Abstract for the October 24, 1997 Robotics Institute Seminar

This page is provided for historical and archival purposes only. While the seminar dates are correct, we offer no guarantee of informational accuracy or link validity. Contact information for the speakers, hosts and seminar committee are certainly out of date.

Mobile Robot Subsystems for Structured and Unstructured Environments

Alonzo Kelly
Robotics Institute
Carnegie Mellon University

4:00pm, Adamson Wing, Baker Hall

While general solutions to robotics problems remains a long term goal, it is no secret that judicious exploitation of the constraints of particular environments and applications leads to much improved performance in real-world problems.

Furthermore, it is often the case that problems which seem insurmountable at the subsystem level are mitigated or even eliminated at the system level when provided with a consistent real-world context.

In this talk, I will discuss my work in autonomous outdoor mobility and factory automation from the perspective that what was initially thought to be difficult - wasn't. In particular:

  1. Perception.
    Perception complexity is REDUCED dramatically by a consistent application of the assumption of high speed motion. I will show a PC-based stereo vision system which generates real-time 5 Hz high density range images sufficient for rough terrain mobility.

  2. Planning.
    The higher the degree of mobility limitations imposed by vehicle dynamics, the EASIER planning becomes if the assumption is used consistently. I will show how a dynamics-first approach to planning eliminates many of the problems of more general planning.

  3. Position Estimation.
    A basic technique of industrial automation is to substitute repeatability for accuracy in structured environments. This technique enables blind machinery to operate effectively by teach and playback. Unfortunately, for mobile robots, cheap repeatable position estimation has never been available. I will introduce a new approach to position estimation in structured environments based on large scale image mosaics that enables cheap teach and playback control of Autonomous Guided Vehicles.

Biographical Sketch:
Alonzo Kelly started working in Robotics in 1982 when he joined the remote manipulation systems division of Spar Aerospace in Canada working on varous space, industrial, and nuclear robotics projects.

He joined the RI Ph.D. program in 1990 concentrating on outdoor mobile robotics on the DARPA UGV program, also spending time at JPL working in stereo vision and terrain classification. He recieved his Ph.D. from, and joined the Robotics Institute in 1995.

He is currently a Project Scientist at the Robotics Institute where he works on the Ford forktruck automation program at the Robotics Engineering Consortium and the TUGV HMMWV automation program at the Field Robotics Center.

His research interests include perception, planning, control, position estimation and human interfaces for mobile robots. He is particularly interested in sensors and algorithms for mapping, obstacle avoidance, object recognition, and pose refinement because these perceptual capabilities can create useful mobile robots with today's technology.

For appointments, please contact the speaker: alonzo@frc.ri.cmu.edu

Last Modified on: Tue Oct 14, 1997

The RI Seminar Home Page: www.frc.ri.cmu.edu/~mcm/seminar.html

Martin C. Martin, <mm+@cmu.edu>