AI at the Edge

Chapter 3. The Hardware of Edge AI

It’s now time to meet the devices, algorithms, and optimization techniques that power edge AI applications. This chapter is designed to provide a broad overview of the most important technical elements of the field. By the end of it, you’ll have the building blocks necessary to start the high-level planning of an edge AI product.

Sensors, Signals, and Sources of Data

Sensors are electronic components that give devices the power to measure their environments and detect human input. They range from extremely simple (trusty old switches and variable resistors) to mind-blowingly sophisticated (light detection and ranging [LIDAR] and thermal imaging cameras). Sensors provide our edge AI devices with the streams of data that they use to make decisions.

Beyond sensors, there are other sources of data that our devices can tap into. These include things like digital device logs, network packets, and radio transmissions. Although they have a different origin, these secondary data streams can be just as exciting as sources of information for AI algorithms.

Different sensors provide data in different formats. A few data formats are commonly encountered in edge AI applications. They can be summarized as follows:

Time series

Time series data represents the change in one or more values over time. A time series may contain multiple values from the same physical sensor—for example, a single sensor component may provide readings of both temperature and humidity. ...