Skip to content

ADR 0007: MVP Topology Selection — Star Now, Mesh-Capable Hardware

Date: 2026-05-08 Status: ACCEPTED Supersedes: ADR 0003 — Hybrid Mesh Topology

Context

ADR 0003 (Jan 2026) chose a Stateless Flooding Mesh as the Member-to-Leader communication topology, on the grounds of "Infrastructure-Zero" coverage and instant adaptation when herds move behind terrain. Three weeks later, proj/topology_comparison.md (Feb 2026) revisited that decision and recommended Star (Cluster) for the MVP — citing a 1–2 yr vs. 5–7 yr battery gap and the engineering risk of meeting the 5-year power budget on a mesh.

That contradiction sat unresolved through Phase 1 planning. With the Innovar PY pilot (50 head, March 2026) approaching as the first field deployment, the topology choice has to be locked before Member-tag firmware is finalized.

Decision

For the MVP and through Phase 2, the Member-to-Leader topology is Mobile Star (Cluster). The hardware is designed to be mesh-capable but ships in single-mode Star firmware; mesh becomes a recompile, not a hardware respin, when (and if) field evidence justifies it.

1. MVP firmware: single-mode Star

  • Members sleep ≥99% of the time and wake on a Leader beacon (LoRa P2P, Class A-style).
  • No multi-hop routing, no TTL handling, no probabilistic rebroadcast in the shipping image.
  • Member firmware is intentionally "dumb" — fewer code paths, fewer bugs, faster certification.

2. Hardware: mesh-capable, single-mode shipped

  • The Member radio + MCU combination is selected to support both topologies natively (Sub-GHz LoRa P2P + 802.15.4g-class mesh on the same silicon).
  • BOM is not sized for dual-mode runtime. The radio is capable of either; the deployed firmware picks one.
  • No feature flag in the shipping image. Switching to mesh later means a new firmware build and a controlled OTA rollout, not a runtime toggle.

3. Mesh stays a planned future state

  • Mesh capability moves from "Phase 1 architecture" to a "Phase 3+ option, gated on field evidence".
  • Promotion from option to plan requires both: (a) measured battery budget on Star firmware in extensive deployment to establish a baseline, and (b) measured stray-cow frequency or paddock-size data showing Star coverage is insufficient.
  • Until then, the marketing narrative is honest: "the herd is mesh-capable" rather than "the herd is the mesh".

Consequences

Positive

  • Schedule risk for Phase 1 drops sharply. Power budget on Star is a textbook calculation; on mesh it is a research problem.
  • 5–7 year battery life is realistic on AA-class energy budgets, preserving the cycle-aligned economics in ADR 0006.
  • Member firmware complexity drops — fewer code paths, faster path to certification (FCC Part 15, ICAR), lower defect risk in the first deployments.
  • Hardware investment is preserved. Choosing mesh-capable silicon now means a future mesh decision costs a recompile, not a board respin.

Negative

  • Range is hard-capped at ~2–3 km from each Leader. Spread paddocks need higher Leader density than the original mesh thesis assumed.
  • Stray-cow risk is real. Animals that drift beyond Leader radius go dark until they return. Mitigated, not eliminated.
  • Marketing narrative shifts. "The Rodeo es la Antena" stays directionally true but reads as aspiration during Phase 1; the field reality is Star-with-mesh-capable-hardware.

Mitigations

  • Leader density planning becomes a deployment-design input, not just a hardware-cost line. Phase 1 paddock surveys must measure max Member-to-Leader distance.
  • Stray-cow telemetry is a Phase 1 success metric — count animals that go silent for >24 h, and correlate with paddock geometry. This is the primary feed for the future mesh-promotion decision.
  • Brand voice update. docs/00_business/brand/voice_es.md should soften "rodeo es la antena" framing for Phase 1 external materials; full mesh narrative returns when mesh ships.

Open Items

  • Member silicon selection — confirm the chosen MCU + radio combination supports both LoRa P2P and 802.15.4g mesh natively (or confirm the explicit alternative). Owner: TBD.
  • Leader-density planning input added to the deployment-design checklist. Owner: TBD.
  • Stray-cow telemetry counter added to the Phase 1 success-metric set. Owner: TBD.
  • Update proj/topology_comparison.md status banner — done in this commit.
  • Mark ADR 0003 as Superseded — done in this commit.