Operation AIRFlag: ESP32-Based Risk Monitoring with Local LLM-Assisted Diagnostics

Abstract

Operation AIRFlag is an educational monitoring prototype that integrates an ESP32-based sensor node, a Python telemetry gateway, deterministic risk scoring, a Streamlit dashboard, and optional locally hosted large language model (LLM)-assisted diagnostics.

The prototype uses DHT11 temperature-humidity, BMP280 pressure, MPU6050 inertial measurement, and infrared obstacle sensors. The ESP32 transmits structured telemetry, which is validated and processed by a Python backend. The backend calculates application-specific risk scores, records events in SQLite, and presents monitoring information through a dashboard. The LLM component provides non-authoritative diagnostic explanations.

The primary telemetry evaluation comprised 450 records collected over 898.063 seconds, with all 450 records passing the configured validity checks. The observations included 439 LOW, two GUARDED, and nine HIGH risk-state records. A separate benchmark evaluated three locally hosted language models across 162 synthetic executions.

The results demonstrate the integration of embedded sensing, telemetry validation, deterministic risk assessment, event logging, and experimental LLM-assisted explanations. This is a laboratory proof-of-concept and does not establish sensor calibration accuracy, validated aviation hazard detection, or certification for aircraft use.