Use this URL to cite or link to this record in EThOS:
Title: Efficient parameterise solutions of predictive control
Author: Khan, Bilal
Awarding Body: University of Sheffield
Current Institution: University of Sheffield
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Access from Institution:
Model based predictive control (MPC) is well established and has gained widespread acceptance in the industry and the academic community. The success of earlier industrial heuristic MPC algorithms motivated the research community to develop several algorithms with improved performance and enlarge the region of attraction. All proposed algorithms to some extent form a trade off between a region of attraction, performance and inexpensive optimisation. This thesis makes contributions in the area of MPC algorithm design and in particular examines to what extent different methods for parameterising the degrees of freedom within the input trajectories can improve aspects of the region of attraction, performance and inexpensive optimisation. Kautz functions are explored to parameterise the input sequences in optimal model predictive control (OMPC). It is shown that this modification gives mechanisms to achieve low computational burden with enlarged region of attraction and without too much detriment to performance. The proposed algorithm based on Kautz function parameterisation guarantees stability and recursive feasibility. It is further explored and a general class of function parameterisation is proposed using higher order orthonormal basis functions. A generalised function based MPC algorithm is formulated with guaranteed convergence and recursive feasibility. The efficacy of the proposed parameterisations within existing MPC algorithms are demonstrated by examples. The general class of function parameterisation is further explored by looking at systematic choices for a particular problem. Systematic mechanisms are discussed to choose the best tuned alternative parameterisation dynamics. The numerical examples demonstrate the efficacy of the systematic mechanisms. It is also shown that generalised function parameterisations are computationally efficient when used to achieve an approximately global region of attraction as compared with OMPC, there is a reduction in number of inequalities to represent the region of attraction, the number of multiparametric solutions (and therefore computational complexity and memory storage) and also the computational time using active set methods. Another avenue explored is the efficacy of generalised function parameterisation of the degree of freedom within a robust MPC algorithm. It extends the work of nominal case to the robust scenario and shows that similar benefits accrue. An algorithm is proposed for the robust MPC using the generalised function parameterisation that enables the use of robust control invariant set to enlarge the region of attraction. Finally, the parameterised solution extends to triple mode approaches to simplify the offline computations. In triple mode MPC algorithms, the first novelty is to propose explicit choices of middle mode using generalised function dynamics as a pragmatic choice without demanding offline computations. The second novel contribution is to parameterise the input sequences for both explicit and implicit choices of the middle mode within triple mode MPC algorithms. The improvements, with respect to existing algorithms, are demonstrated by examples.
Supervisor: Rossiter, Anthony Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available