Year

1993

Degree Name

Doctor of Philosophy

Department

Department of Electrical and Computer Engineering

Abstract

The conventional robotics has proved to be inflexible and non-generic. In this an alternative approach employing a distributed robotics structure is proposed studied. This method provides a uniform and generic medium for manipulation referred to as Distributed Manipulation Environment (DME). The work reported in this thesis concentrates only on the computational aspects of such environment. During the course of the thesis, the concept of Distributed Manipulation Environment is defined and a Computing Platform (CP) model is developed for it. order to match the physical architecture of DME, and also to simplify the design,the Computing Platform is proposed to be parallel and consists of loosely coupledprocessing elements, such as transputer. The Computing Platform model consists of a set of tightly and loosely coupled processes running on parallel processors. In order to improve the efficiency Computing Platform, it is critical to balance the mapping of processes on the and to minimise the communication time between the processes. This is referred to as an optimal model. Two mapping algorithms have been developed for this purpose. One referred to as Automatic Mapping Analyser (aMAP) is aimed at the tightly coupled processes and the other known as Automatic Model Generator (AMG) is developed for loosely coupled processes. The input to AMG is a non-optimal configuration of the Computing Platform model. The AMG computes an optimal or a near-optimal model which is a feasible solution to the NP-complete mapping problem using heuristic algorithms. The output of AMG is the CP model of DME that can be used as a platform for automatic configuration of the DME. The performance of AMG is studied and evaluated through modelling of a ible Manufacturing System available at the University of Wollongong

Share

COinS