Hh3.0

1. Title of the project

Green Node

2. Problem Statement

The global agricultural sector faces a massive adoption gap in precision agriculture. Out of over 500 million small farms globally, only about 10% use digital agricultural services.

• Prohibitive Costs: Commercial sensor nodes cost upwards of $200–$250 per unit, with traditional cellular IoT incurring high recurring costs and requiring expensive trenching.
• Hardware Failures: Many existing solutions fail in harsh agricultural environments, suffering from battery death during cold snaps or UV degradation.
• Sensor Inaccuracies: Low-cost sensors claim to measure nutrients but actually measure soil electrical conductivity, leading to high inaccuracies.
• Data Fatigue: Farmers are overwhelmed with raw data rather than actionable insights.
• Yield vs. Money Disconnect: Agritech often visualizes agronomy but fails to show economic impact.

3. Objective

To build a low-cost, scalable, and open-source plantation monitoring system for the Hardware Hackathon 3.0. Green Node merges robust hardware edge nodes with a sophisticated AI advisory pipeline to bring state-of-the-art agricultural monitoring to farmers. Our core objective focuses on plant survival, optimal yield, and early disease prevention through high-accuracy local ground truth.

4. Existing work

The current market is split between expensive agricultural tech and software-heavy platforms lacking local ground truth:

5. Gap analysis

Despite the rise of precision agriculture, competitors fail due to scaling constraints and disconnected methodologies. Existing systems lack integration between cheap continuous ambient measurement and deep-learning visual diagnostics. Solutions like Fasal capture field data but demand massive capital ($200+ unit cost), while CropIn relies purely on estimates without true real-world soil/plant mapping.

Green Node bridges this gap by decoupling continuous soil telemetry from computationally heavy vision tasks—delivering a comprehensive package for under ₹2,400 per node, combined with actionable, real-time AI-guided triaging devoid of monthly LTE service drops.

6. Novelty

The core novelty resides in our Dual-Node Architecture linked to an AI Advisory Pipeline: Instead of providing raw datasets like "humidity is 78%", Green Node features a visual capture layer. The camera node feeds leaf images into an AI model combined with live ambient data. The system generates intelligent, multi-lingual daily action priorities—e.g., "Node 3 has early signs of Late Blight — inspect row 4." Additionally, we leverage Dynamic Sentinal Polling: Sensor nodes dynamically change their reporting frequency based on internal thresholds (e.g. accelerating ping rates from 30s to 5s during severe moisture drops).

7. Your project description

Green Node is a fully integrated ecosystem designed to deliver actionable intelligence. It incorporates:

1. Hardware Edge Nodes: Custom-designed PCB routing, deploying ESP32-C3 for continuous environmental tracking (temperature, humidity, moisture, light) and ESP32-CAM nodes for visual plant health assessments.
2. Comprehensive Data Backend: Built with FastAPI and an SQLite embedded database to handle data ingestion over local Wi-Fi without internet cellular reliance.
3. AI Health Advisory Pipeline: Contextually combined sensor data and visual image scans generate daily Action Plans translated into Tamil, Telugu, Kannada, Hindi, and Bengali using Sarvam AI, complete with audio TTS capabilities.
4. Earthy Editorial Dashboard: A fully responsive bespoke dashboard providing clear, animated, high-end insights for the modern facility or farmer.

8. List of components and the software used

Hardware Components

• MCU: ESP32-C3 Dev Module (RISC-V) & AI-Thinker ESP32-CAM (ESP32-S).
• Sensors: DHT11 / DHT22 (Temperature & Humidity), Capacitive Soil Moisture Sensor v1.2, GL5528 LDR (Ambient Light), OV2640 (2MP Camera).
• Power: 5V/3.3V custom regulating PCB headers.

Software Stack

• Microcontroller Firmware: C++ (Arduino Framework), DHT sensor library, ArduinoJson.
• Backend Core: Python (FastAPI, Uvicorn, APScheduler), SQLite.
• Frontend Design: Vue.js / VitePress / HTML interfaces.
• Microservices: Gemini Vision models natively proxied through a Sarvam AI API pipeline for multilingual translation and audio rendering.

9. Methodology

To validate our system in a real-world edge scenario, we deployed it in a live indoor bonsai nursery:

• Node N-01 (Camera + Env Node): Captured on-demand 800×600 SVGA SVGA imagery alongside DHT22 inputs whenever the server requested.
• Node N-02 (Soil Node): Pushed telemetry continuously every 30 seconds via HTTP POST, adjusting frequency whenever threshold rules were broken. Data is sent explicitly through local WiFi mDNS structures routing directly to a secure, localized central Python Hub—acting as a firewall from public internet threats.

10. Work flow

1. Boot Sequence: Both nodes power up, synchronize UTC timing via NTP, and POST their IPs to vikram-Vivobook-Go.local:8000/register.
2. Measurement Cycle (Continuous): The Soil Node (N-02) wakes, tracks variables, and evaluates internal thresholds to set loop delays (5s vs 30s) before executing an HTTP POST with JSON data.
3. Capture Cycle (Pull): The central server queries the Vision Node (N-01) at GET /capture. The node streams a JPEG byte packet alongside X-Moisture and X-Temperature custom headers.
4. AI Inference & Translation: The local Hub fuses the image and historical sensor values, passing it to an AI proxy. Sarvam AI retrieves the analysis, translating the diagnostic into local languages.
5. Dashboard Reflection: The end-user operates the Vue dashboard reading the localized Daily Action Plan.

11. Pic of the project/ prototype

12. Demo video

[Insert Link to Live Project Demo Video Here]

13. Create a public github repo and attach the following

The current repository encompasses the complete documentation, firmware, server environment, and backend code designed explicitly according to the 13-point rubric constraints. Repository URL: https://github.com/Vik-Prabhu/Hardware-Hackathon-3.0