Skip to content

ADR 0003: Hybrid Mesh Topology

Date: 2026-01-17 Status: SUPERSEDED by ADR 0007 — MVP Topology Selection (2026-05-08)

Why superseded: The mesh-first decision below was revisited in February 2026 (proj/topology_comparison.md) on power-budget grounds. ADR 0007 keeps mesh as a planned future state on mesh-capable hardware but ships single-mode Star for the MVP and Phase 2. Read this ADR for the original mesh rationale; read ADR 0007 for what is actually being built.

Context

Traditional LoRaWAN uses a "Star" topology where every node speaks directly to a fixed gateway on a tower. * Problem: In extensive "Pampa" environments (hills, forests), Line-of-Sight is often blocked. Building towers is CapEx-prohibitive. * Problem: Competitor mesh solutions (Zigbee) are power-hungry and fail when nodes move (routing table breakage).

Decision

We will implement a Stateless Flooding Mesh (Managed Flood) logic:

  • Cow-as-Repeater: Every Tier 1 node repeats messages from its neighbors.
  • Logic: "Broadcast and Pray" with Time-To-Live (TTL) limits and probabilistic rebroadcast to prevent storms.
  • Time Slicing: Nodes wake up synchronously for milliseconds to transmit/listen, then sleep.

Consequences

  • Positive: Zero Infrastructure. The herd is the network. If the herd moves behind a hill, the mesh adapts instantly without table rebuilding.
  • Negative: Lower total bandwidth than Star topology. Latency is non-deterministic (multi-hop).
  • Mitigation: We use "Exception-Based Reporting" (ADR 0005) to minimize traffic.