Gyvosios sistemos tikriausia labiausiai žadina žmogaus smalsumą ir suteikia įkvėpimo bendrųjų dėsnių ieškojimams. Daugelio jų sudėtinga tvarka ir dinamika vis dar yra neįmintos mįslės ir neišspręsti uždaviniai. Vienas tokių uždavinių yra sinchronizacijos atsiradimas, įtaka ir valdymas neuronų populiacijose. Nors neuronas – sudėtinga biologinė sistema ir matematiniai jo modeliai yra pakankamai komplikuoti, sinchronizacija silpnai įtakoja atskiro neorono ypatybes, bet atspindi visos populiacijos dinamiką. Pasinaudodami šia palankia aplinkybe, galime aproksimuoti neuronų populiaciją labai paprastų globaliai susietųjų (sąveikauja kiekvienas elementas su kiekvienu) osciliatorių ansambliu ir gauti gerą matematinio populiacijos dinamikos modelio sutapimą su realia sistema. Darni neuronų veikla žmogaus organizmui yra gyvybiškai svarbus procesas, kurio sutrikimai dažniausiai turi stiprias neigiamas pasekmes. Empiriškai buvo pastebėta, kad atsiradus Parkinsono ligos simptomams, dalis neuronų sinchronizuojasi. Kai ši sinchronizacija sustabdoma, ligos simptomai žymiai susilpnėja arba visai išnyksta. Medicinoje jau naudojamas aukšto dažnio giluminės smegenų stimuliacijos metodas gydyti šiai ligai yra veiksmingas, bet šis mechanizmas nėra gerai suprantamas ir turi nemažai neigiamų savybių: · metodas yra invazinis, · nėra grįžtamojo ryšio, · žmogaus smegenys yra adaptyvios ir ilgainiui pripranta prie pastovios stimuliacijos, o tai iššaukia stimuliacijos didinimą, · galimos komplikacijos... [toliau žr. visą tekstą] / The phenomenon of the synchronization was observed and studied since XVII century and until today has been the main subject of many researches and core trigger of many devices as well as nature appearences. Though sometimes synchronization is not a desirable process and it is important to learn how to command over it in order to suppress or to strengthen synchronous behaviour in accordance with the results one would like to obtain. This study focuses on controlling the process of synchronization in globally coupled ensemble of oscillators with a configuration of separated observed and stimulated subsystems. The development of such technique could be usefull for suppression of the undesired synchronization of neural networks in the cases like Parkinsonian desease and dystonia. The main advantage of this method is being noninvasive feedback control.

Orientadores: Alberto Vazquez Saa, Marcus Aloizio Martinez de Aguiar / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Física Gleb Wataghin / Made available in DSpace on 2018-08-28T03:40:49Z (GMT). No. of bitstreams: 1 Pinto_RafaelSoares_D.pdf: 7979471 bytes, checksum: b344e1e01031709b8b938dbecb572900 (MD5) Previous issue date: 2015 / Resumo: Sincronização está presente em uma miríade de situações, indo desde vaga-lumes piscando em uníssono na copa das árvores, populações de leveduras ajustando seu metabolismo para um ritmo comum, atividades neurais ocorrendo no cérebro, chegando até as redes de distribuição de energia elétrica, as maiores máquinas construídas pelo homem. Neste trabalho, nós analisamos como se dá o processo de sincronização utilizando o bem conhecido modelo de Kuramoto, estudado incansavelmente nas últimas décadas, quando ele se encontra sobre uma rede complexa, que determina os padrões de interação entre os elementos que compõem a população. A topologia dessas interações determina de maneira crucial a dinâmica do sistema, possibilitando, ou não, a sincronização dos seus elementos. Primeiros, nós analisamos o fenômeno da sincronização explosiva: a correlação de propriedades da rede com a frequência natural dos osciladores altera dramaticamente a natureza da transição de fase do estado não sincronizado para o estado sincronizado. Mostramos que sincronização explosiva ocorre mesmo quando apenas uma pequena fração dos vértices da rede possuem tal correlação, a saber, os vértices mais bem conectados da rede. Além do mais, ajustando o número de vértices onde a correlação é válida, podemos controlar propriedades dessa transição de fase. A seguir estudamos o processo de optimização de topologia para favorecer sincronização. Dado um conjunto de vértices/osciladores com frequências naturais conhecidas e um certo número de links, qual é a melhor topologia, ou seja, o padrão de conexões, que favorece a sincronização? Estudamos esse problema numericamente para o modelo de Kuramoto com inércia, que serve como um modelo simples para analisar as redes de transmissão de energia elétrica, obtendo princípios básicos que devem ser utilizados para o design de tais sistemas. Por fim, ainda no problema de optimização de topologia para favorecer sincronização, obtivemos pela primeira vez de forma analítica as condições para optimização para o modelo de Kuramoto, bem como para uma generalização sua, onde há interações positivas e negativas. Esses resultados analíticos ainda servem para criar algoritmos de optimização mais ecientes que os utilizados atualmente / Abstract: Synchronization is present in a myriad of situations, from the unison ashing of reies in trees, populations of yeast adjusting their metabolism to a common rhythm, neural activities in the brain to the largest machines ever built, the power grids. We analysed how the process of synchronization happens using the well known Kuramoto model, tirelessly studied in the last decades, when it is on top of a complex network, that determines the patterns of interaction between the elements