Cette thèse porte sur l'étude de quelques problèmes de contrôle stochastique dans un contexte de risque de liquidité et d'impact sur le prix des actifs. La thèse se compose de quatre chapitres.Dans le deuxième chapitre, on propose une modélisation d'un problème d'animation de marché dans un contexte de risque de liquidité en présence de contraintes d'inventaire et de changements de régime. Cette formulation peut être considérée comme étant une extension de précédentes études sur ce sujet. Le résultat principal de cette partie est la caractérisation de la fonction valeur comme solution unique, au sens de la viscosité, d'un système d'équations d'Hamilton-Jacobi-Bellman . On enrichit notre étude par la donnée de quelques résultats numériques.Dans le troisième chapitre, on propose un schéma d'approximation numérique pour résoudre un problème d'optimisation de portefeuille dans un contexte de risque de liquidité et d'impact sur le prix des actifs. On montre que la fonction valeur peut être obtenue comme limite d'une procédure itérative dont chaqueitération représente un problème d'arrêt optimal et on utilise un algorithme numérique, basé sur la quantification optimale, pour calculer la fonction valeur ainsi que la politique de contrôle. La convergence du schéma numérique est obtenue via des critères de monotonicité, stabilité et consistance.Dans le quatrième chapitre, on s'intéresse à un problème couplé de contrôle singulier et de contrôle impulsionnel dans un contexte d'illiquidité. On propose une formulation mathématique pour modéliser la distribution de dividendes et la politique d'investissement d'une entreprise sujette à des contraintes de liquidité. On montre que, sous des coûts de transaction et un impact sur le prix des actifs illiquides, la fonction valeur de l'entreprise est l'unique solution de viscosité d'une équation d'Hamilton-Jacobi-Bellman. On propose aussi une méthode numérique itérative pour calculer la stratégie optimale d'achat, de vente et de distribution de dividendes. / The purpose of this thesis is to study some stochastic control problems with liquidity risk and price impact. The thesis contains four chapters.The second chapter is devoted to the modeling aspects of a market making problem in a liquidity risk framework under inventory constraints and switching regimes. This formulation can be seen as an extension of previous studies on this subject. The main result is the characterization of the value functions as the unique viscosity solutions to the associated Hamilton-Jacobi-Bellman system. We further enrich our study with some numerical results.In the third section, we introduce a numerical scheme to solve an impulse control problem under state constraints arising from optimal portfolio selection under liquidity risk and price impact. We show that the value function could be obtained as the limit of an iterative procedure where each step is an optimal stopping problem and we use a numerical approximation algorithm based on quantization procedure to compute the value function and the optimal policy. The main result is to prove the convergence of our numerical scheme using monotonicity, stability and consistency properties.In the fourth section, we study a mixed singular and impulse control problem with liquidity risks and constraints. We propose a mathematical modeling to the dividend and investment policy of a firm operating under uncertain environment and liquidity risks. Our main contribution is to show that, under transaction costs and impact on the illiquid asset price, the firm's value function is the unique viscosity solution of a certain Hamilton-Jacobi-Bellman equation. We also formulated an iterative numerical procedure to compute the optimal dividend and investment policy.

No description available.

The primary goal of this thesis is to study finite element based a priori and a posteriori error estimates of optimal control problems of various kinds governed by linear elliptic PDEs (partial differential equations) of second and fourth orders. This thesis studies interior and boundary control (Neumann and Dirichlet) problems. The initial chapter is introductory in nature. Some preliminary and fundamental results of finite element methods and optimal control problems which play key roles for the subsequent analysis are reviewed in this chapter. This is followed by a brief literature survey of the finite element based numerical analysis of PDE constrained optimal control problems. We conclude the chapter with a discussion on the outline of the thesis. An abstract framework for the error analysis of discontinuous Galerkin methods for control constrained optimal control problems is developed in the second chapter. The analysis establishes the best approximation result from a priori analysis point of view and delivers a reliable and efficient a posteriori error estimator. The results are applicable to a variety of problems just under the minimal regularity possessed by the well-posedness of the problem. Subsequently, the applications of p p - interior penalty methods for a boundary control problem as well as a distributed control problem governed by the bi-harmonic equation subject to simply supported boundary conditions are discussed through the abstract analysis. In the third chapter, an alternative energy space based approach is proposed for the Dirichlet boundary control problem and then a finite element based numerical method is designed and analyzed for its numerical approximation. A priori error estimates of optimal order in the energy norm and the m norm are derived. Moreover, a reliable and efficient a posteriori error estimator is derived with the help an auxiliary problem. An energy space based Dirichlet boundary control problem governed by bi-harmonic equation is investigated and subsequently a l y - interior penalty method is proposed and analyzed for it in the fourth chapter. An optimal order a priori error estimate is derived under the minimal regularity conditions. The abstract error estimate guarantees optimal order of convergence whenever the solution has minimum regularity. Further an optimal order l l norm error estimate is derived. The fifth chapter studies a super convergence result for the optimal control of an interior control problem with Dirichlet cost functional and governed by second order linear elliptic PDE. An optimal order a priori error estimate is derived and subsequently a super convergence result for the optimal control is derived. A residual based reliable and efficient error estimators are derived in a posteriori error control for the optimal control. Numerical experiments illustrate the theoretical results at the end of every chapter. We conclude the thesis stating the possible extensions which can be made of the results presented in the thesis with some more problems of future interest in this direction.

Mathematical theory of homogenization of partial differential equations is relatively a new area of research (30-40 years or so) though the physical and engineering applications were well known. It has tremendous applications in various branches of engineering and science like : material science ,porous media, study of vibrations of thin structures, composite materials to name a few. There are at present various methods to study homogenization problems (basically asymptotic analysis) and there is a vast amount of literature in various directions. Homogenization arise in problems with oscillatory coefficients, domain with large number of perforations, domain with rough boundary and so on. The latter one has applications in fluid flow which is categorized as oscillating boundaries. In fact ,in this thesis, we consider domains with oscillating boundaries. We plan to study to homogenization of certain optimal control problems with oscillating boundaries. This thesis contains 6 chapters including an introductory Chapter 1 and future proposal Chapter 6. Our main contribution contained in chapters 2-5. The oscillatory domain under consideration is a 3-dimensional cuboid (for simplicity) with a large number of pillars of length O(1) attached on one side, but with a small cross sectional area of order ε2 .As ε0, this gives a geometrical domain with oscillating boundary. We also consider 2-dimensional oscillatory domain which is a cross section of the above 3-dimensional domain. In chapters 2 and 3, we consider the optimal control problem described by the Δ operator with two types of cost functionals, namely L2-cost functional and Dirichlet cost functional. We consider both distributed and boundary controls. The limit analysis was carried by considering the associated optimality system in which the adjoint states are introduced. But the main contribution in all the different cases(L2 and Dirichlet cost functionals, distributed and boundary controls) is the derivation of error estimates what is known as correctors in homogenization literature. Though there is a basic test function, one need to introduce different test functions to obtain correctors. Introducing correctors in homogenization is an important aspect of study which is indeed useful in the analysis, but important in numerical study as well. The setup is the same in Chapter 4 as well. But here we consider Stokes’ Problem and study asymptotic analysis as well as corrector results. We obtain corrector results for velocity and pressure terms and also for its adjoint velocity and adjoint pressure. In Chapter 5, we consider a time dependent Kirchhoff-Love equation with the same domain with oscillating boundaries with a distributed control. The state equation is a fourth order hyperbolic type equation with associated L2-cost functional. We do not have corrector results in this chapter, but the limit cost functional is different and new. In the earlier chapters the limit cost functional were of the same type.

No description available.

While stochastic optimal control, together with associate formulations like Reinforcement Learning, provides a formal approach to, amongst other, motor control, it remains computationally challenging for most practical problems. This thesis is concerned with the study of relations between stochastic optimal control and probabilistic inference. Such dualities { exempli ed by the classical Kalman Duality between the Linear-Quadratic-Gaussian control problem and the filtering problem in Linear-Gaussian dynamical systems { make it possible to exploit advances made within the separate fields. In this context, the emphasis in this work lies with utilisation of approximate inference methods for the control problem. Rather then concentrating on special cases which yield analytical inference problems, we propose a novel interpretation of stochastic optimal control in the general case in terms of minimisation of certain Kullback-Leibler divergences. Although these minimisations remain analytically intractable, we show that natural relaxations of the exact dual lead to new practical approaches. We introduce two particular general iterative methods ψ-Learning, which has global convergence guarantees and provides a unifying perspective on several previously proposed algorithms, and Posterior Policy Iteration, which allows direct application of inference methods. From these, practical algorithms for Reinforcement Learning, based on a Monte Carlo approximation to ψ-Learning, and model based stochastic optimal control, using a variational approximation of posterior policy iteration, are derived. In order to overcome the inherent limitations of parametric variational approximations, we furthermore introduce a new approach for none parametric approximate stochastic optimal control based on a reproducing kernel Hilbert space embedding of the control problem. Finally, we address the general problem of temporal optimisation, i.e., joint optimisation of controls and temporal aspects, e.g., duration, of the task. Specifically, we introduce a formulation of temporal optimisation based on a generalised form of the finite horizon problem. Importantly, we show that the generalised problem has a dual finite horizon problem of the standard form, thus bringing temporal optimisation within the reach of most commonly used algorithms. Throughout, problems from the area of motor control of robotic systems are used to evaluate the proposed methods and demonstrate their practical utility.

