STMicroelectronics AEKD-AICAR1

Overview

AEKD-AICAR1 from STMicroelectronics is a hardware development kit for automotive edge AI, integrating an SPC58EC Chorus 4 MB MCU board (AEK-MCU-C4MLIT1), SPI-based 2.8" TFT LCD (AEK-LCD-DT028V1), MEMS sensor connector board (AEK-CON-SENSOR1), and AIS2DW12 motion sensor adapter (STEVAL-MKI206V1). It runs a pretrained LSTM neural network to classify four vehicle states in real time: parking, normal road, bumpy road, and skidding.

For engineers reviewing the AEKD-AICAR1 datasheet, AEKD-AICAR1 pinout, AEKD-AICAR1 application, or AEKD-AICAR1 equivalent, this kit supports rapid prototyping of AI-driven automotive state detection with on-device inference, sensor data streaming, flash-resident fallback classification, and retrainable neural network deployment via SPC5-STUDIO-AI.

Technical Context

The kit implements a long-short term memory (LSTM) recurrent neural network directly on the SPC58EC MCU, processing raw 3-axis acceleration data from the AIS2DW12 sensor at configurable sampling rates. Classification output is rendered in real time on the integrated resistive-touch TFT display.

When the MEMS sensor is disconnected, the system automatically switches to classifying prelogged acceleration data stored in MCU flash memory - enabling robust operation under intermittent sensor conditions without host intervention or external storage.

Key Specifications

Availability

AEKD-AICAR1 is available at Aetrix Electronics and suitable for automotive ADAS validation, predictive maintenance prototyping, and edge AI algorithm development requiring stable component supply and reference-grade hardware fidelity.

Supply support for AEKD-AICAR1 includes scheduled delivery planning, volume procurement assistance, BOM continuity management, traceable sourcing, and lifecycle availability coordination for OEM customers, industrial embedded developers, connected-device designers, and electronics production programs.

Manufacturer

STMicroelectronics is a global semiconductor leader specializing in automotive, industrial, and embedded systems solutions, with deep expertise in microcontrollers, sensors, and AI-accelerated firmware tools.

The AEKD-AICAR1 belongs to ST's AutoDevKit ecosystem - designed to accelerate AI-on-the-edge development for automotive use cases by tightly coupling validated hardware, pretrained models, and MCU-optimized neural network tooling.

FAQ

What neural network architecture does AEKD-AICAR1 implement?

The AEKD-AICAR1 implements a long-short term memory (LSTM) recurrent neural network trained to classify four distinct vehicle dynamics states. The model runs natively on the SPC58EC MCU using the SPC5-STUDIO-AI runtime library, with weights and inference logic compiled into flash-resident executable code - no external AI accelerator or host processor required.

Can I replace the AIS2DW12 sensor with another ST MEMS device?

Yes - the AEK-CON-SENSOR1 board uses a standard DIL-24 socket compatible with multiple ST MEMS sensors, including LIS2DH12, LIS2DS12, and AIS3624DQ. Electrical interface and pin mapping are preserved across supported devices, allowing direct hardware substitution without PCB modification or firmware changes.

How is classification maintained when the MEMS sensor is disconnected?

Upon sensor disconnection, the AEKD-AICAR1 automatically switches to classifying pre-recorded acceleration data stored in internal MCU flash memory. This fallback mode ensures continuous state classification during sensor maintenance, calibration, or failure - preserving functional safety awareness without requiring external data sources or reinitialization.

What software tools are required to retrain or modify the neural network?

Retraining requires ST's AutoDevKit plugin (STSW-AUTODEVKIT) and SPC5-STUDIO-AI toolchain. Users import labeled sensor data, define LSTM topology, train offline (e.g., in Python/TensorFlow), then convert the model to C-compatible inference code optimized for the SPC58EC's Cortex-M3 core and memory constraints.