Graduate Exam Abstract
July 31, 2006, 10 am
Failure-Tolerant Path Planning for Redundant Manipulators
Abstract: This work addresses the issue of finding locations in a
robot's workspace that are guaranteed reachable in the
presence of joint failures. This guarantee is useful for
manipulators performing tasks in remote or hazardous
environments that preclude direct human intervention, and for
tasks where a robot failure could possibly pose a significant danger.
The first part of this work considers start and goal workspace
locations that are guaranteed reachable in the presence of joint
failures and obstacles in the workspace. The method of finding a
solution is to find a set of possible obstacle-free paths
that guarantees the existence of a solution. It is shown that this
set makes up a simply-connected, obstacle-free surface with no
internal local minimum or maximum in the manipulator's configuration
space. A necessary condition and a sufficient condition are presented
that state the requirements to possibly find a solution.
The second part of this work considers the problem of identifying a
region in the workspace where failure tolerance can be achieved.
This is performed by finding a suitable set of artificial joint limits
of the robot such that the failure-tolerant workspace region exists.
Characterization of the self-motion manifolds that identify the boundaries
of this region is presented. Once a failure-tolerant workspace region
is identified, robot tasks can be specified within this region.
Adviser: Anthony Maciejewski
Non-ECE Member: Wade Troxell, ME
Member 3: Iuliana Oprea, Math
Addional Members: Rodney Roberts
Program of Study: