Back to search

National Technical University of Athens (NTUA)

Παραμετρική βελτιστοποίηση: θεωρία, αλγόριθμοι και εφαρμογές στο βέλτιστο και τον προβλεπτικό έλεγχο στοχαστικών διακοπτικών συστημάτων και δικτυωμένων συστημάτων υπό περιορισμούς

Abstract

dc:description

One of the most important and fundamental problems in control is the development of systematic tools for the computation of controllers for general constrained non-linear and hybrid systems, that can guarantee stability, feasibility and optimality for the closed-loop system. Model Predictive Control (MPC) plays the central role in this research area, the application of which, till recently, was limited to systems with large sampling time, due its demanding on-line computations. However, recently the emergence of parametric optimization in optimal control of constrained systems allowed the analytical solution of such kind of problems, off-line, and the derivation of the explicit optimal and predictive control law. The explicit control law is usually a piecewise affine mapping, and on-line, its implementation requires merely the determination of the region of the state-space to which the current state of the system belongs. The main topic of the present doctoral thesis is the development of new algorithms for parametric optimization and their application to the explicit solution of optimal and predictive control problems for constrained stochastic switching systems and constrained networked control systems. To be more specific, in the present thesis, notions and tools stemming from Variational Analysis are employed in order to develop a new algorithm for convex parametric quadratic optimization problems. This kind of problems usually appear in model predictive control of constrained linear systems, and this new proposed approach comes to fill in the gaps of existing algorithms proposed in the literature. In particular, unlike most methods proposed in the literature, the proposed algorithm guarantees the enumeration of all the critical regions for the parametric optimization problem. Next, convex parametric piecewise quadratic optimization problems are studied, and the first systematic algorithm in the literature for their solution is proposed. Here too, the basic mathematical tool is the graphical derivative of the solution mapping. The algorithm, coupled with dynamic programming allows the explicit solution of optimal control problems for various classes of constrained systems that have never been solved analytically before. Specifically, we introduce constrained Markovian Switching Systems, a class of stochastic switching systems, where a Markov chain orchestrates the switching among its various subsystems. Appropriate notions of invariance and stability are established, and Lyapunov like sufficient conditions are derived for this kind of systems. Furthermore, finite and infinite horizon optimal control are studied, and sufficient conditions that the terminal cost and terminal set of the finite horizon optimal control problem must satisfy, in order to guarantee invariance and mean square stability for the Markovian switching system in closed-loop with the model predictive controller. Next, the thesis focuses on networked control systems, a rapidly evolving research area. Specifically, a new MPC algorithm is proposed for linear continuous-time networked systems, that guarantees constraint satisfaction in the continuous-time. Under the assumption that the time-varying transmission delay takes its values according to a Markov chain, the networked control system is cast under the Markovian switching framework, developed in the present thesis. However, since the application of explicit MPC is limited to medium scale systems due to the exponential increase in the number of critical regions with respect to the dimension of the problem, a new on-line MPC algorithm is proposed in the last part of the thesis. The proposed approach is based on the reformulation of the quadratic optimization problem as the unconstrained minimization of a smooth, convex piecewise quadratic function. A piecewise smooth Newton method with exact line search is proposed for the solution of the reformulated problem. The proposed algorithm is compared against state-of-the-art solvers in a series of benchmark problems, demonstrating considerably smaller running times. This result is very important since it renders MPC applicable to large-scale systems with small sampling times.

Degree

thesis:*
Grantor dc:publisher
National Technical University of Athens (NTUA)
Year dc:date
2010

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Patrinos, Panagiotis
  • Πατρινός, Παναγιώτης

Subjects

dc:subject × 22

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/25573

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Patrinos, Panagiotis; Πατρινός, Παναγιώτης. Παραμετρική βελτιστοποίηση: θεωρία, αλγόριθμοι και εφαρμογές στο βέλτιστο και τον προβλεπτικό έλεγχο στοχαστικών διακοπτικών συστημάτων και δικτυωμένων συστημάτων υπό περιορισμούς. National Technical University of Athens (NTUA), 2010. http://hdl.handle.net/10442/hedi/25573