CCNY Lecture Series on Computer Vision, Robotics and Human-Computer Interaction


Title: Embedded, Real-Time Control of a Resource-Constrained CRAWLER Robot

Professor  Richard M. Voyles
Department of Computer Science
University of Minnesota


Date: Wednesday, November 9, 2005
Time: 11:30 AM - 12:30 PM
Room: Steinman Hall  T-623 (EE conference room)

Abstract

Size matters for many robotic applications. Large size may be required for great reach or mass, for example, while small size may be required to access confined spaces or for stealth. As Site Director of the NSF Safety,  Security, and Rescue Research Center (SSR-RC), I am particularly interested in small robots for access to confined spaces for urban search and rescue. This is important as small size results in severe constraints on resources and drives the need for miniaturization. This talk will address of variety of projects of the SSR-RC aimed at resource-constrained robots.

The bulk of this talk focuses on work supporting the CRAWLER - a Cylindrical Robot for Autonomous Walking and Lifting during Emergency Response (a.k.a. TerminatorBot). The talk begins with a brief description of CRAWLER, itself. A soda can-size robot with dual-use limbs for locomotion and manipulation, CRAWLER's two, three-degree-of-freedom limbs can manipulate small objects with a pinch grasp, or drag the body of the robot along rough terrain. In its locomotion modes, the weight of the body is supported by its environment much like a cold-blooded reptile. The talk continues with a description of a novel metric for classifying terrain based on "gait bounce". Gait bounce is the vertical motion of objects in the visual scene resulting from the cyclic motions of the limb gaits. Gait bounce can be measured naturally by visual servoing errors with no additional sensors. From the gait bounce signature, the immediate terrain can be classified for self-adaptation or other needs. Self-adaptation is important for this robot due to its complexity. The human/computer interface can overwhelm typical users in the field. Therefore, we are developing a novel "wearable joystick" for commanding the robot in gloves-on hazardous environments. Based on these needs for motion control and computer vision, the talk concludes with preliminary work on a single-chip, embedded, multiprocessor system for real-time control and self-adaptation of the hardware/software system.

Biography

 Dr. Richard Voyles received his Ph.D. in Robotics from Carnegie Mellon University in 1997, Master degree in Mechanical Engineering from Stanford University in 1989, and B.S. degree in Electrical Engineering from Purdue University in 1983. Currently, he is the associate professor in the department of computer science and engineering at the University of Minnesota, and site Director of the NSF Safety, Security, and Rescue Research Center at University of Minnesota.


The lecture series is supported by CCNY School of Engineering, and a planning grant from NSF Minority Institutional Infrastructure program.