Backward stochastic differential equations appear in many areas of research including mathematical finance, nonlinear partial differential equations, financial economics and stochastic control. The first existence and uniqueness result for nonlinear backward stochastic differential equations was given by Pardoux and Peng (Adapted solution of a backward stochastic differential equation. System and Control Letters, 1990). They looked for an adapted pair of processes {x(t) / y(t)} / t is in [0 / 1]} with values in Rd and Rd&times / k respectively, which solves an equation of the form: x(t) + int_t^1 f(s,x(s),y(s))ds + int_t^1 [g(s,x(s)) + y(s)]dWs = X. This dissertation studies this paper in detail and provides all the steps of the proofs that appear in this seminal paper. In addition, we review (Cvitanic and Karatzas, Hedging contingent claims with constrained portfolios. The annals of applied probability, 1993). In this paper, Cvitanic and Karatzas studied the following problem: the hedging of contingent claims with portfolios constrained to take values in a given closed, convex set K. Processes intimately linked to BSDEs naturally appear in the formulation of the constrained hedging problem. The analysis of Cvitanic and Karatzas is based on a dual control problem. One of the contributions of this thesis is an algorithm that numerically solves this control problem in the case of constant volatility. The algorithm is based on discretization of time. The convergence proof is also provided.

<p>Current indirect optimal control problem (IOCP) solvers, like beluga or PINs, use symbolic math to derive the necessary conditions to solve the IOCP. This limits the capability of IOCP solvers by only admitting symbolically representable functions. The purpose of this thesis is to present a framework that extends those solvers to derive the necessary conditions of an IOCP with fully numeric methods. With fully numeric methods, additional types of functions, including conditional logic functions and look-up tables can now be easily used in the IOCP solver.</p> <p><br></p> <p>This aim was achieved by implementing algorithmic differentiation (AD) as a method to derive the IOCP necessary conditions into a new solver called Giuseppe. The Brachistochrone problem was derived analytically and compared Giuseppe to validate the automatic derivation of necessary conditions. Two additional problems are compared and extended using this new solver. The first problem, the maximum cross-range problem, demonstrates a trajectory can be optimized indirectly while utilizing a conditional density function that switches as a function of height according to the 1976 U.S. atmosphere model. The second problem, the minimum time to climb problem, demonstrates a trajectory can be optimized indirectly while utilizing 6 interpolated look up tables for lift, drag, thrust, and atmospheric conditions. The AD method yields the exact same precision as the symbolic methods, rather than introducing numeric error as finite difference derivatives would with the benefit of admitting conditional switching functions and look-up tables. </p>

Simulation modeling is important for resource management and operational strategy within the industry. Optimation AB specializes in modeling and simulation of complex systems using Dymola, but also offers solutions for decision support by solving simplified optimal control problems (OCPs). Since simulation models can be exported as functional mock-up units (FMUs), interfacing the underlying equations, this thesis explores the use of FMUs to formulate and solve OCPs in Python, proposing a workflow based on the softwares CasADi, Rockit and IPOPT. Test cases of increasing complexity, including a cogeneration plant OCP, were employed to evaluate the workflow. Promising results were obtained for simplified models, though scaling, initial guesses and solver settings require further consideration. Collocation demonstrated the fastest convergence time and overall robustness. It could be concluded that integrating FMUs into OCPs is feasible, although complex models require modifications. This suggest that creating simplified component libraries in Dymola, tailored for optimization, could improve method implementation and re-usability.

<p dir="ltr">The increased interest in deep space missions is creating an increased interest in cislunar space. The need for fast and efficient methods of traversing the lunar vicinity in creases as more spacecraft enter the region. This investigation discusses methods of low thrust transfer design in order to create low cost and low time of flight transfers. Indirect optimization is employed to compute minimum energy and minimum fuel transfers in the circular restricted three body problem. Sigmoid smoothing techniques are leveraged to ap proximate the optimal bang-coast-bang solution with continuous functions. The minimum fuel solution is employed as an initial guess to target an inertially fixed thrust direction transfer. This process is applied to a variety of cislunar orbital transfer problems. Transfers are constructed between orbits in the L1 halo, L2 halo, distant retrograde, and L4 short period orbit families. The resulting trajectories are compared to impulsive and free transfers from the literature based on the required propellant mass and time of flight.</p>