Adaptive Locomotion of Soft Microrobots
Networked Control and Communication
Controller Learning using Bayesian Optimization
Event-based Wireless Control of Cyber-physical Systems
Model-based Reinforcement Learning for PID Control
Learning Probabilistic Dynamics Models
Gaussian Filtering as Variational Inference
How does HuggieBot compare to a human hugging partner?

Social touch is fundamental to human connection, providing emotional support and reducing stress. When we receive a comforting hug from a trusted person, our bodies respond with physiological changes that promote relaxation, including releasing oxytocin counterbalancing cortisol. However, in moments of acute stress, not everyone has access to a supportive human hug. This lack of social support touch inspired our work to design and iteratively improve HuggieBot [], the first interactive hugging robot with visual and haptic perception to answer the question: could a robot provide similar comfort to a human hugging partner?
In our latest study [], we investigated how different forms of social touch impact stress recovery by analyzing physiological, neurohormonal, and behavioral responses to various post-stress interactions. Participants underwent the Trier Social Stress Test (TSST), a well-established protocol for inducing acute psychosocial stress. Then, they experienced ten minutes with one of five post-stress conditions: no hug, a passive robot hug, an active robot hug (HuggieBot), a robotic woman hug, or a natural woman hug. After a rest period, all participants then experienced ten minutes with the active robot hug.
To assess the effects of these interactions, we collected salivary samples at seven different time points, which we later analyzed for oxytocin and cortisol. Participants completed surveys at eight different time points. We continuously measured their heart rate throughout the experiment. Additionally, we analyzed six hugging behavioral responses during the hug sessions to quantify how participants interacted with the hugging agent during each condition. Behavioral measurements we coded included the number of hugs participants chose to exchange with the hugging agent, the number of times participants rested their head on the hugging agent, the percent time (of the ten minutes) they spent hugging the agent, and the minimum, average, and maximum time spent hugging the agent. By combining physiological, hormonal, and behavioral data, we aim to uncover the extent to which robotic hugs can provide meaningful emotional support post-acute stress.
Members
Publications