Parasol Seminar Fall 2004 | Parasol Laboratory Intranet

Parasol Seminar Fall 2004

Friday September 03, 2004. 3:00 pm
Room 302 HRBB

Network human with Robot: Scalable Algorithms and Systems for Collaborative Teleoperation

Dezhen Song.
Department of Computer Science, Texas A&M University

Abstract

Networked Robots: Systems that couple communication to control of one or more robots via networks which often include sensors and remote human operators. As one type of Networked robotic systems, collaborative teleoperation (CT) systems allow many users to simultaneously share control of a single remote physical resource such as a robot or human "explorer", with applications in education, health care, journalism, and entertainment. A primary challenge is scalable methods for computing consensus commands. In this talk, I will present results from two CT systems: one with a robotic webcamera (the Co-Opticon) and one with a human (the Tele-Actor).

In the Co-Opticon, $n$ users share control of a single robotic pan, tilt, zoom webcamera. We formulate a new resource allocation problem and a series of exact and approximate geometric algorithms for solving it. We propose a new similarity metric for the degree of satisfaction for each user, which improves over the nonlinear Intersection Over Union (IOU) metric. For an approximation bound $\epsilon$, our best algorithm runs in $O((n+1/ \epsilon^3)\log^2 n)$ time. These algorithms also apply to satellite image selection problem with applications in weather prediction, disaster response, search and rescue, surveillance, and defense applications.

The Tele-Actor system allows groups of online users to collaboratively "direct" a human exploring a remote environment. We develop an unsupervised scoring metric based on density based clustering to assess individual behavior in terms of leadership and collaboration.

I have implemented both systems and they have been extensively field tested with students and online users. Future research will focus on how new technologies such as Internet 2, broadband videoconferencing, and wireless networking can be used to enhance collaborative teleoperation and its applications in education and health care.

Some of this work is joint with Profs. Ken Goldberg, Sariel Har-Peled, Vladlen Koltun, and Frank van der Stappen.

Biography

Dr. Dezhen Song is an assistant professor in Department of Computer Science, Texas A&M University. Song completed his PhD in Industrial Engineering and Operations Research, with emphasis on Computer Science, at UC Berkeley (Advisor: Prof. Ken Goldberg, IEOR and EECS). Song received his B.S. in process control and M.S. in automation from Zhejiang University. His research areas include robotics, automation, networking, optimization, multimedia, and sensor networks. Song is a founding member of IEEE Robotics and Automation Society Technical Committee on Networked Robotics.

Department of Computer Science | Dwight Look College of Engineering | Texas A&M University