of the population. The topology of this network's determines crucially the possible dynamics of the systems, allowing, or not, the synchronization of its elements. We rst discuss the phenomenon of explosive synchronization, where the correlation between properties of the network and the oscillators changes drastically the nature of the phase transition separating the incoherent state from the synchronized state.We show that explosive synchronization can occur even when a small subset of the vertices are correlated. It is necessary that only the hubs, vertices with highest degrees, show the correlation. Moreover, adjust the fraction of correlated vertices allows us to control properties of the phase transition. Next we study the optimization of the topology to favor synchronization. Given a set of vertices/oscillators with know natural frequencies and a certain number of links, which is the best topology, its pattern of interactions, to favor synchronization? We studied this problem to a generalized Kuramoto model (Kuramoto model with inertia) that is used as a simple tool to model power grids, obtaining in this way simple rules that can be applied to the design of such systems that already helps the synchronization of its elements. In our nal contribution, still in the optimization of the topology problem, we were able, for the first time, to obtain analytically the conditions of optimization for the Kuramoto model, as well as for one of its generalizations, where there can exist positive and negative interactions between the elements. Beyond the signicant fact that the conditions can be know analytically, these results can be used to obtain faster optimization algorithms that the current ones / Doutorado / Física / Doutor em Ciências / 2012/09357-9 / CAPES

none

Feedback control is used in almost every aspect of modern life and is essential in almost all engineering systems. Since no mathematical model is perfect and disturbances occur frequently, feedback is required. The design of a feedback control has been widely investigated in finite-dimensional space. However, many systems of interest, such as fluid flow and large structural vibrations are described by nonlinear partial differential equations and their state evolves on an infinite-dimensional Hilbert space. Developing controller design methods for nonlinear infinite-dimensional systems is not trivial. The objectives of this thesis are divided into multiple tasks. First, the well-posedness of some classes of nonlinear partial differential equations defined on a Hilbert space are investigated. The following nonlinear affine system defined on the Hilbert space H is considered z ̇(t)=F(z(t))+Bu(t), t≥0 z (0) = z0, where z(t) ∈ H is the state vector and z0 is the initial condition. The vector u(t) ∈ U, where U is a Hilbert space, is a state-feedback control. The nonlinear operator F : D ⊂ H → H is densely defined in H and the linear operator B : U → H is a linear bounded operator. Conditions for the closed-loop system to have a unique solution in the Hilbert space H are given. Next, finding a single bounded state-feedback control for nonlinear partial differential equations is discussed. In particular, Lyapunov-indirect method is considered to control nonlinear infinite-dimensional systems and conditions on when this method achieves the goal of local asymptotic stabilization of the nonlinear infinite-dimensional system are given. The Kuramoto-Sivashinsky (KS) equation defined in the Hilbert space L2(−π,π) with periodic boundary conditions is considered. ∂z/∂t =−ν∂4z/∂x4 −∂2z/∂x2 −z∂z/∂x, t≥0 z (0) = z0 (x) , where the instability parameter ν > 0. The KS equation is a nonlinear partial differential equation that is first-order in time and fourth-order in space. It models reaction-diffusion systems and is related to various pattern formation phenomena where turbulence or chaos appear. For instance, it models long wave motions of a liquid film over a vertical plane. When the instability parameter ν < 1, this equation becomes unstable. This is shown by analyzing the stability of the linearized system and showing that the nonlinear C0- semigroup corresponding to the nonlinear KS equation is Fr ́echet differentiable. There are a number of papers establishing the stabilization of this equation via boundary control. In this thesis, we consider distributed control with a single bounded feedback control for the KS equation with periodic boundary conditions. First, it is shown that sta- bilizing the linearized KS equation implies local asymptotical stability of the nonlinear KS equation. This is done by establishing Fr ́echet differentiability of the associated nonlinear C0-semigroup and showing that it is equal to the linear C0-semigroup generated by the linearization of the equation. Next, a single state-feedback control that locally asymptot- ically stabilizes the KS equation is constructed. The same approach to stabilize the KS equation from one equilibrium point to another is used. Finally, the solution of the uncontrolled/state-feedback controlled KS equation is ap- proximated numerically. This is done using the Galerkin projection method to approximate infinite-dimensional systems. The numerical simulations indicate that the proposed Lyapunov-indirect method works in stabilizing the KS equation to a desired state. Moreover, the same approach can be used to stabilize the KS equation from one constant equilibrium state to another.

