Drivbox Telemetry & Crash Detection
Edge AI sensory fusion with high-speed CAN telemetry and low-latency emergency streaming.
< 15ms
Telemetry Latency
50 Hz
Sampling Rate
- Project Overview
An embedded vehicle safety telemetry system built on Raspberry Pi and ESP32 microcontrollers. Runs quantized on-device TensorFlow Lite models to predict crash and rollover events in real-time.
- Architectural Challenge
Processing noisy IMU accelerometer arrays alongside CAN bus engine telemetry at 50Hz without dropping packets on resource-constrained embedded microcontrollers.
- Engineering Solution
Developed modular C++ firmware for the ESP32 to handle sensor filtering and high-frequency buffering, pushing telemetry frames over WebSocket channels with sub-15ms latency.