Machine Learning with Arduino Course

Lesson 1 • Arduino IDE Setup and Navigation

Guides installation and configuration of the Arduino IDE. Students gain the toolchain fluency needed to write, compile, and upload code reliably.

Lesson 2 • Reading Sensors and Serial Output

Introduces analogRead and digitalRead for sensor input. Students learn to stream data to the Serial Monitor, enabling basic data collection.

Lesson 3 • Arduino Hardware Overview

Covers board anatomy, pin types, and power specifications. Establishes the physical foundation needed for all subsequent circuit-building activities.

Lesson 4 • Essential Electronics Concepts

Introduces voltage, current, resistance, and Ohm's Law. Provides the electrical theory required to design safe, functional circuits throughout the course.

Lesson 1 • Transferring Data to a Host Computer

Covers serial streaming and SD file transfer to a PC. Bridges the hardware data collection stage with the Python-based preprocessing pipeline.

Lesson 2 • Sampling Strategy and Data Quality

Addresses sampling frequency, noise, and missing values. Ensures collected data meets the quality standards required for reliable ML model training.

Lesson 3 • Sensor Types and Signal Characteristics

Surveys common sensor categories and their output signal types. Builds selection criteria used when designing ML data collection rigs.

Lesson 4 • Wiring Multi-Sensor Circuits

Covers I2C, SPI, and UART communication protocols for connecting multiple sensors. Students build circuits that feed several data streams simultaneously.

Lesson 1 • Python and Library Environment Setup

Configures a Python environment with NumPy, pandas, and scikit-learn. Establishes the software stack used for all preprocessing and modelling tasks.

Lesson 2 • Feature Engineering from Sensor Signals

Derives statistical and frequency-domain features from raw signals. Converts time-series sensor data into discriminative features for classifiers.

Lesson 3 • Loading and Inspecting Sensor Data

Uses pandas to load CSV logs and perform exploratory analysis. Identifies data quality issues before applying any transformations.

Lesson 4 • Cleaning and Imputing Missing Values

Applies strategies for handling nulls, duplicates, and sensor glitches. Produces a complete, consistent dataset ready for feature engineering.

Lesson 1 • Supervised Learning Fundamentals

Defines classification vs. regression and the bias-variance tradeoff. Frames algorithm selection decisions used throughout the chapter.

Lesson 2 • Decision Trees and Random Forests

Builds tree-based models suited for small, interpretable embedded classifiers. Covers pruning and ensemble methods to improve generalisation.

Lesson 3 • Support Vector Machines for Sensor Data

Applies linear and RBF kernel SVMs to sensor classification tasks. Tunes C and gamma hyperparameters using grid search.

Lesson 4 • K-Nearest Neighbors and Naive Bayes

Trains lightweight classifiers with minimal memory footprint. Evaluates their suitability for real-time inference on constrained hardware.

Lesson 1 • Actuator Response to Predictions

Triggers LEDs, buzzers, or servos based on classifier output. Completes the sense-infer-act loop for a fully autonomous embedded ML system.

Lesson 2 • Model Conversion and Size Constraints

Explains flash and SRAM limits of common Arduino boards. Guides model simplification to fit within hardware memory budgets.

Lesson 3 • Exporting Models with EloquentML

Uses the EloquentML library to auto-generate C++ model headers. Students integrate exported headers into Arduino sketches without manual coding.

Lesson 4 • Integrating Inference into Arduino Sketches

Embeds the model header and writes the inference loop in the Arduino IDE. Connects live sensor readings to the predict() function.

Lesson 1 • Gesture and Motion Recognition Project

Applies the full TFLite pipeline to classify IMU-based gestures. Consolidates neural network deployment skills in a complete end-to-end project.

Lesson 2 • Deploying TFLite on Arduino

Integrates the TFLite Micro library and model array into an Arduino sketch. Students run live inference using the interpreter API.

Lesson 3 • Neural Network Fundamentals

Covers perceptrons, activation functions, and feedforward architecture. Provides the conceptual base for building networks suited to sensor classification.

Lesson 4 • Converting Models to TensorFlow Lite

Applies post-training quantisation and exports .tflite files. Reduces model size and latency for microcontroller deployment.

Lesson 1 • Unsupervised Learning Concepts

Defines clustering, density estimation, and anomaly detection paradigms. Establishes when unsupervised methods outperform supervised alternatives.

Lesson 2 • K-Means Clustering on Sensor Data

Applies K-means to group sensor readings into operational states. Cluster assignments serve as pseudo-labels for downstream anomaly thresholding.

Lesson 3 • Deploying Anomaly Detection on Arduino

Ports threshold logic and lightweight models to Arduino sketches. Students trigger alerts when anomaly scores exceed calibrated limits.

Lesson 4 • Statistical Threshold-Based Anomaly Detection

Uses z-scores and IQR to flag sensor readings outside normal ranges. Implements lightweight detection logic deployable directly on Arduino.

Lesson 1 • Data Collection and Preprocessing Pipeline

Executes the full data acquisition and cleaning workflow for the chosen application. Produces a validated, ML-ready dataset specific to the project domain.

Lesson 2 • Documentation and Project Presentation

Produces technical documentation and delivers a live demonstration. Develops communication skills for presenting embedded ML work to technical audiences.

Lesson 3 • Project Scoping and Requirements

Defines problem statement, sensor selection, and success metrics. Produces a project brief that guides all subsequent design and build decisions.

Lesson 4 • Model Training and Optimisation

Trains, tunes, and selects the best model for the project constraints. Balances accuracy, model size, and inference speed for the target board.