1. Delivery Rate Comparison
The core question: what percentage of messages actually arrive?
Delivery Rate by Scenario
Managed Flooding vs WalkFlood — higher is better
Key finding: WalkFlood delivers 3.7x–22x more messages than managed flooding. In small/medium networks, WalkFlood achieves 99–100% delivery while flooding collapses to 16–87% under half-duplex.
2. TX Efficiency
Delivery rate alone is not enough — how many radio transmissions does each approach burn? Log scale reveals the gap.
Total TX Count (log scale)
Lower is better — less airtime consumed, more capacity for other traffic
At Bay Area 1200-node scale, WalkFlood uses 6.5x fewer TX than managed flooding while delivering 22x more messages. In small networks, the TX savings are 10x.
3. The Half-Duplex Cascade Problem
Half-duplex radios cannot transmit and receive simultaneously. When a mountain node broadcasts, all neighbors are blocked from receiving other traffic. This cascade effect is why flooding collapses at scale.
Flood: Mountain Node Broadcasts
234 neighbors all blocked simultaneously.
Massive collision cascade.
WalkFlood: Directed TX
1 neighbor receives. 233 others remain free.
Minimal channel impact.
This is the fundamental difference. A flood TX blocks N neighbors. A directed TX blocks 1. With 234 neighbors on a mountain node, that is a 234x difference in channel impact per transmission.
4. Path Quality — Why Dijkstra Beats BFS
BFS finds the shortest path (fewest hops). Dijkstra finds the most reliable path (best link quality product). In real terrain, these are very different.
Mountain Path (BFS choice: 2 hops)
Links: [0.05, 0.05] → P(delivery) = 0.05 × 0.05 =
0.25%
With 3 retries: 1 − (1−0.0025)³ =
0.75%
Hill Path (Dijkstra choice: 5 hops)
Links: [0.7, 0.6, 0.5, 0.6, 0.7] → P(delivery) = 0.7 × 0.6 × 0.5 × 0.6 × 0.7 =
8.82%
Per-hop with 3 retries: [0.973, 0.936, 0.875, 0.936, 0.973]
P(end-to-end) = 0.973 × 0.936 × 0.875 × 0.936 × 0.973 =
72.6%
Path Reliability: Mountain (2-hop) vs Hill (5-hop)
Per-hop retries transform a mediocre path into a reliable one — but only if individual links are decent
The 5-hop hill path is 97x more reliable than the 2-hop mountain path (72.6% vs 0.75%). BFS would choose the mountain. Dijkstra chooses the hills. Reliable path > short path.
5. Scaling Behavior
The most surprising result: WalkFlood improves with more nodes while flooding degrades.
Delivery Rate vs Network Size
Bay Area topology: 235 nodes vs 1,200 nodes
More nodes = more flooding collisions (6% → 4% delivery). But more nodes also = more overhearing = richer route tables. WalkFlood goes from 84% to 88% delivery. Density helps WalkFlood, hurts flooding.
6. Phase Breakdown — How Messages Get Delivered
For Bay Area 235 nodes: which protocol phase delivers each message?
Message Delivery by Phase (Bay Area 235n, 200 messages)
WalkFlood’s four-phase fallback in action
80 messages (40%) delivered by direct routing — zero exploration needed. 7 by random walk. 80 by scoped mini-flood. 33 dropped (unreachable partitions). The protocol correctly escalates: try cheap first, expensive only when needed.