Thesis (Ph. D.)--University of Texas at Austin, 2004. / Supervisor: Philip J. Morrison. Vita. Includes bibliographical references.

- / -

Recent work has established that a group theoretical viewpoint of completely integrable dynamical systems with N degrees of freedom yields an algorithm that provides new information concerning the symmetry transformation group structure of this class of dynamical systems. The work presented here rests heavily on the results presented in reference and it is recommended that the reader consult this reference for a more rigorous discussion of the results given in this thesis.

Energy-shaping techniques are used to expand the range of autonomous motion of unmanned aerial systems without prohibitively {color{black}increasing the computational cost of the resultant controller}. Passivity-based control presents a method to implement a static, nonlinear state feedback control law that stabilizes the motion of an aircraft with a large region of attraction. {color{black} The energy-based control scheme is applied to both multirotor and fixed-wing aircraft}. Multirotor aircraft dynamics are cast into a port-Hamiltonian System and the concept of trajectory tracking using canonical feedback transformation is implemented to construct a cross-track controller. Fixed-wing aircraft dynamics are cast in port-Hamiltonian form and a passivity-based nonlinear control law for steady, wings-level flight of a fixed-wing aircraft to a specified inertial velocity (speed, course, and climb angle) is constructed. Results in simulations and experiments suggest robustness, and a large region of attraction of the controller. The control law extended to support time-varying inertial velocity tracking that incorporates banking to turn. The results are extended by including a line-of-sight guidance law and varying the direction as a function of position relative to a desired path, rather than as a function of time. The control law and the associated proof of stability follow similarly to that of the time-varying directional stabilization problem. The results are supported with simulations as well as experimental flight tests. / Doctor of Philosophy / This dissertation presents an alternative but intuitive approach to regulate unmanned aerial vehicles' flight that would allow for more maneuverability {color{black} than conventional methods}. This scheme relies on modifying the energy of the system to achieve the desired motion and leverages the properties of the aircraft rather than eliminating them and imposing different properties. This approach is applied to both fixed-wing and aircraft and quadcopters. Simulations and experimental flights have show the efficacy of this approach compared to other more established methods.

The problem considered here is one of finding the minimal Hamiltonian chain of a graph. A single chain must traverse all 𝑛 vertices of a graph with the minimal distance. The proposed procedure reduces a large problem into several smaller problems and uses a branch and bound algorithm to find the minimal Hamiltonian chain of each partitioned subproblem. The graph is decomposed and partitioned into subproblems with the use of necessary conditions for the existence of a Hamiltonian chain. This process is only applicable to graphs with relatively few incident edges per vertex. The branch and bound algorithm makes use of concepts developed by Nicos Christofides. Hamiltonian chains are derived by using minimal spanning trees. / Master of Science

Dans cette thèse, on s'intéresse à deux aspects différents des systèmes dynamiques intégrables. La première partie est dévouée à l'étude de trois familles de systèmes hamiltoniens intégrables : les systèmes de pliage de Kapovich et Millson sur les espaces de modules de polygones 3D de longueurs de côtés fixées, les systèmes de Gelfand-Cetlin introduits par Guillemin et Sternberg sur les orbites coadjointes du groupe de Lie U(n), et une famille de systèmes définie par Nohara et Ueda sur la variété grassmannienne Gr(2,n). Dans chaque cas on montre que les fibres singulières de l'application moment sont des sous-variétés plongées et on en donne des modèles géométriques sous la forme de variétés quotients. La deuxième partie poursuit une étude initiée par Zung et Minh sur les actions totalement hyperboliques de Rn sur des variétés compactes de dimension n, qui apparaissent naturellement lors de l'étude des systèmes non-hamiltoniens intégrables dont toutes les singularités sont non-dégénérées. On s'intéresse au flot engendré par l'action d'un vecteur générique de Rn. On donne une définition d'indice pour ses singularités qu'on relie à la théorie de Morse classique, et on utilise ce flot pour obtenir des résultats sur le nombres d'orbites de dimension donnée. Une étude plus poussée est effectuée en dimension 2, et en particulier sur la sphère S2, où les orbites de l'action dessinent un graphe plongé dont on analyse la combinatoire. On termine en construisant explicitement des exemples d'actions hyperboliques en dimension 3 sur la sphère S3 et dans l'espace projectif RP3. / In this thesis, we are interested in two different aspects of integrable dynamical systems. The first part is devoted to the study of three families of integrable Hamiltonian systems: the systems of bending flows of Kapovich and Millson on the moduli spaces of 3D polygons with fixed side lengths, the Gelfand-Cetlin systems introduced by Guillemin and Sternberg on the coadjoint orbits of the Lie group U(n), and a family of integrable systems defined by Nohara and Ueda on the Grassmannian Gr(2,n). In each case we prove that the fibers of the momentum map are embedded submanifolds for which we give geometric models in terms of quotients manifolds. In the second part we carry on with a study initiated by Zung and Minh of the totally hyperbolic actions of R^n on compact n-dimensional manifolds that appear naturally when investigating integrable non-hamiltonian systems with nondegenerate singularities. We study the flow generated by the action of a generic vector in Rn. We define a notion of index for its singularities and we use this flow to obtain results on the number of orbits of given dimension. We investigate further the 2-dimensional case, and more particularly the case of the sphere S2, where the orbits of the action draw an embedded graph of which we analyse the combinatorics. Finally, we provide explicit examples of totally hyperbolic actions in dimension 3, on the sphere S3 and on the projective space RP3.

