Back to search

University of Sheffield

Efficient Real-Time Solutions for Nonlinear Model Predictive Control with Applications

Abstract

dc:description.abstract

Nonlinear Model Predictive Control is an advanced optimisation methodology widely used for developing optimal Feedback Control Systems that use mathematical models of dynamical systems to predict and optimise their future performance. Its popularity comes from its general ability to handle a wide range of challenges present when developing control systems such as input/output constraints, complex nonlinear dynamics multi-variable systems, dynamic systems with significant delays as well as handling of uncertainty, disturbances and fault-tolerance. One of the main and most important challenges is the computational burden associated with the optimisation, particularly when attempting to implement the underlying methods in fast/real-time systems. To tackle this, recent research has been focused on developing efficient real-time solutions or strategies that could be used to overcome this problem. In this case, efficiency may come in various different ways from mathematical simplifications, to fast optimisation solvers, special algorithms and hardware, as well as tailored auto-generated coding tool-kits which help to make an efficient overall implementation of these type of approaches. This thesis addresses this fundamental problem by proposing a wide variety of methods that could serve as alternatives from which the final user can choose from depending on the requirements specific to the application. The proposed approaches focus specifically of developing efficient real-time NMPC methods which have a significantly reduced computational burden whilst preserving desirable properties of standard NMPC such as nominal stability, recursive feasibility guarantees, good performance, as well as adequate numeric conditioning for their use in platforms with reduced numeric precision such as ``floats'' subject to certain conditions being met. One of the specific aims of this work is to obtain faster solutions than the popular ACADO toolkit, in particular when using condensing-based NMPC solutions under the Real-Time Iteration Scheme, considered for all practical purposes the state-of-the-art standard real-time solution to which all the approaches will be bench-marked against. Moreover, part of the work of this thesis uses the concept of ``auto-generation'' for developing similar tool-kits that apply the proposed approaches. To achieve this, the developed tool-kits were supported by the Eigen 3 library which were observed to result in even better computation times than the ACADO toolkit. Finally, although the work undertaking by this thesis does not look into robust control approaches, the developed methods could be used for improving the performance of the underlying ``online'' optimisation, eg. by being able to perform additional iterations of the underlying SQP optimisation, as well as be used in common robust frameworks where multi-model systems must be simultaneously optimised in real-time. Thus, future work will look into merging the proposed methods with other existing strategies to give an even wider range of alternatives to the final user.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Sheffield
Year dc:date.issued
2021

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Gonzalez Villarreal, Oscar Julian
Advisor dc:contributor.advisor
  • Rossiter, John Anthony
ORCID person:identifier.orcid

Rights

Language dc:language.dcterms_iso639_2
eng

Identifiers

dc:identifier.*
Identifier
uk.bl.ethos.837180
Dc Identifier Grantnumber
687281

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Gonzalez Villarreal, Oscar Julian. Efficient Real-Time Solutions for Nonlinear Model Predictive Control with Applications. doctoral thesis, University of Sheffield, 2021.