Search

Integrating On-chip AI/ML with BLE and Matter for Next-Gen Endpoint Devices

a controller screen shows the different smart home systems it can control, including temperature, lights, locks and others.

Smart homes are no longer only part of the future. Today’s home systems can be controlled on a smartphone, from anywhere. Image source: Shutterstock

Not long ago, controlling home systems like lights and heat and air conditioning while at work or on vacation was considered a lavish, high-tech indulgence. Now, smart homes are not just a reality, but they are also here to stay. You can program the perfect home temperature when going to bed, download washing machine presets, cook your dinner and have it ready for you when you arrive home, have your lights turn a certain color when the smoke alarm is triggered, or have the smart speakers coordinate a light show in sync with the music playing, and even schedule the vacuum robot. 

There are real-time applications in the smart home such as streaming audio to TWS headsets from smartphones and tablets, controlling VR/AR headsets through accessories, smart glasses that connect to smartphones, video game controllers, keyboards and mice that control central platforms such as gaming consoles, and PCs. Smart homes are connected, and you can control them from almost anywhere.

At the core of smart homes are endpoint devices sensing, processing, and sharing data. And the reach of this connection goes beyond smart homes. In fact, the endpoint devices industry is growing exponentially, with the number of connected devices projected to reach 29 billion by 2030, according to Statista. 

The demand for smarter, more capable devices that can process data locally and make intelligent decisions in real-time is also growing. As such, artificial intelligence and machine learning (AI/ML) capabilities are being directly integrated into endpoint devices alongside low-power wireless connectivity technologies, such as Bluetooth Low Energy (BLE) and Matter.

BLE and Matter: Enabling Low-Power, Interoperable Endpoint Connectivity

Bluetooth Low Energy (BLE) is regarded as the standard for low-power wireless communication in Edge devices. Specifically engineered to minimize power consumption, BLE is ideal for battery-operated devices, such as sensors in a variety of settings, wearable devices, and other IoT devices that need to operate for extended periods. BLE also establishes connections faster and offers low latency communication, which is essential for applications that depend on real-time updates or interactions, from smart home control systems to industrial applications. 

Matter is an open-source standard for smart home technology that enables devices to work with Matter-certified ecosystems using a unified protocol. Its main purpose is to ensure that devices from different manufacturers can interact seamlessly, providing a level of interoperability that has been challenging in the smart home industry because of the prevalence of proprietary technologies. 

While Matter itself is an application layer protocol, it is designed to work over standard wireless networking  technologies, including Ethernet, Wi-Fi, and Thread. Thread is built on the familiar and proven IEEE 802.15.4 radio technology and offers low power mesh networking or endpoint products. It is IP based, enabling secure cloud access. Matter runs on top of Thread and uses Thread’s underlying features and security capability. It builds on top of that layer with added security for Matter enabled devices. 

Combined, BLE and Matter provide a powerful foundation for low-power, interoperable endpoint devices. BLE enables efficient wireless communication between devices and gateways, while Matter ensures that devices can interact seamlessly within a unified smart home ecosystem. By leveraging both technologies, endpoint device manufacturers can design products that are energy-efficient, secure, and compatible with a wide range of devices and platforms.

Addressing the Challenges: On-Chip AI/ML 

While BLE and Matter address the connectivity and interoperability challenges of endpoint devices, these devices still need to become smarter and more capable. Typically, ML algorithms are deployed in the cloud, with endpoint devices collecting data from sensors and sending it to remote AI/ML servers for processing. 

Although commonly used, this approach has several disadvantages in low-power endpoint applications. Transmitting data to and from the cloud can make the process less efficient, increasing latency, an issue for real-time needs. Relying on cloud connectivity increases power consumption, compromising battery life, and any breaks in the network would disrupt functionality. In addition, sending sensitive data, such as biometric or location information, to remote servers raises potential data privacy and security risks.

Enter on-chip AI/ML. It enables endpoint devices to process data locally without relying on cloud connectivity. Running ML models directly on the device minimizes latency and power consumption, while also enhancing sensitive data security and privacy.

However, designing ML models for on-chip implementation brings its own set of challenges. Endpoint devices are typically resource-constrained in terms of memory and processing power; as a result, most off-the-shelf MCUs struggle to handle the computations of AI/ML algorithms. 

To address these issues, a new class of general purpose MCUs and MPUs with embedded neural processing units (NPUs) are being purpose-built to accelerate ML workloads while minimizing power consumption.

Balletto™ MCUs: Purpose-Built for the Edge

Alif Semiconductor’s Balletto™ family of wireless MCUs are the world’s first to combine a BLE 5.3+ and 802.15.4 radio subsystem supporting Matter with an Arm Cortex-M55 processor, Helium Vector Extension , and a dedicated Ethos-U55 NPU for on-device hardware-accelerated AI/ML workloads. This unique architecture enables tiny, ultra-low power endpoint devices to incorporate sophisticated AI functionality built-in with wireless connectivity.

A World of Possibilities

Balletto MCUs allow intelligent endpoint devices to process complex data locally, make decisions based on that data, and communicate those insights or actions over BLE or Matter networks, while drawing minimal power. These abilities open up a world of possibilities — but how? 

In smart homes, for example, a VR headset can recognize gestures based on advanced AI models for body function sensor fusion in the handheld controller or a wristband and wirelessly transmitted over BLE, smart security camera with on-chip ML can analyze video footage in real-time to detect and classify objects, people, and activities. Whether it’s a package being delivered or a child coming home from school, by processing the data locally, the camera can send a notification to the homeowner’s smartphone over BLE or trigger an alarm system over Matter, without relying on cloud servers. 

In industrial settings, asset trackers with predictive maintenance ML models can be attached to machinery and use on-board sensors to monitor key parameters like vibration, temperature, and power consumption. By analyzing this data with on-chip ML models, the asset tracker can detect patterns that indicate potential equipment failure. Then it can transmit this insight over a BLE/Matter network to a central system, alerting maintenance teams to schedule repairs before a failure occurs, reducing downtime, extending equipment lifespan, and lowering overall maintenance costs. 

a person controls smartfactory systems via a portable computing device.

Smart industry BLE-enabled control systems support not only cost savings, but they also represent a proactive approach to maintenance. Image source: Shutterstock

On-chip AI/ML and low-power wireless connectivity also shine in wearable devices, especially for health-related applications. A smartwatch with integrated biosensors can track the user’s heart rate, blood oxygen levels, skin temperature, and activity levels. Glucose monitors can give the user real-time values, and smart glasses can adjust focus when necessary. 

On-device ML algorithms can analyze biometric data in real-time to detect anomalies and patterns, such as an irregular heart rhythm or a sudden drop in blood sugar, to inform immediate treatment decisions. Processing the data locally also ensures user privacy, as sensitive health information is not transmitted to or stored on remote servers. 

Combining energy-efficient ML processing and low-power wireless connectivity on a single chip allows manufacturers to create intelligent and power-efficient devices for smart homes, industrial machines, or wearables. By processing data locally with on-chip ML, these devices can deliver fast, secure, and reliable performance. BLE and Matter connectivity also lets these devices communicate seamlessly with each other and with other connected devices. Alif Semiconductor’s Balletto MCUs enable on-device intelligence while maintaining power efficiency and interoperability in a wide range of applications. Learn more about On-Chip AI/ML in the white paper, “Machine Learning Transforms IoT Devices.”

X

(Required)
This field is for validation purposes and should be left unchanged.