AffectaPocket robot reads heart rate through touch, no wearables

Ubiquitous companion robots promise immediate anxiety relief for children, but their effectiveness depends on continuously monitoring physiological states without cumbersome wearables. In a new arXiv paper (2608.04242), Turjja Datta and Morten Roed Frederiksen tackle this by embedding a photoplethysmography (PPG) sensor directly into AffectaPocket, a pocket-sized companion robot. PPG uses light to detect blood volume changes, enabling heart rate tracking through skin contact. The challenge: handheld tactile interactions generate motion artifacts that corrupt the signal. The authors built a two-stage filtering pipeline—first, an onboard Inertial Measurement Unit (IMU) rejects high-variance segments caused by movement; second, a confidence-based smoothing algorithm stabilizes recovery periods after motion. This allows the robot to autonomously estimate the child's heart rate during natural holding and petting.

The researchers validated the system in a within-subjects study with 26 participants, comparing AffectaPocket's readings against a common wrist-worn sensor. The filtering strategy significantly reduced Mean Absolute Percentage Error and achieved statistical equivalence to ground truth measurements (p<0.05). Notably, analysis of short-duration interactions revealed that the sensor requires longer stability periods to converge—meaning quick, fidgety touches are less reliable than sustained holding. Despite this limitation, the work demonstrates that self-contained physiological sensing in companion robots is feasible. A robot that can sense stress from touch could adjust its behavior in real time, unobtrusively, and reduce reliance on external wearables. This is a step toward more autonomous, emotionally responsive robots for pediatric mental health applications.