Description

A project-driven guide to designing, training, and deploying artificial intelligence directly on embedded hardware, showing how to build intelligent, autonomous systems under real-world constraints.

Embedded AI is a hands-on, hardware-first guide to building intelligent systems that run locally, in real time, on microcontrollers. Rather than focusing on theory or cloud-based workflows, the book emphasises practical design decisions and the trade-offs engineers face when deploying AI on constrained devices.

Readers begin with the foundations of embedded systems and machine learning, then progress through deep learning architectures, exploratory data analysis, and sensor data preprocessing. The book places particular emphasis on smart sensors, IMU data processing, and sensor fusion. Areas that are critical to embedded AI but rarely covered in ML textbooks. Dedicated chapters cover on-sensor machine learning using finite state machines and decision trees running directly on sensor hardware, and compressed sensing for efficient data acquisition.

Using platforms including Arduino, Raspberry Pi Pico 2, and ST smart sensor evaluation kits, readers work through complete projects including battery state-of-charge prediction, person detection using CNNs, robot arm anomaly detection, real-time audio noise suppression on a custom carrier PCB, hot word detection with a neural decision processor, and a multi-project AI MIDI synthesizer built from custom hardware. Industry-standard tools such as TensorFlow Lite for Microcontrollers and Edge Impulse are used throughout to support real deployments.

The book concludes with a biologically grounded framework for embedded intelligence. Explores modelling brainstem reflexes, limbic valence, cerebellar motor learning, and neocortical reasoning, and examines neuromorphic computing and spiking neural networks as the next frontier for on-device AI.