It is often desired to have control over a process or a physical system, to cause it to behave optimally. Optimal control theory deals with analyzing and finding solutions for optimal control for a system that can be represented by a set of differential equations. This thesis examines such a system in the form of a set of matrix differential equations known as a continuous linear time-invariant system. Conditions on the system, such as linearity, allow one to find an explicit closed form finite solution that can be more efficiently computed compared to other known types of solutions. This is done by optimizing a quadratic cost function. The optimization leads to solving a Riccati equation. Conditions are discussed for which solutions are possible. In particular, we will obtain a solution for a stable and controllable system. Numerical examples are given for a simple system with 2x2 matrix coefficients.

A collection of benchmark examples is presented for the numerical solution of continuous-time algebraic Riccati equations. This collection may serve for testing purposes in the construction of new numerical methods, but may also be used as a reference set for the comparison of methods.

This is the second part of a collection of benchmark examples for the numerical solution of algebraic Riccati equations. After presenting examples for the continuous-time case in Part I, our concern in this paper is discrete-time algebraic Riccati equations. This collection may serve for testing purposes in the construction of new numerical methods, but may also be used as a reference set for the comparison of methods.

Solving the linear quadratic regulator (LQR) problem for partial differential equations (PDEs) leads to many computational challenges. The primary challenge comes from the fact that discretization methods for PDEs typically lead to very large systems of differential or differential algebraic equations. These systems are used to form algebraic Riccati equations involving high rank matrices. Although we restrict our attention to control problems with small numbers of control inputs, we allow for potentially high order control outputs. Problems with this structure appear in a number of practical applications yet no suitable algorithm exists. We propose and analyze solution strategies based on applying model order reduction methods to Chandrasekhar equations, Lyapunov/Sylvester equations, or combinations of these equations. Our numerical examples illustrate improvements in computational time up to several orders of magnitude over standard tools (when these tools can be used). We also present examples that cannot be solved using existing methods. These cases are motivated by flow control problems that are solved by computing feedback controllers for the linearized system. / Ph. D.

This work focuses on a comparison of numerical methods for linear quadratic regulator (LQR) problems defined by parabolic partial differential equations. In particular, we study various methods for computing functional gains to boundary control problems for the heat equation. These methods require us to solve various equations including the algebraic Riccati equation, the Riccati partial differential equation and the Chandrasekhar partial differential equations. Numerical results are presented for control of a one-dimensional and a two-dimensional heat equation with Dirichlet or Robin boundary control. / Ph. D.

Neste trabalho estudamos as equações de Riccati para a filtragem de sistemas lineares com saltos Markovianos a tempo discreto. Obtemos uma condição geral para estabilidade do filtro ótimo obtido pela equação algébrica de filtragem, e que também é válida para que não haja multiplicidade de soluções. Revisitamos também a questão da existência, chegando a uma condição em termos da sequência de ganhos de um observador de Luenberger. Estes resultados usaram cadeias de Markov em escala reversa de tempo, inspirando a explorar a dualidade entre filtragem e controle em sistemas com reversão na cadeia, chegando a uma relação simples de dualidade. / In this work, we studied Riccati equations for filtering Markovian jump linear systems in discrete time. We found a general condition for the stability of the optimal filter obtained via the coupled algebraic Riccati equation, and it is also valid for uniqueness of solutions. We revisit the topic of existence of solutions of the Riccati and obtain a condition in terms of the sequence of gains of a Luenberger observer. These results used Markov chains in reverse time scale, inspiring us to explore the duality between filtering and control in systems with chain reversion, arriving at a simple relation of duality.

Neste trabalho estudamos as equações de Riccati para a filtragem de sistemas lineares com saltos Markovianos a tempo discreto. Obtemos uma condição geral para estabilidade do filtro ótimo obtido pela equação algébrica de filtragem, e que também é válida para que não haja multiplicidade de soluções. Revisitamos também a questão da existência, chegando a uma condição em termos da sequência de ganhos de um observador de Luenberger. Estes resultados usaram cadeias de Markov em escala reversa de tempo, inspirando a explorar a dualidade entre filtragem e controle em sistemas com reversão na cadeia, chegando a uma relação simples de dualidade. / In this work, we studied Riccati equations for filtering Markovian jump linear systems in discrete time. We found a general condition for the stability of the optimal filter obtained via the coupled algebraic Riccati equation, and it is also valid for uniqueness of solutions. We revisit the topic of existence of solutions of the Riccati and obtain a condition in terms of the sequence of gains of a Luenberger observer. These results used Markov chains in reverse time scale, inspiring us to explore the duality between filtering and control in systems with chain reversion, arriving at a simple relation of duality.

Can we capture the explosive nature of volatility skew observed in the market, without resorting to non-Markovian models? We show that, in terms of skew, the Heston model cannot match the market at both long and short maturities simultaneously. We introduce Abi Jaber (2019)'s Lifted Heston model and explain how to price options with it using both the cosine method and standard Monte-Carlo techniques. This allows us to back out implied volatilities and compute skew for both models, confirming that the Lifted Heston nests the standard Heston model. We then produce and analyze the skew for Lifted Heston models with a varying number N of mean reverting terms, and give an empirical study into the time complexity of increasing N. We observe a weak increase in convergence speed in the cosine method for increased N, and comment on the number of factors to implement for practical use.

A collection of benchmark examples is presented for the numerical solution of continuous-time algebraic Riccati equations. This collection may serve for testing purposes in the construction of new numerical methods, but may also be used as a reference set for the comparison of methods.

This is the second part of a collection of benchmark examples for the numerical solution of algebraic Riccati equations. After presenting examples for the continuous-time case in Part I, our concern in this paper is discrete-time algebraic Riccati equations. This collection may serve for testing purposes in the construction of new numerical methods, but may also be used as a reference set for the comparison of methods.