Multimodal robots gain contactless breathing monitor for rescue ops
Robots can now measure your breathing rate from 8 meters away, even in pitch black.
Deep Dive
A modality-adaptive contactless respiratory monitoring framework for heterogeneous mobile robots uses onboard edge computing, fusing RGB, thermal, NIR, and low-light cameras with keypoint-guided chest ROI extraction and SQI-based filtering. Tested on quadruped and wheeled platforms, RGB works up to 8m, NIR up to 6m, low-light up to 8m in complete darkness, and thermal only at short range. The framework generalizes across platforms without per-platform retuning, enabling autonomous triage in hazardous search-and-rescue settings.
Key Points
- Fuses RGB, thermal, NIR, and low-light cameras with brightness-adaptive sensor selection for variable lighting
- Keypoint-guided chest ROI extraction handles posture changes; SQI filtering ensures reliable respiratory rate estimates
- Tested on quadruped and wheeled robots without per-platform tuning; RGB and low-light both achieve 8m range, NIR up to 6m, thermal short-range only
Why It Matters
Enables autonomous, contactless triage in disaster zones, reducing responder risk and improving operational safety.