Thursday, May 24, 2012
Mechanical and Civil Engineering Seminar
Tactile Acceleration Cues for Haptic and Robotic Systems
Katherine Kuchenbecker, Assistant Professor, Mechanical Engineering and Applied Mechanics, University of Pennsylvania
When you touch objects in your surroundings, you feel a rich array of haptic cues that reveal each items identity and enable you to accomplish a wide variety of tasks. Though much research has centered on the hand's sensitivity to spatial pressure distributions and steady-state forces, the human sense of touch also excels at perceiving vibrations up to 1000 Hz. These signals are particularly prevalent at contact transition points during tool-mediated interactions, and they are detected via the Pacinian Corpuscles, mechanoreceptors that function much like omnidirectional accelerometers. Inspired by the importance of these cues, this talk will present three different projects from the Penn Haptics Group that have all benefited from the use of high-bandwidth three-axis accelerometers. First, VerroTouch uses accelerometers to enable the operator of a telerobotic surgery system to feel the contact interactions that are instantaneously occurring between each robotic instrument and the rest of the operative site. We have demonstrated the feasibility of this approach through in vivo testing on a porcine model, and we have shown that surgeons performing in vitro tasks significantly prefer the availability of this feedback. Second, the haptic modeling and rendering paradigm of haptography uses accelerometers to enable an individual to quickly capture the feel of tool-mediated contact with a real surface and then authentically recreate that experience for another user at a later time. In addition to fundamental work on highly realistic haptic textures, we are using this approach to create a haptic simulator for dental caries detection. Third, our human-inspired approach to robotic grasping uses accelerometers and tactile pressure sensors to enable the PR2 robot to handle delicate objects without crushing or dropping them. We have demonstrated the robustness of this approach through testing on fifty everyday objects, and our code has been integrated into the standard manipulation stack in ROS, which is used by many researchers worldwide.