Design Of Rubble Analyser Probe Using ML
2023 · AIP Conference Proceedings
A TinyML probe that listens for survivors inside earthquake rubble while logging environmental vitals.
The rubble analyser combines a remotely operated four-wheel platform with an embedded audio classifier. An Arduino Nano 33 BLE Sense processes sounds such as calls for help, breathing, coughs, and background noise. Table 2 of the open manuscript reports 89.83% audio classification accuracy.
Alongside audio classification, onboard sensors stream temperature, humidity, air quality, and pressure so rescuers get both human-presence cues and environmental risk in one pass into the void.
An ESP32-CAM supplies the visual feed, while the operator inspects predictions and sensor readings through the prototype workstation. The original photograph below shows the assembled electronics and drive system. The reported audio result and sensor-testing results are separate measurements; neither establishes operational rescue readiness.
Listen where a camera cannot tell the whole story.
A four-wheel probe uses remote control and an ESP32-CAM feed to help the operator inspect its surroundings.
The paper’s Table 2 reports 89.83% audio classification accuracy. Its sensor-testing aggregate is a separate result.
- TinyML
- Disaster response
- IoT