Back to articles
Wireless connectivity with embedded machine learning (ML) opens up a whole world of opportunity. Connected devices that can directly process the information they gather on-device can be used to make the planet greener, cleaner, and a better place to live.
But taking any ML IoT project from idea to production can be a major challenge. A lot of hardware and software engineering is needed just to get the tools set up before an end product can start to take shape.
Nordic’s Thingy:53 has recently joined the Thingy:52 and Thingy:91 in Nordic’s prototyping platform range. It offers a quick route to IoT prototypes and proofs-of-concept because the developer doesn’t have to worry about creating hardware integration. Instead, they can focus all their hard work on application firmware development and perfecting their embedded ML models.
But the Thingy:53 brings the developer much more than that: through Nordic’s collaboration with Edge Impulse, a U.S.-based ‘tinyML’ specialist, the prototyping platform is supplied with Edge Impulse firmware and Nordic’s nRF Edge Impulse mobile app. The firmware and the app enable developers to rapidly collect data and test embedded machine learning (ML) models on the Thingy:53. ML will be key to tomorrow’s advanced IoT edge processing.
What is Nordic Thingy:53?
The Thingy:53 is a multi-protocol IoT prototyping platform well-suited for embedded ML projects. It is equipped with Nordic’s nRF5340 SoC, the company’s most advanced multi-protocol wireless chip. The SoC features two Arm Cortex-M33 processors; the application processor is optimized for performance and the network processor is optimized for efficiency. The application processor can handle the most advanced algorithms and computational tasks associated with embedded ML.
- Read more: Meet the nRF5340, Nordic’s New Dual-Core Flagship SoC
- Read more: Why does the nRF5340 have two cores?
To cater to a wide range of embedded ML applications Thingy:53 boasts a suite of sensors, including a built-in inertial measurement unit (six-axis with accelerometer, gyroscope, and magnetometer); a low power accelerometer; temperature, humidity, air quality, and air pressure sensor; digital MEMS microphone; and color and light sensor.
Thingy:53 also includes battery charging and range extension powered by Nordic. Specifically, the product includes the nPM1100, Nordic’s first power management IC (PMIC), and the nRF21540 RF FEM, a range extender.
Making embedded ML development easier
Getting started on embedded ML designs is easier with Thingy:53. The Edge Impulse firmware collects training and test data from the prototype platform’s built-in sensors. The Thingy:53 then forwards the data using Bluetooth LE over the air to the mobile app. The mobile app then forwards the data to Edge Impulse Studio (a Cloud-based development platform for ML), where it is used to build an embedded ML model.
The resulting embedded ML model can then be deployed for testing to the Thingy:53 over the air using the same mobile app. Inferencing is performed on the Thingy:53, and the inferencing results can be viewed directly in the app to confirm that the ML model works as intended when running on the nRF5340 SoC.
The firmware and the app enable developers to take full advantage of the advanced sensors of the Thingy:53 in applications like voice recognition or movement pattern detection. The low-power accelerometer and the Thingy:53’s Pulse Density Modulation (PDM) microphone can also wake the nRF5340 from sleep on motion or sound events. This is useful for creating low-power embedded ML applications, allowing the device to remain asleep and save power when there's nothing to register and no inputs to process.
Working with Nordic’s software development kit
Thingy:53 firmware can be changed by using either a PC or Mac and the nRF Connect for Desktop software or from the nRF Programmer app on a smartphone. This allows a range of firmware samples and applications to be downloaded wirelessly over a Bluetooth LE link from a smartphone to the prototyping platform. Samples include Bluetooth mesh light switch and Bluetooth mesh sensor.
- Read more: Getting started with Thingy:53
Thingy:53 has full support in Nordic’s nRF Connect SDK (software development kit). The SDK supports firmware development for new ML devices across all product categories, from basic to highly advanced. It integrates the Zephyr RTOS and a wide range of samples, application protocols, protocol stacks, libraries, and hardware drivers.
The SDK simplifies porting modules, libraries, and drivers from one application to another, thus further reducing development time. And by enabling developers to pick and choose the essential software components for their application, they can make optimum use of memory space.
Thousands of applications
The potential use cases for IoT applications using embedded ML are limited only by the developer’s imagination. But turning those ideas into prototypes can be daunting. Thingy:53 can support almost any embedded ML application the developer can think of, from the most basic to the highly advanced. And better yet, the platform makes it easy for the engineer to turn their application idea into a prototype in short order - ready for testing and further development.
