Next-Gen Edge AI and Data Management

Overview

ITTIA DB Lite extends the capabilities of the STM32N6 microcontroller, enabling it to function as a secure, intelligent edge platform for data-centric applications. Leveraging STM32N6’s Cortex-M55 core with Helium vector extensions and integrated Neural Processing Unit (NPU), ITTIA DB Lite provides efficient time-series storage, compression, and querying to manage large volumes of sensor data in real time. This allows developers to reliably capture high-frequency data streams, run deterministic queries, and feed structured datasets directly into AI/ML models accelerated by the NPU. Features like transactional integrity and secure storage integrate seamlessly with the STM32N6’s TrustZone-M architecture, ensuring data reliability and protection in mission-critical environments. By transmitting only insights instead of raw streams, STM32N6 devices with ITTIA DB Lite save bandwidth, reduce cloud dependency, and deliver local intelligence for predictive maintenance, anomaly detection, and industrial automation. In short, ITTIA DB Lite transforms STM32N6 into a high-performance, AI-enabled edge database platform for next-generation IoT and embedded systems.

Case Study

The STM32N6 series introduces an Arm Cortex-M55 core with Helium (M-Profile Vector Extension) and integrated Neural Processing Unit (NPU) for efficient on-device AI acceleration. With advanced memory, high-speed interfaces, and secure architecture, STM32N6 is positioned for AI-enabled IoT, industrial automation, and smart devices. Pairing STM32N6 with ITTIA DB Lite allows developers to manage sensor and application data with structured storage, time-series queries, compression, and secure access, while also feeding AI workloads directly on the MCU. This combination enables deterministic data processing, local analytics, and AI inference at the edge—ideal for mission-critical IoT applications.

Use Case: Industrial Predictive Maintenance Gateway

Challenge

A manufacturing company needs an edge gateway that can:

• Collect vibration, temperature, and acoustic data at high frequency.
• Store weeks of time-series data for local analytics and compliance.
• Run AI models for early fault detection using the built-in NPU.
• Transmit only insights and alerts to the factory cloud system.

System Architecture

• MCU: STM32N6 with Cortex-M55 + NPU
• Sensors: Vibration (MEMS accelerometer), temperature, acoustic microphone
• Comms: Ethernet + Wi-Fi + CAN-FD
• Software Stack:
  • OS: FreeRTOS/ Eclipse ThreadX RTOS /Zephyr
  • Database: ITTIA DB Lite
  • AI Model: AI engine + STM32Cube.AI leveraging NPU acceleration  
  • Visualization: ITTIA Analitica
  • Data distribution to MPUs: ITTIA Data Connect
  • Security: TrustZone-M, hardware crypto for secure storage 

Implementation 

1. Schema Design

A time-series table captures sensor readings:

CREATE TABLE sensor_data (
    sample_time   TIMESTAMP PRIMARY KEY,
    vibration_RMS FLOAT NOT NULL,
    temperature_C FLOAT NOT NULL,
    acoustic_dB   FLOAT NOT NULL
);

CREATE TABLE anomaly_events ( 
    event_time TIMESTAMP PRIMARY KEY,
    score      FLOAT NOT NULL,
    severity   VARCHAR(32) NOT NULL
);

2.Data Logging (C API Example) 

#include "predictive_maintenance_database.h" // generated from schema

int main() {
   db_init_ex(DB_API_VER, NULL);
   open_predictive_maintenance_database("predictive_maintenance", NULL);
   // ...
}

void log_sensor(uint64_t ts_ms, float vib, float temp, float sound) {
    db_t db;
    db_connect(&db, "predictive_maintenance", "ingestion", NULL, NULL);
    const sensor_data_row_t row = {
        .sample_time   = ts_ms * 1000, // to microseconds
        .vibration_RMS = vib,
        .temperature_C = temp,
        .acoustic_dB   = sound
    };
    put_sensor_data(db, row);
    db_disconnect(db);
} 

• Data logged at 1 kHz frequency.
• Compression reduces storage footprint by ~45%. 

3. On-Device AI 

• Query retrieves last 1,000 vibration samples: 

   

  sensor_data_row_t data_array[1000];
  size_t result_count = 0;
  scan_sensor_data_by_PK_desc(db, data_array, 1000, &result_count);
  // ...
  float vib = data_array[0].vibration_RMS;
  // ...

   

• Results fed into NPU-accelerated ML model for anomaly detection.
• Detected faults logged in anomaly events table with severity level.

4. Data Distribution

• Summaries uploaded hourly over Ethernet to factory MES/SCADA system.
• Critical anomalies transmitted immediately as MQTT messages.
• Raw data remains local, reducing bandwidth >90%. 

Results 

Benefits

1. Edge intelligence: NPU-accelerated AI + ITTIA DB Lite enable real-time predictive maintenance.
2. Efficient data handling: Compression + SQL queries optimize memory and performance.
3. Secure operations: TrustZone-M and secure DB storage protect industrial IP.
4. Resilient connectivity: Works offline, shares only insights when cloud is available.
5. Scalability: Schema and architecture can be reused across STM32U5 (low power) or STM32H7 (high performance).

Conclusion

The STM32N6 + ITTIA DB Lite platform allows developers to design next-generation industrial and IoT systems that combine high-performance data management with AI inference at the edge. This makes it an ideal solution for predictive maintenance, smart factories, energy monitoring, and intelligent sensor hubs, where reliability, efficiency, and real-time insight are essential.