The ballbot research platform:  The project used a novel mobile robot research platform created with previous NSF funding.  The robot is the vanguard of a new class of mobile robots called "ballbots," which balance in place and move about in indoor environments using a single, spherical wheel. The particular emphasis of the research is to explore how such ballbots can interact physically with people by the exchange of forces.  This field of endeavor is referred to as physical human-robot interaction, or pHRI.  The ballbot can move at human walking speed along tortuous curved paths and is capable of navigating through a building by referring to a stored map.  Most importantly, the ballbot responds compliantly when touched or shoved, allowing safe, gentle interaction. This behavior is unique to ballbots, and is not found in traditional mobile robots. Three specific interactions were investigated: leading a person by the hand; assistance in rising from a chair; and cooperative carrying of a bulky object.

At the outset, numerous physical and computational improvements were made to the ballbot platform to support pHRI.  Among these was the addition of a pan/tilt turret on the top of the ballbot with a 3D camera, laser rangefinder, array microphone, and speakers.  We added speech dialog capability so the ballbot could speak and understand simple speech commands.  We made many changes to the ballbot's software, adopting the popular Robot Operating System to make it easier to develop new capabilities.

New motion planning algorithm: A ballbot must constantly balance and carefully maintain the relationship between its center of mass and the point of support beneath the ball wheel on the floor. Motion planning for ballbots is not straightforward because it must always respect the natural dynamics of the machine according to the laws of physics.  We invented a new method for planning motion trajectories for ballbots, based on a property called "differential flatness."  The new method allows ballbots to move gracefully and rapidly along lengthy paths inside a building, while avoiding stationary obstacles and moving obstacles such as people.

Human approaching study: In one study, we evaluated ballbot trajectories that approach people engaged in a cooperative task. We assume that robots for human assistance must be able to operate safely around people and have a safe appearance; if individuals are uncomfortable with a robot and the way it moves they may simply choose not to engage with it. We found that subjects were entirely comfortable around the ballbot and did not perceive it as dangerous, despite its 75 Kg mass.

Sit-to-stand assistance: Many older persons require assistance in rising from a chair.  Often, this assistance is provided by a  family member or care giver. It is a leading cause of back injuries. We performed experiments with the ballbot helping subjects out of chairs, where the ballbot used its ability to lean to provide assistive forces.  Following human to human experiments to assess force profiles, the assistive action was performed by the ballbot leaning up to 12 degrees, applying up to 120 N (30 lbs.) of force to help  subjects rise out of chairs.

Broader Impacts of the Work:  The research performed under this grant showed that the ballbot was capable of fast, graceful motion through cluttered environments while avoiding obstacles and people up to normal human walking speed. Performance greatly exceeds that of traditional statically stable mobile robots as well as humanoid robots.  Experiments in physical human robot interaction show the potential for using this type of robot for human assistance applications, especially for elderly or sight-i...

Please report errors in award information by writing to: awardsearch@nsf.gov.