AI Smart Bin Sensor based on LoRaWAN

Client

A waste management technology company providing smart city sensor solutions for optimizing municipal waste collection routes.

Challenge

Accurately measuring waste bin fill level through a 60mm diameter sensor hatch, rejecting false readings from rain, snow, and birds landing on the bin rim.

Product Concept & Architecture

The STM32WLE5 LoRa SoC integrates both the application MCU and LoRa transceiver. A VL53L3CX Time-of-Flight ranging sensor measures distance to waste surface.

Engineering Solution

The firmware implements a multi-modal rejection filter: ToF readings outside expected bin geometry are discarded. A 10-sample median filter smooths the signal. When fill level crosses 80%, the reporting interval shortens from 6 hours to 30 minutes. The TinyML model predicts 'bin full' time by learning historical fill patterns per bin location.

Result

±3% fill level accuracy. 3+ year battery life at 6-hour reporting. Route optimization reduced collection costs by 30%.