Learning Whole-Body Human-Robot Haptic Interaction in Social Contexts

Abstract

This paper presents a learning-fromdemonstration (LfD) framework for teaching human-robot social interactions that involve whole-body haptic interaction, i.e. direct human-robot contact over the full robot body. The performance of existing LfD frameworks suffers in such interactions due to the high dimensionality and spatiotemporal sparsity of the demonstration data. We show that by leveraging this sparsity, we can reduce the data dimensionality without incurring a significant accuracy penalty, and introduce three strategies for doing so. By combining these techniques with an LfD framework for learning multimodal human-robot interactions, we can model the spatiotemporal relationship between the tactile and kinesthetic information during wholebody haptic interactions. Using a teleoperated bimanual robot equipped with 61 force sensors, we experimentally demonstrate that a model trained with 121 sample hugs from 4 participants generalizes well to unseen inputs and human partners.

A HUG taxonomy of humans with potential in human–robot hugs

“…Other researchers focus on the development of humanoid robots designed for providing hugs. Yamane has developed a realistic, real-time, and comfortable remote-controlled hugging robot 11,13 , though requiring the robot operator's remote presence to hug users. Block and Kuchenbecker's team have been dedicated to developing the HuggieBot, a robot that could offer distinctly anthropomorphic and comfortable hugs 14 .…”

“…Current solutions for hugging robots either require remote operator control [11][12][13] or only provide non-reciprocating, non-adaptive, or pre-defined embrace [14][15][16] . To ensure a positive user experience, hugging robots should not only be humanoid but also understand human body language and respond accordingly.…”