MATSYA-Node：以四層核心系統架構競逐 MATSYA 2026

TL;DR for the Team: A basic IoT tracker or cloud app will fail at sea because there is no internet 20–50 km offshore, fishermen have wet hands and zero time for manual data entry, and they fiercely protect their secret fishing spots.

Here is our winning 4-Tier "Passive Edge + Privacy Mesh" Architecture (MATSYA-Node)—designed to score maximum marks across Technical Feasibility (50 pts), Real-World Impact (40 pts), and Innovation (30 pts).

1. High-Level System Architecture Diagram

[ TIER 1: WET POD (Tethered in Water / Net) ]
  ├── Waterproof DS18B20 (Sea Surface & Depth Temperature)
  ├── MS5837-30BA (Water Pressure / Depth Sensor)
  └── Low-Cost Piezo Hydrophone + Optical Turbidity Sensor
          │ (Marine Tether / Waterproof Connector)
          ▼
[ TIER 2: DECK EDGE BOX (60-Sec Clip-on Gunwale / Wheelhouse) ]
  ├── Compute: Raspberry Pi Zero 2W / ESP32-S3 (Edge AI NPU)
  ├── Vision: 8MP Wide-Angle Sony IMX219 (Angled at Haul/Deck Area)
  ├── Navigation: u-blox NEO-M8N GPS + 9-DOF IMU (Roll/Pitch/Drift/Vibration)
  ├── Mesh Radio: SX1262 LoRa 865-867 MHz (5–15 km Boat-to-Boat Offline Mesh)
  └── Power: 12V Boat Battery Tap + 10W Solar Trickle + 10,000mAh LiFePO4
          │ (Local Wi-Fi AP / BLE - 100% Offline at Sea)
          ▼
[ TIER 3: FISHERMAN VERNACULAR COMPANION (Smartphone / Audio Prompt) ]
  ├── Offline Map Engine: Pre-synced INCOIS PFZ + Live Local Thermal Fronts
  ├── Voice AI: Offline Voice Prompts (Tulu, Kannada, Malayalam, Tamil)
  ├── Digital Catch Ledger: Auto-logged weight/species -> Instant Harbour Receipt
  └── Privacy Vault: Exact GPS pins AES-256 encrypted locally (Fisherman Only)
          │ (Store-and-Forward Sync via 4G / Harbour Wi-Fi on Return)
          ▼
[ TIER 4: HARBOUR CLOUD & SCIENTIFIC INTELLIGENCE (CMFRI / INCOIS) ]
  ├── Privacy Sanitizer: Strips exact GPS -> Maps to ~8-10km Uber H3 Hex Grids
  ├── Spatiotemporal Engine: PostGIS + TimescaleDB (Depth x Season x Species)
  └── Ecosystem Dashboard: CPUE (Catch Per Unit Effort), Juvenile Bycatch & PFZ Validation

2. Deep Dive into the 4 Architectural Tiers

Tier 1 & Tier 2: Clip-On Hardware Node ("Zero Boat Modification")

• 60-Second Portable Mount: Clamps onto any traditional motorized country craft (gillnetter) or multi-day trawler using a heavy-duty marine C-clamp. Housed in an IP67 polycarbonate enclosure built to survive saltwater spray, humidity, and deck washdowns.
• Passive Ocean & Vessel Sensing:
  • Water Pod: Continuously records Sea Surface Temperature (SST), depth profile temperature, turbidity, and ambient acoustic energy.
  • IMU Engine State Classifier: Uses 9-DOF accelerometer/gyroscope vibration signatures to automatically classify whether the boat is Steaming, Searching, Drifting, or Hauling Nets—calculating exact fuel burn per activity and surface current drift without touching the boat's engine wiring.
• Zero-Touch Haul Vision (Edge AI):
  • Fishermen will never type catch logs by hand. When the IMU detects net-hauling motion, the deck camera wakes up and runs an INT8-quantized YOLOv8n-seg model on the edge.
  • Automatically identifies dominant species (Indian Mackerel, Oil Sardine, Seer Fish, Prawns, Ribbonfish), estimates crate/net biomass (kg), and flags juvenile bycatch percentages.