Feedback control is used in almost every aspect of modern life and is essential in almost all engineering systems. Since no mathematical model is perfect and disturbances occur frequently, feedback is required. The design of a feedback control has been widely investigated in finite-dimensional space. However, many systems of interest, such as fluid flow and large structural vibrations are described by nonlinear partial differential equations and their state evolves on an infinite-dimensional Hilbert space. Developing controller design methods for nonlinear infinite-dimensional systems is not trivial. The objectives of this thesis are divided into multiple tasks. First, the well-posedness of some classes of nonlinear partial differential equations defined on a Hilbert space are investigated. The following nonlinear affine system defined on the Hilbert space H is considered z ̇(t)=F(z(t))+Bu(t), t≥0 z (0) = z0, where z(t) ∈ H is the state vector and z0 is the initial condition. The vector u(t) ∈ U, where U is a Hilbert space, is a state-feedback control. The nonlinear operator F : D ⊂ H → H is densely defined in H and the linear operator B : U → H is a linear bounded operator. Conditions for the closed-loop system to have a unique solution in the Hilbert space H are given. Next, finding a single bounded state-feedback control for nonlinear partial differential equations is discussed. In particular, Lyapunov-indirect method is considered to control nonlinear infinite-dimensional systems and conditions on when this method achieves the goal of local asymptotic stabilization of the nonlinear infinite-dimensional system are given. The Kuramoto-Sivashinsky (KS) equation defined in the Hilbert space L2(−π,π) with periodic boundary conditions is considered. ∂z/∂t =−ν∂4z/∂x4 −∂2z/∂x2 −z∂z/∂x, t≥0 z (0) = z0 (x) , where the instability parameter ν > 0. The KS equation is a nonlinear partial differential equation that is first-order in time and fourth-order in space. It models reaction-diffusion systems and is related to various pattern formation phenomena where turbulence or chaos appear. For instance, it models long wave motions of a liquid film over a vertical plane. When the instability parameter ν < 1, this equation becomes unstable. This is shown by analyzing the stability of the linearized system and showing that the nonlinear C0- semigroup corresponding to the nonlinear KS equation is Fr ́echet differentiable. There are a number of papers establishing the stabilization of this equation via boundary control. In this thesis, we consider distributed control with a single bounded feedback control for the KS equation with periodic boundary conditions. First, it is shown that sta- bilizing the linearized KS equation implies local asymptotical stability of the nonlinear KS equation. This is done by establishing Fr ́echet differentiability of the associated nonlinear C0-semigroup and showing that it is equal to the linear C0-semigroup generated by the linearization of the equation. Next, a single state-feedback control that locally asymptot- ically stabilizes the KS equation is constructed. The same approach to stabilize the KS equation from one equilibrium point to another is used. Finally, the solution of the uncontrolled/state-feedback controlled KS equation is ap- proximated numerically. This is done using the Galerkin projection method to approximate infinite-dimensional systems. The numerical simulations indicate that the proposed Lyapunov-indirect method works in stabilizing the KS equation to a desired state. Moreover, the same approach can be used to stabilize the KS equation from one constant equilibrium state to another.

