Ultra-Low-Power Wi-Fi and AI/ML are a Match Made in the Cloud for AIoT

Today’s AI-driven reality is seeing artificial intelligence (AI) and machine learning (ML) moving toward the very edge of the network – even the smallest of IoT devices will soon run AI/ML algorithms. This ongoing evolution is also dubbed Artificial Intelligence of Things (AIoT). In this blog, Ravi Subramanian, Product Marketing Director at Silicon Labs, explains AIoT and how the AI/ML-accelerated SiWx917 Wi-Fi SoC simplifies Edge AI for IoT device makers.

The world has been very data- and IoT-driven already for a decade or longer – most industrial processes, various types of appliances, smart homes, and smart buildings are already controlled and monitored through IoT, transmitting data between the Edge and the Cloud. Unlike the early days of M2M, where tiny data packets were sent periodically over narrow-bandwidth radio, today’s Cloud- and Ecosystem-connected IoT devices can transmit big chunks of data and stream video over Wi-Fi. However, today astronomical amounts of data are generated, and decisions need to be made fast, making Cloud computing inefficient for end-users – the latency of processing data in the Cloud is too long, and the cost of moving and storing data centrally is getting too high.

The AIoT Challenge

AIoT is an industry-wide evolution that IoT device makers are now tapping to avoid the two commonly known end-user disadvantages of Cloud computing - latency and cost. 

In the early days of IoT, devices like sensors could collect data near the source and perform simple computations locally on the device’s microprocessor to avoid processing all readings in the Cloud. However, back in the day, the lack of sheer computing power and ultra-low-power Wi-Fi connectivity limited the use cases and field operation time for resource-constrained devices.

The AIoT challenge for device makers is building high-performing, always-on cloud-connected Wi-Fi IoT devices with ample computing and AI/ML processing capacities yet operating months or even years on a battery while keeping the design and BoM costs at bay. 

Luckily, Moore’s Law is still in effect, helping Silicon Labs’ engineers squeeze ever more capabilities on a single semiconductor die. The SiWx917, which will be the lowest-power Wi-Fi 6 SoC available when it releases later this year, features an integrated AI/ML accelerator, application-dedicated microprocessor core, and heaps of memory in a single package, enabling powerful computing and fast AI/ML processing even on small battery-powered IoT devices.

Integrated Ultra-Low-Power Wi-Fi and AI/ML Acceleration  

SiWx917-powered wireless IoT devices can process data, audio, or images and apply advanced AI algorithms and fast ML inferencing to make more sophisticated decisions locally at the Edge. The benefits of ultra-low-power Edge AI computing are related to time and money (as always). Instead of going back to the Cloud for demanding computations, the IoT sensors can now handle them locally, quickly, and without initiating a Wi-Fi connection, transmitting data, and waiting for the response. 

Should Cloud computing be needed, the SiWx917 delivers ultra-low-power, always-on Wi-Fi 6 connectivity with a double battery life compared to competing SoCs. It can maintain the Wi-Fi context active on the access point (WLAN-associated mode) even when the sensor is in sleep, enabling a quick connection ramp-up and high-throughput transmission to reduce the latency and energy consumption of Cloud computing. 

Thanks to the integrated AI/ML accelerator (Matrix Vector Processor) on the SiWx917 SoC, the end-device, a sensor for example, can process data right on its Wi-Fi radio unit (i.e., SiWx917) without having to wake up and occupy its central microprocessor, reducing the end-device’s power consumption and further extending the battery life and recharging interval. 

In conclusion, ultra-low-power Wi-Fi and AI/ML are indeed a perfect match for the AIoT and the SiWx917 offers both for device makers in a single package, spiced up with an advanced sensor hub, best-in-class security, and more.   

Wi-Fi and AI/ML – A Match Made in the Cloud 

As the world becomes increasingly AI-driven, even the smallest IoT devices will be enabled with AI/ML capability. You, as a device maker, must find a fast, cost-efficient, and reliable solution to accelerate your AIoT development. 

Silicon Labs’ SiWx917 Wi-Fi 6 SoC is a complete and optimized Edge computing solution for battery-powered IoT devices. It doubles the Wi-Fi 6 battery life compared to competing SoCs without compromising on wireless performance and AI/ML computing power. For device makers, this reduces the Bill of Material and development costs, simplifies design, and improves reliability for Industrial IoT, building automation, smart home, smart cameras, and other applications.

Learn more about Silicon Labs SiWx917 and the AI/ML solution.