What Is Embedded AI and How Is It Used in ADAS Systems?

AI is becoming increasingly common in many types of devices, from industrial equipment and smart cameras to automotive driving assistance systems. This trend has increased interest in Embedded AI and Edge AI within the embedded systems industry.

The main reason is the need for devices to analyze data and respond to events by themselves, without sending all information to external servers or cloud platforms.

For automotive applications, where cameras and sensors generate large amounts of data, processing information directly on the device helps improve response time, reduce data transfer requirements, and increase overall system efficiency.

What Is Embedded AI? 

Embedded AI is the integration of AI capabilities into embedded systems, allowing devices to analyze data, recognize patterns, and make decisions directly on the device.

Common examples include:

• - Object detection from camera images
• - Image classification
• - Sensor data analysis
• - Pattern and behavior recognition

The key concept of Embedded AI is moving AI processing closer to where data is generated, enabling faster response and more efficient operation.

Embedded AI vs. Edge AI: What's the Difference? 

Although Embedded AI and Edge AI are often mentioned together, they describe different concepts.

• - Edge AI refers to running AI processing at or near the source of data instead of sending all information to a centralized system or cloud.
• - Embedded AI focuses on integrating AI capabilities into embedded devices or systems.

In many applications, both concepts work together. Embedded AI devices act as the hardware platform that enables Edge AI applications to operate efficiently.

Why Are Automotive Systems Increasingly Using Embedded AI? 

Modern vehicles use more cameras and sensors for applications such as driver monitoring, in-cabin monitoring, and environmental sensing. These systems continuously generate data, and many applications require real-time analysis. Examples of Embedded AI applications in automotive systems include:

• - Driver monitoring systems
• - Driver drowsiness and distraction detection
• - In-cabin monitoring systems
• - Occupant detection
• - Real-time image processing from cameras

By processing data closer to the source, Embedded AI helps reduce response time and minimizes the amount of data transferred to external systems.

What Are the Challenges of Developing Embedded AI for Automotive Applications?

Although Embedded AI improves system responsiveness, real-world implementation requires consideration of several factors, including AI processing capability, camera support, power consumption, and memory limitations. As a result, semiconductor manufacturers are developing microcontrollers with integrated AI acceleration capabilities to support Embedded AI and Edge AI applications.

STM32N6 from STMicroelectronics for Embedded AI Applications 

For automotive systems such as driver assistance and in-cabin monitoring, the ability to run AI directly on the device is becoming an important factor in system design. STM32N6 is an STM32 MCU series featuring the ST Neural-ART Accelerator™, a Neural Processing Unit (NPU) developed by STMicroelectronics specifically for AI workloads.

• - The NPU helps accelerate AI model and neural network processing by reducing the workload on the main CPU. This enables faster data analysis while improving power efficiency compared with running AI processing only on the CPU.
• - With up to 600 GOPS of AI acceleration performance, the NPU inside STM32N6 supports computer vision applications and real-time image analysis, which are important capabilities for modern automotive systems.

In addition to the NPU, STM32N6 includes a dedicated Computer Vision Pipeline, Image Signal Processor (ISP), and MIPI CSI-2 interface for camera connectivity, enabling direct image processing on the device.

These capabilities help developers build Embedded AI systems that support Edge AI applications, including:

• - Driver Monitoring Systems (DMS)
• - In-Cabin Monitoring Systems
• - Computer Vision applications require real-time processing without relying entirely on cloud computing

By integrating an NPU into the MCU architecture, STM32N6 provides a platform for developing Embedded AI applications that require local AI processing directly on the device.

Summary 

Embedded AI enables devices to run AI directly on the hardware, while STM32N6 provides a platform that helps developers build Edge AI applications for automotive systems through dedicated AI acceleration and computer vision capabilities.

With features such as the ST Neural-ART Accelerator™, Computer Vision Pipeline, and camera interface support, STM32N6 helps simplify the development of intelligent automotive applications that require real-time AI processing.