Skip to content

Technologies

Reference Materials and Resources

Core Platform Architecture

The platform is a universal patchbay for business logic, developed based on decades of experience in Mission Critical Systems.

  • Foundation: Built on NATS.io (JetStream), the neural system of modern distributed computing.

    • Why NATS?: It solves the "Distributed Systems" problem. Unlike HTTP (REST) which fails when a link breaks, NATS is "Dial-Tone" reliable.
    • Real-World Scale:
      • Walmart: Uses NATS for edge-to-cloud sync in thousands of stores.
      • Mastercard: Reliable financial messaging.
      • Tinder: Real-time matching at massive concurrency.
    • Resiliency Benefit: If the satellite link drops, the system buffers every event (Store-and-Forward). When the link returns, the data "replays" in order. Zero data loss.

  • Capacities:

    • Orchestration: Deploys logic ("Gears") to the edge.
    • Observability: The platform provides cloud-native dashboards and apps for real-time herd visualization.
    • Interface: GenAI-Ready API allowing natural language queries ("Which paddock has the most lame cows?") by feeding structured JSON stats to LLMs.

For more details, refer to the internal architectural documentation.

Scientific Validation & Academic Basis

The architecture doesn't guess; we implement proven science.

The "Mesh" Physics (LoRaWAN Research)

Academic papers confirm that Multi-Hop LoRa Mesh is superior for livestock in broken terrain compared to single-hop Star topologies.

Lameness Algorithms (Wageningen / Lincoln NZ)

Our "Lameness Gear" (Wasm) is based on peer-reviewed biomechanics.

  • The Science: Lame cows change their "Lying Bouts" frequency and duration days before visual limping appears.
  • Benchmarks:

Hardware Components (Reference Spec)

The following represents the Gen 1 Reference Design. The architecture is hardware-agnostic and can run on equivalent industrial processors.

Role Component Technology Rationale
Member Node SoC STM32WLE5 (Cortex-M4 + LoRa) Integrated Low Power SoC. Support for LoRa P2P. Target Bom <$15.
Member Node Battery Li-SOCl2 (Primary AA) High Density. Target 5+ Years life.
Leader Node Modem Quectel BG770A-SN Skylo/NTN Certified. Supports NB-IoT over Satellite and Cellular.
Leader Node Power e-peas PMIC + LTO/Supercap SODAQ Architecture. Harvests energy efficiently for "Zero Maintenance".
Network Cluster LoRa P2P (915 MHz) Superior penetration vs Bluetooth (used by 701x).

Network Topology: Dynamic (Runtime Switching)

Answer: Yes, the topology is defined in Software (Runtime), not Hardware.

We use a Hybrid State Machine on the STM32WLE5. The hardware is identical; the behavior changes based on the environment.

The "Beacon" Protocol

  1. Default State (Star/Cluster):
    • The tag listens for the Leader's Beacon (sent every minute).
    • If heard: It syncs, sleeps, and sends data directly. (Battery Life: 5+ Years).
  2. Fallback State (Mesh/Repeater):
    • If the Beacon is LOST (stray cow, behind a hill):
    • The tag switches to "Search Mode". It sends a "Help" ping.
    • Nearby cows (who can hear the Leader) hear the "Help" ping and temporarily act as Repeaters.
    • Trade-off: Power increases during this event, but data is saved.

Why this wins?

  • Voisin/Flat: 100% Star Mode. Maximum battery.
  • Basalt/Hills: The system automatically degrades into a partial Mesh only where needed (in the blind spots), preserving the battery of the rest of the herd.
graph LR
    subgraph Star["Star Topology (Competitor)"]
        G1((Gateway))
        G1 --- C1[Cow 1]
        G1 --- C2[Cow 2]
        G1 --- C3[Cow 3]
        G1 -.->|Out of Range| CX[Cow X]
    end
graph TD
    subgraph Mesh["Mobile Mesh (Cattle Tracking)"]
        Cow[Member Node]
        Leader[Leader Hub]
        Cloud[Platform Cloud]
    end
    M2 --- M3[Member 3]
        M3 --- MX[Member X]
        style MX fill:#4CAF50,stroke:#2E7D32
    end
  • Proven Scale:
    • Wirepas: Used in Oslo Smart Grid (700,000 devices).
    • Amazon Sidewalk: City-wide coverage via consumer mesh.
    • Maersk: Logistics tracking in mesh-hostile ports.

Software Stack

Layer Technology Purpose
Edge OS Zephyr RTOS Real-time efficiency, broad driver support.
Logic/Gears TinyGo / Wasm Safe, portable code execution on the edge.
Communication NATS / JetStream Cloud-native messaging and persistence.
Analytics DuckDB Fast SQL analytics on the Platform layer (not on-tag).

Edge Intelligence (The "Brain" on the Cow)

To reduce satellite costs, Leader Nodes process data locally using WebAssembly (Wasm) modules. Only high-confidence alerts are uploaded.

  • Lameness Gear: Analyzes accelerometer gait patterns.
  • Estrus Gear: Detects mounting behavior.
  • Theft Gear: Detects non-biological acceleration (>20km/h) indicative of truck transport.

[!IMPORTANT] The Moat: While the hardware is based on reference designs, the Wasm Gears represent JAAB Tech's core Intellectual Property—the "Biological Brain" that others cannot easily replicate.

Security & Trust (Zero Trust Architecture)

The platform operates in "hostile" environments (Internet is optional, physical access is easy).

  • Leader Node: Solar-powered gateway. Designed with a 20% energy buffer to survive 14 days of winter overcast (Winter Solstice Margin).
  • Data Sovereignty: The rancher owns the data. We provide the "Utility", not the "Walled Garden".
  • Immutable Logs: The Snake Protocol acts as a tamper-proof flight recorder.
  • Traffic Management: To prevent saturation, the mesh uses Exception-Based Reporting.
    • Quiet Mode: Healthy animals send 1 heartbeat per hour.
    • Alarm Mode: Lame, Estrus, or Speeding animals transmit instantly.
    • Result: 99% of bandwidth is preserved for critical events.

Key References