<p>We are interested in backward-in-time solution techniques for evolutionary PDE problems arising in fluid mechanics. In addition to their intrinsic interest, such techniques have applications in recently proposed retrograde data assimilation. As our model system we consider the terminal value problem for the Kuramoto-Sivashinsky equation in a l D periodic domain. The Kuramoto-Sivashinsky equation, proposed as a model for interfacial and combustion phenomena, is often also adopted as a toy model for hydrodynamic turbulence because of its multiscale and chaotic dynamics. Such backward problems are typical examples of ill-posed problems, where any disturbances are amplified exponentially during the backward march. Hence, regularization is required to solve such problems efficiently in practice. We consider regularization approaches in which the original ill-posed problem is approximated with a less ill-posed problem, which is achieved by adding a regularization term to the original equation. While such techniques are relatively well-understood for linear problems, it is still unclear what effect these techniques may have in the nonlinear setting. In addition to considering regularization terms with fixed magnitudes, we also explore a novel approach in which these magnitudes are adapted dynamically using simple concepts from the Control Theory.</p> / Thesis / Master of Science (MSc)

Synchronization of an ensemble of oscillators is an emergent phenomenon present in several complex systems, ranging from biological and physical to social and technological systems. The most successful approach to describe how coherent behavior emerges in these complex systems is given by the paradigmatic Kuramoto model. For decades, this model has been traditionally studied in globally coupled topologies. However, besides being intrinsically dynamical, complex systems exhibit very heterogeneous structure, which can be represented as complex networks. This thesis is dedicated to the investigation of fundamental problems regarding the collective dynamics of Kuramoto oscillators coupled in complex networks. First, we address the effects on network dynamics caused by the presence of triangles, which are structural patterns that permeate real-world networks but are absent in random models. By extending the heterogeneous degree mean-field approach to a class of configuration model that generates random networks with variable clustering, we show that triangles weakly affect the onset of synchronization. Our results suggest that, at least in the low clustering regime, the dynamics of clustered networks are accurately described by tree-based theories. Secondly, we analyze the influence of inertia in the phases evolutions. More precisely, we substantially extend the mean-field calculations to second-order Kuramoto oscillators in uncorrelated networks. Thereby hysteretic transitions of the order parameter are predicted with good agreement with simulations. Effects of degree-degree correlations are also numerically scrutinized. In particular, we find an interesting dynamical equivalence between variations in assortativity and damping coefficients. Potential implications to real-world applications are discussed. Finally, we tackle the problem of two intertwined populations of stochastic oscillators subjected to asymmetric attractive and repulsive couplings. By employing the Gaussian approximation technique we derive a reduced set of ODEs whereby a thorough bifurcation analysis is performed revealing a rich phase diagram. Precisely, besides incoherence and partial synchronization, peculiar states are uncovered in which two clusters of oscillators emerge. If the phase lag between these clusters lies between zero and &pi;, a spontaneous drift different from the natural rhythm of oscillation emerges. Similar dynamical patterns are found in chaotic oscillators under analogous couplings schemes. / Sincronização de conjuntos de osciladores é um fenômeno emergente que permeia sistemas complexos de diversas naturezas, como por exemplo, sistemas biológicos, físicos, naturais e tecnológicos. A abordagem mais bem sucedida na descrição da emergência de comportamento coletivo em sistemas complexos é fornecida pelo modelo de Kuramoto. Durante décadas, este modelo foi tradicionalmente estudado em topologias completamente conectadas. Entretanto, além de ser intrinsecamente dinâmicos, tais sistemas complexos possuem uma estrutura altamente heterogênea que pode ser apropriadamente representada por redes complexas. Esta tese é