Tier 3: Offline Fisherman Companion & LoRa Mesh ("Why a Fisherman Will Use It")

Judges explicitly ask: "Why would a fisherman use it?" Our Tier 3 delivers immediate monetary and safety ROI on Day 1:

1. Cuts Fuel Burn by 15–25%: Before leaving harbour, the app caches INCOIS Potential Fishing Zone (PFZ) advisories. Out at sea (with zero mobile network), it overlays live temperature readings from the boat's Tier 1 sensor onto the offline map—alerting the fisherman via Tulu/Kannada voice prompts the moment the boat crosses a thermal front where fish congregate.
2. Free Boat-to-Boat LoRa Safety Mesh (865–867 MHz ISM Band): Creates a 5–15 km offline mesh network between nearby fishing boats for SOS distress relays, sudden squall/rough-sea alerts, and cooperative thermal-front updates without expensive satellite terminals.
3. Automated Trip & Auction Ledger: Generates an automatic digital trip sheet (Fuel used vs. Species & Weight caught) so boat owners and crew can transparently split earnings and use verified catch history to access low-interest bank credit and insurance.

Tier 4: Dual-Ledger Privacy Protocol ("Why a Fisherman Will Trust It")

Judges explicitly ask: "Why would a fisherman trust it?" Secret fishing spots are a fisherman's livelihood. We solve this hardware-up using a Dual-Ledger Split on the Edge Box before any data ever syncs to the cloud:

3. How Our Architecture Answers the 5 Mandatory Ocean Pillars

4. Bill of Materials (BOM) — Proving Affordability & Scalability

A boat burns ₹8,000 to ₹40,000+ in diesel per trip. Our entire hardware node costs ₹4,500 (and drops to ₹3,100 at 1,000-unit scale), meaning fuel savings pay back the hardware cost in 1 to 2 trips.

5. Team Role Split by Architectural Module

To lock in the 30 cumulative points for Multi-Domain Team Composition, here is who owns what across our 4 tiers:

1. Member 1 — Embedded & Marine Hardware Lead (Tier 1 & 2): Sensor interfacing (Temp/Depth/Turbidity), 12V/Solar power management, IP67 waterproofing, and LoRa 865 MHz radio mesh.
2. Member 2 — Edge AI & Computer Vision Lead (Tier 2): Training & INT8 quantization of YOLOv8n-seg on Indian marine species (Mackerel, Sardine, Seer, Prawn) + IMU boat-activity classifier.
3. Member 3 — Offline Mobile & Privacy Systems Lead (Tier 3): Offline map caching (INCOIS PFZ), Tulu/Kannada voice alerts, Dual-Ledger encryption, and Uber H3 spatial anonymization.
4. Member 4 — Ocean Data Science & GIS Lead (Tier 4): PostGIS + TimescaleDB cloud pipeline, CPUE stock calculation, and CMFRI/INCOIS scientific dashboard.
5. Member 5 — Product, Field Ops & Fisherman Validation Lead: BOM economics, Mangaluru Old Port (Bunder) / Malpe harbour fisherman interviews, real-world crate testing, and pitch storytelling.

6. Execution Roadmap Across All 3 Phases

• Phase 1 (Idea Submission): Submit this complete 4-Tier Architecture, Dual-Ledger Privacy specification, BOM economics, and early fisherman interview insights from Mangaluru harbour.
• Phase 2 (Proof of Concept — Nov 2026):
  • Working benchtop + enclosure prototype reading live temperature, GPS, IMU state, and LoRa packets.
  • Trained YOLOv8n model tested on real fish crates at Mangaluru Bunder / Malpe harbour showing live species & weight estimation.
  • Functional mobile app demo showing exact local GPS vs. H3 anonymized cloud sync.
• Phase 3 (Grand Finale at CHRIST University — Jan 2027):
  • Fully waterproofed IP67 MATSYA-Node field-tested on a local fishing boat, complete with fisherman usability feedback and live end-to-end stage demonstration.