An AI-enabled smart cycling helmet system with multisensor fusion and graded hazard warning

Abstract

Cycling safety remains a pressing concern, yet conventional helmets provide only passive impact protection without the ability to anticipate dangers. This paper presents an AI-enabled smart cycling helmet system with multi-sensor fusion and graded hazard warning based on the ESP32 microcontroller. The system integrates a tri-axial accelerometer, ultrasonic distance sensor, temperature and humidity sensor, ambient light sensor, gas sensor, and infrared wear detection sensor for comprehensive environmental and rider state awareness. A lightweight AI model deployed on the ESP32 fuses multi-source sensor data in real time to perform precise fall detection, obstacle proximity prediction with dynamically adjusted warning thresholds, and riding pattern recognition. A three-tier color-coded warning mechanism provides risk-appropriate alerts: Level 1 activates the buzzer with yellow LED flashing, Level 2 combines rapid yellow LED flashing with the buzzer, and Level 3 triggers red strobe lighting, continuous buzzer alarm, and voice broadcast, with automatic help requests sent to a mobile application. The ambient light sensor automatically activates white illumination in dark environments. A micro cooling fan is activated when temperature exceeds 30°C or harmful gas is detected, with fan speed dynamically regulated via PWM. Experimental results demonstrate a fall detection sensitivity of 95.0%, specificity of 93.0%, and risk classification accuracy of 94.0%, with end-to-end warning response latency within 500 ms. The electronic modules are integrated into a conventional helmet with weight increase below 10%, maintaining wearing comfort.