In the Arms of a Robot: Designing Autonomous Hugging Robots with Intra-Hug Gestures

TL;DR

Six new guidelines for designing interactive hugging robots are presented and users found the robot more natural, enjoyable, and intelligent in the last phase of the experiment than in the first, and thought robots were nicer to hug.

Abstract

Hugs are complex affective interactions that often include gestures like squeezes. We present six new guidelines for designing interactive hugging robots, which we validate through two studies with our custom robot. To achieve autonomy, we investigated robot responses to four human intra-hug gestures: holding, rubbing, patting, and squeezing. Thirty-two users each exchanged and rated sixteen hugs with an experimenter-controlled HuggieBot 2.0. The robot's inflated torso's microphone and pressure sensor collected data of the subjects' demonstrations that were used to develop a perceptual algorithm that classifies user actions with 88\% accuracy. Users enjoyed robot squeezes, regardless of their performed action, they valued variety in the robot response, and they appreciated robot-initiated intra-hug gestures. From average user ratings, we created a probabilistic behavior algorithm that chooses robot responses in real time. We implemented improvements to the robot platform to create HuggieBot 3.0 and then validated its gesture perception system and behavior algorithm with sixteen users. The robot's responses and proactive gestures were greatly enjoyed. Users found the robot more natural, enjoyable, and intelligent in the last phase of the experiment than in the first. After the study, they felt more understood by the robot and thought robots were nicer to hug.

Figures (22)

• Figure 1: The four intra-hug gestures that our hugging robot, HuggieBot, can perform, either in response to a user action or proactively when it does not detect any user actions. Despite their importance during prolonged hugs between humans, no prior hugging robot has been able to detect and respond to intra-hug gestures.
• Figure 2: Views of HuggieBot 2.0 TheSixHugCommandments ready for a hug and hugging a user. This custom human-sized hugging robot has two padded arms, an inflated torso, and a face screen mounted to a rigid frame. A camera above the screen visually senses the user at the start of the interaction, and torque sensors on the shoulder flexion and elbow flexion joints are used to embrace the user with a comfortable pressure. A microphone and pressure sensor in the back chamber of the torso are used to detect user contact and detect and classify gestures. The user ends the hug by releasing the robot's torso and/or leaning back against the arms.
• Figure 3: Front and side views of HuggieBot 2.0 TheSixHugCommandments without the robe and sweatshirt so that the HuggieChest, heating pads, and foam arm padding can be clearly seen.
• Figure 4: A matrix showing the user ratings of the appropriateness of each possible robot response to the four intra-hug actions that users performed in the action-response elicitation study. The color of each square represents the average rating, following the legend shown at right. The dots in each cell show the individual rating of each user, consistently ordered based on their average score from low to high. The pale horizontal lines in each square show the ratings of 0 (hate), 5 (neutral), and 10 (love).
• Figure 5: A matrix showing the user ratings of the quality of the four robot responses, using the same visualization approach as Fig. \ref{['fig:BehaviorMatrix']}.