Parallel processing tools in adaptive and self tuning control
Parallel processing tools in adaptive and self tuning control
The accuracy and effectiveness of a control system depends greatly on the sampling rate. Along with each individual plant comes a practical minimum sampling period. If the accuracy of the control of a plant falls below certain criteria, two options are available: either a more complex plant model can be chosen so that it more accurately represents the plant trajectory; or the sampling rate can be increased. However, a more complex model is only an option if the fault lies in the model; and a faster sampling rate is only an option if the computational overheads do not dictate the minimum length of the sampling period. Most control systems will be working at the absolute limit with the sampling period dictated entirely by how long it takes to calculate a set of control inputs from the sampled input and output or state information.
Real-time solutions provide an alternative, through parallel processing, for increased accuracy of control. By finding solutions to run existing algorithms faster, more complex plant models may be implemented and/or sampling rates may be increased.
This thesis develops methods for parallelising existing adaptive control systems before introducing novel solutions to the adaptive control of linear and nonlinear plants through the use of multiple model switching schemes. Formerly non-implementable due to the computational intensity involved, these novel methods are made practical by the high degree of parallelisation that is possible in their algorithms and the much faster calculations that are therefore possible through the use of parallel processing. The parallel concepts of speed-up and scalability are introduced and used for evaluation purposes throughout. Wherever relevant, the parallel results are directly compared against the sequential ones.
University of Southampton
Brown, David Matthew
c307da0e-f2a2-4167-a49a-1ff879ad5ca1
2000
Brown, David Matthew
c307da0e-f2a2-4167-a49a-1ff879ad5ca1
Brown, David Matthew
(2000)
Parallel processing tools in adaptive and self tuning control.
University of Southampton, Doctoral Thesis.
Record type:
Thesis
(Doctoral)
Abstract
The accuracy and effectiveness of a control system depends greatly on the sampling rate. Along with each individual plant comes a practical minimum sampling period. If the accuracy of the control of a plant falls below certain criteria, two options are available: either a more complex plant model can be chosen so that it more accurately represents the plant trajectory; or the sampling rate can be increased. However, a more complex model is only an option if the fault lies in the model; and a faster sampling rate is only an option if the computational overheads do not dictate the minimum length of the sampling period. Most control systems will be working at the absolute limit with the sampling period dictated entirely by how long it takes to calculate a set of control inputs from the sampled input and output or state information.
Real-time solutions provide an alternative, through parallel processing, for increased accuracy of control. By finding solutions to run existing algorithms faster, more complex plant models may be implemented and/or sampling rates may be increased.
This thesis develops methods for parallelising existing adaptive control systems before introducing novel solutions to the adaptive control of linear and nonlinear plants through the use of multiple model switching schemes. Formerly non-implementable due to the computational intensity involved, these novel methods are made practical by the high degree of parallelisation that is possible in their algorithms and the much faster calculations that are therefore possible through the use of parallel processing. The parallel concepts of speed-up and scalability are introduced and used for evaluation purposes throughout. Wherever relevant, the parallel results are directly compared against the sequential ones.
Text
797786.pdf
- Version of Record
More information
Published date: 2000
Identifiers
Local EPrints ID: 467057
URI: http://eprints.soton.ac.uk/id/eprint/467057
PURE UUID: f6996ebd-6093-431d-9ba3-158bdb8a97e6
Catalogue record
Date deposited: 05 Jul 2022 08:10
Last modified: 16 Mar 2024 20:57
Export record
Contributors
Author:
David Matthew Brown
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics