Autonomous Robo+c Manipula+on

Transcription

1 Autonomous Robo+c Manipula+on Jing Xiao, IEEE Fellow Robo+cs Research Lab, Professor of Computer Science College of Compu+ng and Informa+cs Site Director NSF Industry/University Coopera+ve Research Center Robots and Sensors for the Human Well being University of North Carolina at CharloMe Vice President for Member Ac+vi+es IEEE Robo+cs and Automa+on Society

2 Real-time Adaptive Motion Planning [Vannoy and Xiao ]

3 Con+nuum manipulators are inspired by invertebrate structures found in nature, suitable for whole arm manipula+on three sections By the courtesy of Ian Walker

4 Progressive Grasping For n-section manipulator: Compute an initial pose with section m in contact with the object Curl sec. m+1 to n, one by one, without self-collision Force-closure check the final grasping configuration; re-adjust the spiral if needed. J. Li and J. Xiao, Progressive Generation of Force-Closure Grasps for an n-section Continuum Manipulator, IEEE IRCA, 2013.

5 Progressive grasping in clumered space Move all arm sections simultaneously in small steps towards the respective targets for the tips while avoiding collision. Achieve force-closure grasp without self collision J. Li and J. Xiao, Progressive Grasping of an n-section Continuum Manipulator in cluttered space, IEEE/RSJ IROS, 2013.

6 Constrained mo+on Raster for Scanning inspec+on (posi+on constrained) J. Li, J. Xiao, R. Grizzi, J. Lindberg, Inspection in Cluttered Space with a Continuum Manipulator, CARPI 2014.

7 3 D object detec+on and pose es+ma+on in clumered scenes Object Detec0on: Iden+fy known objects and fine their posi+ons and orienta+ons Open Challenge: iden+fy par0ally occluded objects Use RGB D Cameras to provide both 2D color and depth informa+on of a scene Build appearance based models of objects based on smooth surface segments and their visual features Conduct object detec+on and pose es+ma+on from a single image [Teng and Xiao ]

8 Pose es+ma+on results original reconstruction (left and back views) original reconstruction (three views)

9 Robo+c assembly in the presence of uncertainty Use exploratory compliant mo+on of easyto make contacts to obtain accurate pose es+mates for the required part ma+ng mo+on, even without force sensing Interleave exploratory compliant mo+on and nominal compliant mo+on [Saric, Xiao, and Shi ; Meeussen et al. 2008]

10 Hap+cs Haptic device the study to provide the tactile sensation in the manipulation of virtual objects through displaying simulated force/torque

11 Hap+c Simula+on Contacts between rigid bodies Mul+ region contact detec+on Contacts between a rigid object and an elas+c body Virtual assembly Optical fiber assembly [Chou and Xiao 2005, Luo and Xiao 2006]

12 Simulation of Grasping Convex Hollow [Cui and Xiao 2008]

13 Haptic feedback for virtual training

14 NSF Industry/University Cooperative Research Center (I/UCRC) Robots and Sensors for the Human Well-being The only NSF I/UCRC Center focused on Robotics Technologies Multiple university sites: U. Minnosota, U. Penn, U. Denver, UNC Charlotte, Purdue U.

15 Center Opera+on Conduct relevant research funded by industrial or government partners through membership fees Membership fee is minimum $35K/year NSF covers some university overhead, leaving more $$ for research Tried and true intellectual property agreement Members can access to research results of all sites

16 I/UUNC CharloCe Site Leverage NSF funding and center infrastructure and exper+se with membership contribu+ons to build trusted, long term partnerships upon shared principles and values in order to advance research of relevance to members Expand member R&D efforts with a focus on computa+ondriven robo+c and interac+ve systems augmented by sensing and data analysis, to improve the safety, capability, and wellbeing of humans Provide Access to academic exper+se and train graduate students to be poten+al recruits