Industrial Platform Lift Safety System (ADAS MVP)

Designed and delivered an MVP Advanced Driver Assistance System (ADAS) for industrial platform lifts — combining embedded real-time control, multi-zone spatial sensing, and AI vision to enforce safety shutdowns before accidents occur.

The system runs on a distributed architecture: an ESP32-S3 custom PCB handles all real-time hardware — polling three advanced ToF sensors (left, right, top) over I2C, executing safety decision logic, and driving a 30A relay to physically cut platform motor power within <300ms of hazard detection. A Raspberry Pi 5 runs a Linux daemon hosting a YOLOv26-powered Python pipeline for operator presence monitoring, boundary violation detection, and mobile phone distraction detection — communicating back to the ESP32 via USB/UART.

Three safety triggers are enforced in parallel: an operator stepping outside the defined control boundary, a detected distraction event (phone usage sustained >2-3 seconds), and proximity intrusion flagged by ToF sensors at ±2-3 cm accuracy across a 4-metre range. A Flask web GUI provides live sensor readouts, configurable thresholds, and AI confidence tuning. A manual override button with time-bound bypass handles false-positive edge cases. The system architecture aligns with ISO 13849 SRP/CS input-logic-output safety function decomposition.