IoT Embedded Computing: The next wave of Smart Things

Smart Lifestyle

IoT Embedded Computing: The next wave of Smart Things

Look inside any of the billions of smart things connected to the IoT and what would you see? Most will have one or more microcontrollers (MCUs) to do some or all of the “thinking” for these connected objects.

by Michele Scarlatella

An MCU takes information it collects locally using sensors and perhaps sends it off to a centralized computer for storage or remote actions. Maybe the MCU combines it with information collected at other nodes in order to do something useful. Consider a simple example: a smart thermometer in the home. With one in every room measuring temperature, these might turn a space heater or window air conditioner (AC) on or off, or they could operate centrally to control a more efficient zone heater or AC.

So far, so good. Now think of a more complex situation in our information overloaded society where having smart things to think for us can be extraordinarily convenient and efficient. At a municipal level, smart street lights that adapt to cloud cover, sun or moon brightness, and pedestrian traffic or vehicle-mounted sensors, connected to navigation systems, that identify roadways in need of repair and communicate that information to the appropriate maintenance crew.

The makers of smart and connected things are getting wise to the need for security in their products

Michele Scarlatella
Communication Director, Microcontrollers & Digital ICs at STMicroelectronics

Michele Scarlatella STmicroelectronics


All of these embedded systems, among thousands of others, are specialized computing platforms that rely on one or more MCUs in each node to provide distributed computing capabilities. All the while, MCUs are evolving to support the next wave of connected objects and their interactions with increasingly complex networks of devices and services. This evolution is traveling along the usual path of providing higher computing capabilities with more storage, better graphics and media processing, and higher-featured, integrated peripherals. Today, there are two additional demands that have not traditionally been associated with microcontrollers – embedded security and artificial intelligence (AI). The makers of smart and connected things are getting wise to the need for security in their products. The type of security required depends on the nature of the “thing” or the value of its knowledge and data, and its potential for harm. Fitness monitors, for example, may not contain very valuable information locally but they could be hacked and used to give backdoor access to a network. You don’t want hackers crawling around in the contents of your smartphone, and you certainly don’t want them taking control of a car while you, or the person next to you, is driving.

Fortunately, chip vendors like STMicroelectronics are stepping up to meet these demands for rising levels of security. ST provides a flexible security-product portfolio ranging from the integrated security features in its 32-bit microcontrollers to the highest levels of security based on dedicated secure elements. The combination of a general-purpose microcontroller and secure element is a solution that reaches the very highest safeguarding requirements for IoT applications, while at the same time simplifying the product architecture, security validation, and certification. Putting all critical features into a single, highly protected device is much simpler than having security distributed inside a very complex product.

On its STM32 MCUs and in ST’s automotive MCUs, the company has begun to embed security using a range of recognized practices, including countermeasures against remote software and board-level attacks, trusted execution environment (TEE) capabilities, self-evaluation solutions, and integrated one-time programmable (OTP) memory. It also produces embedded hardware security modules (eHSMs) that provide cryptographic processing and can safeguard and manage digital keys for strong authentication.

For even stronger protection, ST has a range of secure elements. The STSAFE family is one of these important building blocks for highly secure applications. STSAFE is a complete ecosystem of independently certified turnkey solutions designed to ensure device identity as well as system and network integrity. Devices from this family are easily linked to general purpose MCUs to deliver enhanced security or they can be used standalone for authentication or security in consumables covering a wide range of applications and use cases. MCUs in embedded systems aren’t only increasing their raw processing power. They are also learning to do things in a smarter way. This is where artificial intelligence comes in. AI is a key part of the evolution of the Internet of Things and is necessary to make many of the IoT services and applications work, as well as being a driver for the kinds of IoT devices that will exist in the future.

IoT Embedded Computing: STM32 Microcontroller ST
STM32 Microcontroller ST
provides a flexible security-product portfolio ranging from the integrated security features in its 32-bit microcontrollers to the highest levels of security based on dedicated secure elements.

AI is rapidly gaining ground in a very large number of products and use cases. In some, AI processing is done by powerful processing machines located in the cloud, taking in and operating on vast amounts of data and spitting out answers. But AI processing is also quickly spreading down into the nodes themselves, performing dedicated applications without the transmission-time delays, bringing the advantages of real-time response and lower power requirements that result from on-demand (rather than permanent) connectivity. AI embedded into local devices holds great promise for a wide range of applications like predictive maintenance and real-time object detection, classification, and tracking, among others, where it might be applied in smart factories or in smart vehicles.

Today, general-purpose MCUs, like the STM32 family, can be used to run AI algorithms and neural-network models, using mapping tools that allow these features to be adapted to existing architectures. In the next step specialized hardware will be available to execute these programs more efficiently, allowing AI capabilities to be brought down to the lowest node in the network in an economical and energy-efficient manner.

Leave a Reply

Your email address will not be published. Required fields are marked *