Accelerator-Equipped MCU Brings Greater Access to AI in Cars

Artificial intelligence (AI) in cars has barely scratched the surface of possible uses. Currently, AI tends to primarily target a few applications, such as self-driving or accident detection.

To make machine learning in cars more accessible, STMicroelectronics developed the Stellar P3E. According to ST, it’s the first automotive microcontroller to ship with the company’s Neural-ART Accelerator, as opposed to a similar MCU without a dedicated machine learning IP. 

By offloading AI workloads from the main cores, the neural processing unit (NPU) significantly accelerates inference execution and supports real-time, AI-based virtual sensing.

The ST Neural-ART Accelerator is embedded directly in silicon, enabling instantaneous AI inference that creates “smart” virtual sensors for higher-performance, more reliable powertrain control without additional hardware. It’s also the first time ST is using a variant of its Neural-ART Accelerator outside an STM32 device.  

Analyzing Signals at the Source

Stellar P3E with Neural-ART embedded AI acceleration is said to be able to handle complex multi-sensor data streams directly at the actuation level. This enables predictive maintenance and anomaly detection, identifying potential failures before they manifest themselves. 

For example, analyzing sensing and current signatures at the source can detect bearing wear, anode oxidation, and thermal anomalies months before they cause a failure.

For automotive AI, computational challenges remain one of the primary issues, in part because vehicles often have very strict constraints. ST notes, however, that Stellar P3E allows for virtual sensing capabilities that can emulate physical sensors in inaccessible locations (e.g., inside an engine block) to provide things such as real-time parallel control of front and rear axles without added latency.

Additional Benefits of the NPU

X-in-1 powertrain architectures require high-performance, real-time intelligence. A software-driven approach reduces hardware and complexity, boosting reliability and improving efficiency and range all without compromising real‑time behavior or safety.

In software-defined vehicles, memory is a key driver of innovation. Stellar P3E integrates extensible xMemory (provides high-density, non-volatile memory based on phase-change memory technology), delivering up to twice the density of traditional embedded flash. It helps ensure robust, automotive-qualified performance to support evolving software features throughout the vehicle’s lifetime

The new MCU features six Cortex-R52+ running at up to 500 MHz, allowing engineers to use more sensors and write more complex programs. The Stellar P3E also supports two cores in lockstep and two additional cores in lockstep or split-lock.

Lockstep is a synchronization method where processes or threads execute in a coordinated manner, maintaining the same pace. Lockstep means that both cores mirror each other, so that one immediately takes over if the other fails. Split-lock is a technique that allows processes to operate independently on different parts of data, reducing contention and improving performance.

Electric motors often run above 10,000 rpm and thus require extreme precision. Stellar P3E provides 98 channels of successive-approximation-register/sigma-delta (SAR/SD) analog-to-digital conversion (ADC), eight channels of low-power ADC, 10 integrated timer modules for jitter-free signal control, and redundant sensing. Overall, it delivers the control required for advanced applications such as dual-motor control. 

A scalable gigabit Ethernet interface is included to connect short distances while shielding the connections from noisy environments. This is particularly important when using the Stellar P3E in an X-in-1 electronic control unit (ECU) to consolidate the onboard charger, DC-DC converter, and other components into a single system. 

Samples of the MCU are available in limited quantities; ST expects full production by the end of 2026. Software development solutions, including the AI tools, are available today.