dedicada à investigação de problemas fundamentais da dinâmica coletiva de osciladores de Kuramoto acoplados em redes. Primeiramente, abordamos os efeitos sobre a dinâmica das redes causados pela presença de triângulos padrões que estão omnipresentes em redes reais mas estão ausentes em redes gerados por modelos aleatórios. Estendemos a abordagem via campo-médio para uma variação do modelo de configuração tradicional capaz de criar topologias com número variável de triângulos. Através desta abordagem, mostramos que tais padrões estruturais pouco influenciam a emergência de comportamento coletivo em redes, podendo a dinâmica destas ser descrita em termos de teorias desenvolvidas para redes com topologia local semelhante a grafos de tipo árvore. Em seguida, analisamos a influência de inércia na evolução das fases. Mais precisamente, generalizamos cálculos de campo-médio para osciladores de segunda-ordem acoplados em redes sem correlação de grau. Demonstramos que na presença de efeitos inerciais o parâmetro de ordem do sistema se comporta de forma histerética. Ademais, efeitos oriundos de correlações de grau são examinados. Em particular, verificamos uma interessante equivalência dinâmica entre variações nos coeficientes de assortatividade e amortecimento dos osciladores. Possíveis aplicações para situações reais são discutidas. Finalmente, abordamos o problema de duas populações de osciladores estocásticos sob a influência de acoplamentos atrativos e repulsivos. Através da aplicação da aproximação Gaussiana, derivamos um conjunto reduzido de EDOs através do qual as bifurcações do sistema foram analisadas. Além dos estados asíncrono e síncrono, verificamos a existência de padrões peculiares na dinâmica de tal sistema. Mais precisamente, observamos a formação de estados caracterizados pelo surgimento de dois aglomerados de osciladores. Caso a defasagem entre estes grupos é inferior a &pi;, um novo ritmo de oscilação diferente da frequência natural dos vértices emerge. Comportamentos dinâmicos similares são observados em osciladores caóticos sujeitos a acoplamentos análogos.

Dans cette thèse, on étudie des systèmes de particules en interaction dont le comportement est lié à certaines équations aux dérivées partielles lorsque le nombre de particules tend vers l’infini. L’équation de Kuramoto-Sivashinsky modélise par exemple la propagation de certains fronts de flamme, la topographie de la surface d’une couche mince en cours de croissance, et fait apparaître des structures macroscopiques. Un modèle de particules en interaction par un couplage harmonique des vitesses, attractif aux premières vitesses voisines, répulsive aux secondes voisines, associée à des collisions élastiques, produit des profils de vitesses analogues aux fronts de flamme. On observe également la création et l’annihilation d’agrégats de particules. Un autre modèle, où les particules fusionnent lors des collisions en préservant masse et quantité de mouvement, et avec uniquement attraction au plus proche voisin, permet de retrouver un modèle de type gaz sans pression avec viscosité. Ces modèles sont étudiés théoriquement, en particulier les facteurs de mise à l’échelle des forces d’interaction sont précisés pour obtenir les équations correctes dans la limite du grand nombre de particules. Des simulations numériques confirment la validité et la pertinence des modèles. / This work is concerned by systems of interacting particles, which are linked to partial derivative equations when the particle number becomes large enough. The Kuramoto-Sivashinsky equation is actually modeling as well the front flame propagation as the morphology of growing interfaces, in deposition, for example. Moreover, surface periodical macroscopic structuring is occurring. An interacting particle model through an harmonic velocity coupling, attractive with the first velocity-neighbor and repulsive for the second neighbors, associated with elestic collisions. This model thus provides us with velocity profiles close to those of front flame propagation. Creation and annihilation of particle clusters is also observed. Another model, where particle are merging during collisions, while retaining mass and momentum conservation and with only nearest neighbor attraction, allows to recover a viscous pressureless gas model. These models are studied using mathematical tools. Especially interaction scaling factors are determined for obtaining the suitable equations in the large particle number limit. The numerical simulations confirm the relevance of the models.

