This project uses Arduino RP-2040 to implement a simple machine learning model to change the color of the on board tri-LED from BLUE to RED as the temperature sensed by the temperature sensor (thermistor) increases. This project uses CASP software for training the model on the PC and then transfer the trained model on to the board.
In general, this project demonstrates how to do machine learning on any Arduino board with CASP.
Steps to be followed:
Step 1. Install CASP and Arduino board support package from https://aadhuniklabs.com/?page_id=550. Please go through the below video on how to install:
Step 2. Create a simple project to check every thing is installed properly for Arduino Nano RP-2040 Connect. Please see below video for guidance.
Step 3. Use the tutorial project from the CASP installation directory to create and train a simple neural network model that changes RGB values with respect to input temperature data. Transfer the trained model on to the Arduino board. Validate and test for correct operation. All these steps are given in the below video. Please go through it.
EXPANDED TECHNICAL DETAILS
Edge AI and Neural Modeling
This project explores the deployment of Machine Learning (ML) models on low-power 8-bit microcontrollers using the CASP (Computer Aided Simulation Program) platform.
- Model Quantization and Optimization: Focuses on taking complex predictive models (like gesture recognition or predictive maintenance) and "Shrinking" them to fit into the limited 32KB Flash of an Arduino Uno.
- In-Hardware Execution: Unlike cloud-based AI, the inference happens entirely on the Arduino, resulting in sub-millisecond response times and zero dependency on a WiFi connection.
Practical Use Cases
- Smart Vibration Analysis: Can be trained to identify the "Sound" of a healthy motor vs. a failing one, providing a powerful tool for industrial preventative maintenance.