Physical Human-Robot Interaction

Physical Human-Robot Interaction

 

Physical human-robot interaction (pHRI) has been studied mostly for physical assistance and collaboration, and interactions take place mostly at end-effectors.

We have developed a method for modeling haptic interactions, specifically hugs, between a human and robot using Bayesian inference. The model is trained using the data obtained by 120 teleoperated hugs. The inputs to the model are the human motion detected by OpenPose and contact forces measured by 61 force sensors placed throughout the robot's body. The main difficulty is the temporal and spatial sparsity of the tactile information. Experimental results demonstrate that the model is able to adapt to different hug styles, duration, and strength.