Event Location: NSH 1305

Abstract: This work describes methods for advancing the state of the art in mobile robot navigation and physical Human-Robot Interaction (pHRI). An enabling technology in this effort is the ballbot, a person-sized mobile robot that balances on a ball.  This underactuated robot presents unique challenges in planning, navigation, and control; however, it also has significant advantages over conventional mobile robots.  The ballbot is omnidirectional and physically compliant.  Moving requires the ballbot to lean, but this also gives it the ability to achieve both soft, compliant physical interaction and apply large forces.

The completed work presented in this proposal demonstrates the ability to navigate cluttered environments with the ballbot.  Formulating the system as differentially flat enables fast, analytic trajectory planning in the space of obstacles.  Also completed is a method of helping a person stand from a seated position, applying large assistive forces with the ballbot.  Though these tasks may seem disparate, the insights from each endeavor unite to motivate the proposed work of navigating dense crowds and leading people by the hand.

Using the ballbot, we propose physically leading people by the hand to assist in navigation.  This capability could enable robotic assistance to the elderly in walking or perhaps provide useful guidance to the visually impaired in unknown environments.  

We also propose autonomous navigation through densely crowded environments.  In such situations, not only is collision with people sometimes imminent, it can be helpful.  Crowded environments pose unique challenges for robotic navigation, and we believe that by permitting and controlling contact, navigation through crowds can actually be achieved more successfully.

These new concepts in physically assistive navigation and robot planning in crowds will be formulated, implemented, and experimentally validated on the ballbot.  Qualitative and quantitative tests will be completed to assess the feasibility and tractability of these methods.  If successful, we believe they will offer an advantageous approach to physical interaction with robots.

Committee:Ralph Hollis, Chair

Jodi Forlizzi

George Kantor

Bill Smart, Oregon State University