There are many problems that lead to analysis of dynamical systems with phase variables of two types, slow and fast ones. Such systems are called slow-fast systems. The dynamics of such systems is usually described by means of different versions of perturbation theory. Many questions about accuracy of this description are still open. The difficulties are related to presence of resonances. The goal of the proposed thesis is to establish some estimates of the accuracy of the perturbation theory for slow-fast systems in the presence of resonances. We consider slow-fast Hamiltonian systems and study an accuracy of one of the methods of perturbation theory: the averaging method. In this thesis, we start with the case of slow-fast Hamiltonian systems with two degrees of freedom. One degree of freedom corresponds to fast variables, and the other degree of freedom corresponds to slow variables. Action variable of fast sub-system is an adiabatic invariant of the problem. Let this adiabatic invariant have limiting values along trajectories as time tends to plus and minus infinity. The difference of these two limits for a trajectory is known to be exponentially small in analytic systems. We obtain an exponent in this estimate. To this end, by means of iso-energetic reduction and canonical transformations in complexified phase space, we reduce the problem to the case of one and a half degrees of freedom, where the exponent is known. We then consider a quasi-linear Hamiltonian system with one and a half degrees of freedom. The Hamiltonian of this system differs by a small, ~ε, perturbing term from the Hamiltonian of a linear oscillatory system. We consider passage through a resonance: the frequency of the latter system slowly changes with time and passes through 0. The speed of this passage is of order of ε. We provide asymptotic formulas that describe effects of passage through a resonance with an improved accuracy O(ε3/2). A numerical verification is also provided. The problem under consideration is a model problem that describes passage through an isolated resonance in multi-frequency quasi-linear Hamiltonian systems. We also discuss a resonant phenomenon of scattering on resonances associated with discretisation arising in a numerical solving of systems with one rotating phase. Numerical integration of ODEs by standard numerical methods reduces continuous time problems to discrete time problems. For arbitrarily small time step of a numerical method, discrete time problems have intrinsic properties that are absent in continuous time problems. As a result, numerical solution of an ODE may demonstrate dynamical phenomena that are absent in the original ODE. We show that numerical integration of systems with one fast rotating phase leads to a situation of such kind: numerical solution demonstrates phenomenon of scattering on resonances, that is absent in the original system.

A pair of nonidentical two-level atoms, separated by a fixed distance R, interact through photon exchange. The system is described by a state vector which is assumed to be a superposition of four "essential states": (1) the first atom is excited, the second one is in the ground state, and no photon is present, (2) the first atom is in its ground state, the second one is excited, and no photon is present, (3) both atoms are in their ground states and a photon is present, and (4) both atoms are excited and a photon is also present. The system is initially in state (1). The probabilities of each atom being excited are calculated for both the minimally-coupled interaction and the multipolar interaction in the electric dipole approximation. For the minimally-coupled interaction Hamiltonian, the second atom has a probability of being instantaneously excited, so the interaction is not retarded. For the multipolar interaction Hamiltonian, the second atom is not excited before the retardation time, which agrees with special relativity. For the minimally-coupled interaction the nonphysical result occurs because the unperturbed Hamiltonian is not the energy operator in the Coulomb gauge. For the multipolar Hamiltonian in the electric dipole approximation the unperturbed Hamiltonian is the energy operator. An active view of unitary transformations in nonrelativistic quantum electrodynamics is used to derive transformation laws for the potentials of the electromagnetic field and the static Coulomb potential. For a specific choice of unitary transformation the transformation laws for the potentials are used in the minimally-coupled second-quantized Hamiltonian to obtain the multipolar Hamiltonian, which is expressed in terms of the quantized electric and magnetic fields.

