While autonomous robots excel at repetitive tasks in controlled environments, the world in which most humans live is messy, constantly changing, and filled with other people. Important opportunities exist for helping humans in these unstructured environments, particularly as our population ages, but new approaches are needed for robots to be as successful inside homes, clinics, and hospitals as they already are in factories. We are thus working to discover whether and how human-robot interaction can benefit humans by designing, building, and evaluating new systems targeted at particular user populations.

To facilitate study of a variety of forms of human-robot interaction, we constructed what we call the Robot Interaction Studio in one of our laboratories. It uses a commercial markerless motion-capture system to estimate the pose of a human user in the room in real time, so that the robot being tested can act on this perceptual data. We were particularly curious about the potential benefits of allowing a robot to respond to the user's movements on a continual basis, rather than only at the end of an interaction. Though it is more complex to implement, this kind of feedback from a robot greatly benefits the interaction and deserves further exploration.

Given the importance of the sense of touch for everyday life, we are particularly interested in robots that physically interact with both objects and people to accomplish useful tasks, taking advantage of novel strategies to detect and understand contact. Sometimes the accelerometers built into a robot's grippers are sufficient for detecting impacts, but many commercial robots lack these sensors, and they also have no form of sensitive skin on their rigid links. Our work in this domain often combines the invention of novel practical tactile sensors that can enable a robot to feel contact from the humans with which it is interacting, as well as extensive human-subject studies to investigate how touch-enabled robot reactions affect the interaction.

Many physical interactions that transpire between humans, such as object handovers and hugs, have strong social dynamics that have been carefully studied. Robots that skillfully take part in such interactions may be able to work more effectively with and around humans. We are interested in how such robots should behave and how people react to them in different scenarios. To this end, we are continuing one final investigation with HuggieBot, our previously developed hugging robot.