Skip to content

Cattle Tracking Initiative

Hardware Partners Sync: Status & Open Challenges

Phase 1 Pilot Preparation

layout: image-right image: /hereford_two_tier_network.png

The Core Vision

Infrastructure-Free Two-Tier Network

  • Goal: Eliminate dependency on fixed towers or cellular networks.
  • Member Nodes: Ultra-low power Ear-Tags for 97% of the herd.
  • Leader Nodes: Edge-processing Collars for 3% of the herd.
  • Connectivity: Sub-GHz local mesh to Collar -> NTN Satellite (Skylo) to Cloud.

layout: section

The Ear-Tag (Member Node)

The Weight vs. Cost Dilemma

Partner Context: Gosuncn proposed using an off-the-shelf tag designed for Mengniu (dairy). * The Problem: It integrates a LoRa module on top of an existing design, making it heavy (>30g). * Our Target: < 20g for a standard single-pin ear attachment to avoid sanitary/retention issues. * Strategic Decision: * Use the >30g tag only for Phase 1 Pilot to save NRE cost and time. * Evaluate dual-pin mechanical setup if we use a heavier tag.

The "New Tag" and Mesh Protocol

Partner Context: Lierda promised a new tag in October. * The Challenge: We need preliminary specs now to evaluate if their new tag aligns with our requirements, even if not final. * Mesh Protocol: We must explicitly validate their LoRa/Sub-GHz mesh capabilities. Can it support hundreds of nodes in an infrastructure-free topology without crashing?

layout: section

The Collar (Leader Node)

The Single-Channel LoRa Capacity Trap

The Challenge: Aggregating 100-500 Ear-Tags to a single Collar. * Commercial Gateways (Towers): 8-channel concentrators (e.g., SX1302). Can handle 10,000+ nodes simultaneously. * Collar Transceiver: Single-channel (e.g., SX1262). Can only listen to one frequency/speed at a time. * The Solution: We cannot use standard LoRaWAN for the local network. We need a custom MAC protocol (TDMA / synchronized polling) to prevent massive collisions.

Processing, Memory & Firmware Control

The Hardware Stack: * MCU: Capable MCU with FPU/DSP (optimized for low-power operations). * OS: Zephyr RTOS native support required. * Memory: 8-16MB external SPI NOR Flash to cache 5 days of data (orbital blind spots).

The Firmware Requirement: * We require a low-level C/C++ SDK and a functional template app. We will not outsource our core IP. We must have full control to inject custom algorithms.

Edge Algorithms & Battery Autonomy

Battery Constraint: Collar battery must last 1 year, but cannot listen 24/7 without draining. * Wake-on-Radio: Required to synchronize listening periods. * TinyML (Edge Logic): Our algorithms will continuously evolve based on field feedback. * Over-The-Air (FOTA): We must be able to deploy new, arbitrary TinyML models (e.g., via Edge Impulse) over the satellite link.

layout: section

Phase 1 Pilot Strategy

Parallel Validation Tracks

Because this is a complex integration, we must decouple the risks: 1. Local Cow Network (Sub-GHz): Test body-shadowing effects and the custom MAC protocol collisions. 2. NTN Satellite Link (Skylo): Test coverage, B255 antenna tuning, and payload compression (using bitmaps instead of raw JSON).

The "Handheld Gateway" Pivot

To test the software and network layers immediately without waiting for the Collar's mechanical and battery challenges: * The Concept: Use a handheld/backpack datalogger as the Leader Node for the first pilot. * Why? Battery autonomy is not an issue (recharged nightly), and we can test the LoRa-to-NTN bridge immediately. * Constraint: The handheld must still use a single-channel transceiver to accurately simulate the final Collar capacity.

Summary & Next Steps

  1. Gosuncn (Call Thursday Night): Confirm the FPU MCU, memory specs, and the SDK approach. Accept the heavy tag for Phase 1.
  2. Lierda: Wait for their mesh protocol details and preliminary October tag specs.
  3. Oxeltech: Wait for their technical analysis based on the Research Pack.