Synchronization of an ensemble of oscillators is an emergent phenomenon present in several complex systems, ranging from biological and physical to social and technological systems. The most successful approach to describe how coherent behavior emerges in these complex systems is given by the paradigmatic Kuramoto model. For decades, this model has been traditionally studied in globally coupled topologies. However, besides being intrinsically dynamical, complex systems exhibit very heterogeneous structure, which can be represented as complex networks. This thesis is dedicated to the investigation of fundamental problems regarding the collective dynamics of Kuramoto oscillators coupled in complex networks. First, we address the effects on network dynamics caused by the presence of triangles, which are structural patterns that permeate real-world networks but are absent in random models. By extending the heterogeneous degree mean-field approach to a class of configuration model that generates random networks with variable clustering, we show that triangles weakly affect the onset of synchronization. Our results suggest that, at least in the low clustering regime, the dynamics of clustered networks are accurately described by tree-based theories. Secondly, we analyze the influence of inertia in the phases evolutions. More precisely, we substantially extend the mean-field calculations to second-order Kuramoto oscillators in uncorrelated networks. Thereby hysteretic transitions of the order parameter are predicted with good agreement with simulations. Effects of degree-degree correlations are also numerically scrutinized. In particular, we find an interesting dynamical equivalence between variations in assortativity and damping coefficients. Potential implications to real-world applications are discussed. Finally, we tackle the problem of two intertwined populations of stochastic oscillators subjected to asymmetric attractive and repulsive couplings. By employing the Gaussian approximation technique we derive a reduced set of ODEs whereby a thorough bifurcation analysis is performed revealing a rich phase diagram. Precisely, besides incoherence and partial synchronization, peculiar states are uncovered in which two clusters of oscillators emerge. If the phase lag between these clusters lies between zero and &pi;, a spontaneous drift different from the natural rhythm of oscillation emerges. Similar dynamical patterns are found in chaotic oscillators under analogous couplings schemes. / Sincronização de conjuntos de osciladores é um fenômeno emergente que permeia sistemas complexos de diversas naturezas, como por exemplo, sistemas biológicos, físicos, naturais e tecnológicos. A abordagem mais bem sucedida na descrição da emergência de comportamento coletivo em sistemas complexos é fornecida pelo modelo de Kuramoto. Durante décadas, este modelo foi tradicionalmente estudado em topologias completamente conectadas. Entretanto, além de ser intrinsecamente dinâmicos, tais sistemas complexos possuem uma estrutura altamente heterogênea que pode ser apropriadamente representada por redes complexas. Esta tese é dedicada à investigação de problemas fundamentais da dinâmica coletiva de osciladores de Kuramoto acoplados em redes. Primeiramente, abordamos os efeitos sobre a dinâmica das redes causados pela presença de triângulos padrões que estão omnipresentes em redes reais mas estão ausentes em redes gerados por modelos aleatórios. Estendemos a abordagem via campo-médio para uma variação do modelo de configuração tradicional capaz de criar topologias com número variável de triângulos. Através desta abordagem, mostramos que tais padrões estruturais pouco influenciam a emergência de comportamento coletivo em redes, podendo a dinâmica destas ser descrita em termos de teorias desenvolvidas para redes com topologia local semelhante a grafos de tipo árvore. Em seguida, analisamos a influência de inércia na evolução das fases. Mais precisamente, generalizamos cálculos de campo-médio para osciladores de segunda-ordem acoplados em redes sem correlação de grau. Demonstramos que na presença de efeitos inerciais o parâmetro de ordem do sistema se comporta de forma histerética. Ademais, efeitos oriundos de correlações de grau são examinados. Em particular, verificamos uma interessante equivalência dinâmica entre variações nos coeficientes de assortatividade e amortecimento dos osciladores. Possíveis aplicações para situações reais são discutidas. Finalmente, abordamos o problema de duas populações de osciladores estocásticos sob a influência de acoplamentos atrativos e repulsivos. Através da aplicação da aproximação Gaussiana, derivamos um conjunto reduzido de EDOs através do qual as bifurcações do sistema foram analisadas. Além dos estados asíncrono e síncrono, verificamos a existência de padrões peculiares na dinâmica de tal sistema. Mais precisamente, observamos a formação de estados caracterizados pelo surgimento de dois aglomerados de osciladores. Caso a defasagem entre estes grupos é inferior a &pi;, um novo ritmo de oscilação diferente da frequência natural dos vértices emerge. Comportamentos dinâmicos similares são observados em osciladores caóticos sujeitos a acoplamentos análogos.

The problem of constructing data-based, predictive, reduced models for the Kuramoto–Sivashinsky equation is considered, under circumstances where one has observation data only for a small subset of the dynamical variables. Accurate prediction is achieved by developing a discrete-time stochastic reduced system, based on a NARMAX (Nonlinear Autoregressive Moving Average with eXogenous input) representation. The practical issue, with the NARMAX representation as with any other, is to identify an efficient structure, i.e., one with a small number of terms and coefficients. This is accomplished here by estimating coefficients for an approximate inertial form. The broader significance of the results is discussed.