Dans cette thèse, nous discutons l'influence d'un réseau qui possède une topologie non triviale sur les propriétés collectives d'un modèle hamiltonien pour spins,le modèle $XY$, défini sur ces réseaux.Nous nous concentrons d'abord sur la topologie des chaînes régulières et du réseau Petit Monde (Small World), créé avec le modèle Watt- Strogatz.Nous contrôlons ces réseaux par deux paramètres $\gamma$, pour le nombre d' interactions et $p$, la probabilité de ré-attacher un lien aléatoirement.On définit deux mesures, le chemin moyen $\ell$ et la connectivité $C$ et nous analysons leur dépendance de $(\gamma,p)$.Ensuite,nous considérons le comportement du modèle $XY$ sur la chaîne régulière et nous trouvons deux régimes: un pour $\gamma<1,5$,qui ne présente pas d'ordre longue portée et un pour $\gamma>1,5$ où une transition de phase du second ordre apparaît.Nous observons l'existence d'un état ​​métastable pour $\gamma_ {c} = 1,5$. Sur les réseaux Petit Monde,nous illustrons les conditions pour avoir une transition et comment son énergie critique $\varepsilon_{c}(\gamma,p)$ dépend des paramètres $(\gammap$).Enfin,nous proposons un modèle de réseau où les liens d'une chaîne régulière sont ré-attachés aléatoirement avec une probabilité $p$ dans un rayon spécifique $r$. Nous identifions la dimension du réseau $d(p,r)$ comme un paramètre crucial:en le variant,il nous est possible de passer de réseaux avec $d<2$ qui ne présentent pas de transition de phase à des configurations avec $d>2$ présentant une transition de phase du second ordre, en passant par des régimes de dimension $d=2$ qui présentent des états caractérisés par une susceptibilité infinie et une dynamique chaotique. / In this thesis we discuss the influence of a non trivial network topology on the collective properties of an Hamiltonian model defined on it, the $XY$ -rotors model. We first focus on networks topology analysis, considering the regular chain and a Small World network, created with the Watt-Strogatz model. We parametrize these topologies via $\gamma$, giving the vertex degree and $p$, the probability of rewiring. We then define two topological parameters, the average path length $\ell$and the connectivity $C$ and we analize their dependence on $\gamma$ and $p$. Secondly, we consider the behavior of the $XY$- model on the regular chain and we find two regimes: one for $\gamma<1.5$, which does not display any long-range order and one for $\gamma>1.5$ in which a second order phase transition of the magnetization arises. Moreover we observe the existence of a metastable state appearing for $\gamma_{c}=1.5$. Finally we illustrate in what conditions we retrieve the phase transition on Small World networks and how its critical energy $\varepsilon_{c}(\gamma,p)$ depends on the topological parameters $\gamma$ and $p$. In the last part, we propose a network model in which links of a regular chain are rewired according to a probability $p$ within a specific range $r$. We identify a quantity, the network dimension $d(p,r)$ as a crucial parameter. Varying this dimension we are able to cross over from topologies with $d<2$ exhibiting no phase transitions to ones with $d>2$ displaying a second order phase transition, passing by topologies with dimension $d=2$ which exhibit states characterized by infinite susceptibility and macroscopic chaotic dynamical behavior.

A presente tese analisa diferentes aspectos de sistemas de espalhamento clássico com caos. Espalhamento caótico é uma forma de caos transiente que ocorre em diversos sistemas físicos. Nestes sistemas o espaço de fase é aberto, mas o caos ocorre apenas em uma região restrita do espaço, chamada de região de espalhamento. Os efeitos desta dinâmica apresentam-se em qualquer relação de espalhamento pela presença de conjuntos fractais, que geram hiper-sensibilidade a condições iniciais. Em nosso primeiro trabalho, mostramos que as bifurcações que levam ao caos manifestam-se na Seção de Choque Diferencial (SCD) pela criação de infinitas singularidades arco-íris. Estas singularidades aparecem na forma de cascatas, registrando na SCD todas as transições sofridas pela sela caótica. O segundo trabalho mostra que a introdução de dissipação em sistemas de espalhamento pode limitar a autosimilaridade de conjuntos originalmente fractais. Uma partícula espalhada por potenciais repulsivos encontra regiões não acessíveis, que dependem do valor de sua energia. Estas regiões determinam a estrutura da sela caótica. Com a perda de energia, o cenário de órbitas presas é alterado e, dependendo do valor da dissipação, podem existir nas funções de espalhamento estruturas fractais truncadas. O terceiro estudo aborda a presença de advecção caótica em fluxos sanguíneos. Doenças circulatórias estão geralmente associadas a uma mudança de geometria de artérias ou veias. Essas deformações podem gerar espalhamento caótico das partículas sanguíneas carregadas pelo fluxo. Em nosso trabalho mostramos, a partir de simulações numéricas, que caos pode existir em fluxos sanguíneos e, assim, formar um ciclo no desenvolvimento de anomalias circulatórias. / In this thesis we study different scattering systems with chaos. Chaotic scattering, present in a large variety of physical systems, is a type of transient chaos. While the phase-space of such systems is unbounded, irregular motion occurs only in a bounded area, called the scattering region. Still, any (nontrivial) scattering function relating initial conditions to asymptotic variables contains fractal structures, resulting in a very sharp sensitivity to initial conditions. Our first work shows that bifurcations leading to chaos manifest themselves through an infinitely fine-scale structure of rainbow singularities in the cross section. These singularities appear as cascades, mirroring the bifurcation cascade undergone by the chaotic saddle. The second work shows that the presence of dissipation in scattering systems can limit the auto-similarity of originally fractal structures. Depending on the value of their energy, particles scattered by repulsive potentials find forbidden regions in the space-phase. These regions determinate the structure of the chaotic saddle. With friction, the scenario of trapped orbits changes and, depending on the ammount dissipation, scattering functions follow a truncated fractal structure. Our third study concerns the presence of chaotic advection in blood flows. Typically, circulatory diseases are due to sudden changes on the geometry of vessel walls. These deformations can generate chaotic scattering of blood particles carried by the flow. We show, with numerical simulations, that chaos can occur in blood flows and thus form a hazardous cycle in the further developing of circulatory anomalies.