Norwegian wireless chip designer Nordic Semiconductor has signed a partnership with Edge Impulse in the US to add machine learning and embedded AI to its Bluetooth Low Energy wireless chips.
The deal allows AI frameworks to run on the nRF52 and nRF53 Series controllers to model and understand their operating environments and make intelligent decisions. This is part of a key move to add AI frameworks to smaller, resource-constrained microcontrollers and the deal marks the first time this is available fro Bluetooth controllers. Several companies are developing extensions for RISC-V processor for embedded AI, and ARM is driving its own embedded AI add-on blocks.
The nRF52840 Development Kit now provides access to AI and machine learning using Impulse’s Edge Optimized Neural (EON) compiler. This optimises computer processing and memory use by up to 50 percent when running TinyML on resource-constrained semiconductor devices.
“What AI and machine learning on resource-constrained chips does – which Nordic will now collectively refer to as TinyML – is take the application potential of wireless IoT technologies such as Bluetooth to a whole new level in terms of environmental awareness and autonomous decision making,” said Kjetil Holstad, Nordic’s Director of Product Management.
“Although we have had customers build and run TinyML applications on Nordic’s Bluetooth chips in the past, before now this required quite a high level of mathematical and computer programming expertise using professional science industry and academia software like MATLAB.”
Next: Embedded AI example libraries
One example of the above is two successful projects in the Hackster.io and Smart Parks backed ‘ElephantEdge’ wildlife tracker challenge that employed Nordic’s nRF52840 System-on-Chip (SoC). These included a design by Dhruv Sheth called ‘EleTect’, a TinyML and IoT smart wildlife tracker employing the nRF52840 SoC as well as an accelerometer, camera and microphone. The models are ready for deployment in three forms including a C++ library, Arduino library, and OpenMV library all available on GitHub.
“What our partnership with Edge Impulse will do is remove all the complexity and previous technological barriers-to-entry for our customers wishing to add TinyML features to their Bluetooth applications,” said Holstad. “In fact using Edge Impulse tools, Nordic customers could be up and running TinyML on their applications within an afternoon. And at an ultra-low power consumption level that still supports extended battery operation, even from small batteries.”
Holstad says prime engineering areas for TinyML include but are not limited to audio and vibration where it can be used to establish normal operating patterns and rapidly detect anomalies. Example applications include anti-poaching (listening for gun shots), predictive and preventative maintenance (listening for tell-tale changes in the vibration signature of a public escalator or lift), and utilities (power line failure detection after a storm). But Holstad says all Nordic customer applications stand to benefit from TinyML from asset tracking to wearables.
“What Nordic Semiconductor is doing through its partnership with Edge Impulse is bringing AI and machine learning to the wireless IoT masses,” says Edge Impulse Co-Founder and CEO, Zach Shelby. “By leveraging the fact that every Nordic nRF52 and 53 Series Bluetooth SoC employs at least one powerful Arm core processor on-board, and is architecturally designed for ultra-low power battery operation, this partnership is effectively democratizing access to state-of-the-art TinyML within the Bluetooth market. Given the powerful application benefits of TinyML, this is going to help make the world a lot more reliable and a lot safer.”
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.