Evaluation

Each system was scored 0–10 across seven weighted criteria:

6. Resilience & Adaptive QoS

System 5 handles every critical failure mode with graceful degradation:

• Node failure — Instant failover to cached backup route (0ms). If a border node dies, a second border node takes over. If all borders fail, the cluster falls back to local flooding.
• GPS failure — Neighbor consensus: if 4 of 5 neighbors report geohash "u0x8", the GPS-less node adopts that cluster. If no neighbors have GPS: "homeless" mode with local flooding fallback.
• MQTT bridge failure — LoRa relay subnets activate. Small relay nodes between cities form a physical radio bridge. Slower (more hops) but functional without internet.
• Full internet outage — Local mesh continues operating. Clusters communicate via LoRa relay chains. Cross-continental traffic uses DTN (store-and-forward).

7. Results

Key advantages over Meshtastic's current routing:

• O(hops) vs O(n) — cost depends on path length, not network size. The fundamental scaling improvement.
• Delivery rate — System 5 delivers equal or more messages in most scenarios. At 1000+ nodes, it delivers 7x more than managed flooding. Exception: very sparse networks (Mountain Valley) where no protocol helps.
• Hop limit eliminated — 20 hops cost less than managed flooding costs for 1. Unlocks range and preset freedom.
• All traffic types — unlike next-hop (DMs only), System 5 routes all messages including broadcasts.
• Multi-path resilience — 2-3 cached paths with instant failover, vs. next-hop's single relay.
• Load balancing — proportional distribution prevents bottlenecks. Meshtastic has no load awareness.
• Adaptive QoS — NHS per cluster with priority gate. SOS always gets through.

8. Roadmap & Next Steps