University of Southern California
1305 Newell-Simon Hall
Refreshments 3:15 pm
Talk 3:30 pm
The potential of mobile robot teams for solving real-world problems
remains largely unexploited. In this talk, I will describe some
cooperative multi-robot approaches to traditional single-robot tasks,
such as localization, mapping and exploration. Rather than attempting
to generalize existing single-robot solutions, these cooperative
approaches exploit the availability of more than one robot to solve
tasks in qualitatively different ways. These approaches are also
motivated by the observation that while engineering the environment to
suit the task is often impossible or undesirable, no such limitations
apply when it comes to engineering the robot. We may, for example,
choose to equip our robots with unique beacons, and utilize these robots
as 'mobile landmarks'. For localization tasks, we can deploy such
robots to areas of the environment that otherwise lack useful features.
For concurrent mapping and localization (CML), robots may be used in
pairs to close loops; with two robots it is possible to disambiguate
features in partial maps, and thereby side-step the data-association
problem associated with single-robot CML.
Three specific problems (and their solutions) are described in this
talk: (1) cooperative relative localization, by which robots are able
maintain estimates of their pose relative to team-members; (2)
deployment for coverage, in which robots are dynamically placed in an
environment to form a distributed sensor network; and (3) cooperative
exploration and mapping, in which we seek both to parallelize the
exploration task, and, more significantly, to perform active loop
closure.
Dr. Andrew Howard is a Research Assistant Professor in the Computer
Science Department at the University of Southern California (USC); he
is a member of Robotics Research Laboratory (RRL) and the Center for
Robotics and Embedded Systems (CRES). His research interests include
multi-robot localization, exploration and coordination, distributed
sensor/actuator networks, and simulation of large-scale multi-robot
systems. Dr. Howard received his Ph.D. in Engineering from the
University of Melbourne in 1999, and his B.Sc. (Hons.) in Theoretical
Physics from the same institution in 1991. He served as RoboCup
coordinator at the University of Melbourne prior to joining USC in the
Fall of 2000.
For appointments, please
contact Reid Simmons (reids@cs.cmu.edu)
The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.