Presenters

Source

Bringing AI to the High Seas: Coral Restoration Gets a Kubernetes Boost! 🌊🤖

Ever wondered how cutting-edge technology can make a real-world impact on something as vital as coral reef restoration? Well, get ready to be inspired! Prashant from Mirantis recently shared a fascinating deep-dive into how they leveraged AI at the edge, powered by ONNX inference on K0S, to create an intelligent, self-sufficient monitoring system for a coral restoration project. It’s a story of innovation, cost-effectiveness, and the power of open-source solutions.

Let’s dive into how this incredible feat was achieved!

The Challenge: Monitoring Our Precious Corals 🐠🛠️

The core mission was to monitor a coral restoration project. This involved deploying sensors to collect crucial environmental data like:

  • Temperature 🌡️
  • Nutrient levels 🧪
  • Water flow 💧

These sensors were housed on submersible buoys, floating about 100 meters offshore and powered by solar energy. The critical requirement was to have a stable, simple, provisional tool capable of self-healing, auto-restarting, health checks, and pod rescheduling.

The “Why” Behind Edge AI: Local Predictions for Swift Action 💡🎯

You might be asking, “Why do local predictions on a buoy?” The answer is simple: speed and autonomy.

  • Real-time Nutrient Delivery: The system needed to analyze sensor data and instantly trigger pumps to deliver nutrients to the corals. Waiting for data to travel back to a main base station and then receiving commands would be too slow, potentially impacting the corals’ health.
  • Cost-Effective Solution: The entire setup needed to be low-cost and highly reliable without constant human intervention or expensive infrastructure.

The Tech Stack: A Symphony of Open-Source Power 🎶💾

To achieve these ambitious goals, a carefully selected set of technologies came together:

1. The Stable Platform: Raspberry Pi, K0S, and WasmEdge Runtime 🚀

  • Raspberry Pi: The heart of the operation, providing the necessary processing power for the buoys.
  • K0S (Kubernetes by Mirantis): This is where things get really interesting! K0S is a single binary Kubernetes cluster that’s incredibly easy to install and manage.
    • CNCF Sandbox Project: Demonstrating its growing importance in the cloud-native ecosystem.
    • Vanilla Kubernetes: Ensuring compatibility and a familiar Kubernetes experience.
    • Air-Gapped Installation: Crucially, it doesn’t connect to the internet for dependencies during installation, making it perfect for remote, off-grid deployments.
    • Production Ready & Runs Everywhere: From the edge to the cloud, K0S is designed for resilience.
    • Zero Pain, Zero Friction, Zero Cost: The promise of K0S is its simplicity and affordability.
    • Flexible Storage & Networking: Supports SQLite for local storage and offers IPv6 and custom CNI options.
    • Lightning-Fast Installation: A complete Kubernetes cluster can be up and running in under 3 minutes with just three commands!
  • WasmEdge Runtime: This enabled the execution of WebAssembly (Wasm) applications, offering a lightweight and portable way to run code at the edge.

2. The AI Engine: ONNX Models for Lightweight Inference 🧠✨

  • ONNX (Open Neural Network Exchange): This open format was a game-changer.
    • Portability & Lightweight: ONNX models are designed to be small and efficient, ideal for resource-constrained devices like Raspberry Pis.
    • Interoperability: The open format was essential for collaborating with Japanese and Vietnamese firms, ensuring seamless model sharing.

How It All Came Together: A Data-Driven Ecosystem 🌐

The architecture was elegantly designed for efficiency and autonomy:

  1. Sensor Data Collection: A Wasm application, running on the WasmEdge runtime within K0S, collected data from the various sensors.
  2. Inference on the Edge: This collected data was fed into an ONNX model running on a dedicated Raspberry Pi.
  3. Local Prediction & Action: The inference provided local predictions.
  4. Triggering Nutrient Pumps: The output of the inference was a JSON file that directly triggered GPIO pins. These pins then activated pumps, precisely delivering nutrients to the corals.

This entire process happened locally on the buoy, enabling rapid responses without relying on a constant network connection to a main base station.

The Impact: A Global Network of Intelligent Buoys 🌎🛰️

The success of this project is truly remarkable:

  • Low-Cost, Free, Open-Source Solution: Demonstrating that powerful AI solutions don’t need to break the bank.
  • Scalability: The system has been deployed on around 1,200 buoys worldwide, actively monitoring coral growth.
  • Intelligent Monitoring: Each buoy acts as an independent, intelligent unit, capable of basic predictions and automated actions.

The Future is Edge AI! 🦾📡

This project is a shining example of how Kubernetes, combined with edge AI technologies like ONNX and Wasm, can solve real-world challenges in innovative and sustainable ways. It proves that even in remote and harsh environments, we can deploy powerful, intelligent systems that make a tangible positive impact.

If you’re curious to learn more, be sure to visit Mirantis at booth 8:20! You might just discover your next groundbreaking project idea. ✨

Appendix