The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

PhD Thesis #9715

Title:Accomplishing Task-Invariant Assembly Strategies by Means of an Inherently Accommodating Robot Arm
Date: 1997
Abstract:Despite the fact that the main advantage of robot manipulators was always meant to be their flexibility, they have not been applied widely to the assembly of industrial components in situations other than those where hard automation might be used. We identify the two main reasons for this as the "fragility" of robot operation during tasks that involve contact, and the lack of an appropriate user interface. This thesis describes an attempt to address these problems.We survey the techniques that have been proposed to bring the performance of current industrial robot manipulators in line with expectations, and conclude that the main obstacle in realising a flexible asssembly robot that exhibits robust and reliable behaviour is the problem of spatial uncertainty.Based on observations of the performance of position-controlled robot manipulators and what is involved during rigid-body part mating, we propose a model of assembly tasks that exploits the shape invariance of the part geometry across instances of a task. This allows us to escape from the problem of spatial uncertainty because we are no longer working in spatial terms. In addition, because the descriptions of assembly tasks that we derive are task-invariant, i.e. they are not dependent on part size or location, they lend themselves naturally to a task-level programming interface, thereby simplifying the process of programming an assembly robot.

[Search These Pages] [DAI Home Page] [Comment]