ARTICLE IN PRESS. Dynamic graphs. Electrical Engineering Department, Santa Clara University, Santa Clara, CA 95053, USA

Size: px
Start display at page:

Download "ARTICLE IN PRESS. Dynamic graphs. Electrical Engineering Department, Santa Clara University, Santa Clara, CA 95053, USA"

Transcription

1 Nonlinear Analysis: Hybrid Systems ( ) Dynamic graphs D.D. Šiljak Electrical Engineering Department, Santa Clara University, Santa Clara, CA 95053, USA Received 26 July 2006; accepted 4 August 2006 Abstract Dynamic graphs are defined in a linear space as a one-parameter group of transformations of the graph space into itself. Stability of equilibrium graphs is formulated in the sense of Lyapunov to study motions of positive graphs in the nonnegative orthant of the graph space. Relying on the isomorphism of graphs and adjacency matrices, a new concept of dynamic connective stability of complex systems is introduced. A dynamic interaction coordinator is added to complex interconnected system to ensure that the desired level of interconnections between subsystems is preserved as a connectively stable equilibrium of the overall system despite uncertain structural perturbations. It is shown how the coordinator can be designed to adaptively adjust the interconnection levels in order to assign a prescribed state of the complex multi-agent system as a stable equilibrium point. The equilibrium assignment is achieved by the action of the coordinator which solves an optimization problem involving a two-time-scale system; the coordinator action is slow compared to the fast dynamics of the agents. Polytopic connective stability of the multi-agent systems with a coordinator is established by the concept of vector Lyapunov functions and the theory of M-matrices. c 2007 Elsevier Ltd. All rights reserved. Keywords: Graphs; Dynamic graphs and adjacency matrices; Positive dynamic graphs; Complex interconnected systems; Interconnection coordinator; Multi-agent systems; Equilibrium assignment problem; Two-time-scale optimization problem; Adaptive coordination 1. Introduction Graph theory provides a mathematical model for studying interconnections among elements in natural and manmade systems [1]. Initially, interactions were limited to binary relations among the individual elements represented by vertices of the graph. Subsequently, functions were associated with graphs that assign a real number to each edge of a graph in order to quantify the relation between any pair of elements in a given system [2]. While graphs have been traditionally studied as static objects, in applications of graph theory to social systems, it has long been recognized that graphs should change in time to better reflect the reality of social interactions in groups and organizations. In 1940, Radcliffe-Brown [3] wrote:... a consideration of the continuity of social structure through time, a continuity which is not static like that of a building, but a dynamic continuity (my italics), like that of organic structure of a living body. A natural way to introduce changes in graphs is to weight vertices and/or edges of a graph by their existence probability [4]. A generic problem in the context of probabilistic graphs has been to find the probability that a graph remains connected despite random drop out of vertices and edges. Tel.: address: DSiljak@scu.edu X/$ - see front matter c 2007 Elsevier Ltd. All rights reserved. doi: /j.nahs

2 2 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) Dynamic graphs have been introduced in stability studies of complex dynamic systems to study the effect of uncertain interconnections between the subsystems on the stability of the overall system [5,6]. The notion of connective stability was introduced within the context of Lyapunov stability to determine conditions under which a complex system remains stable despite structural perturbations whereby subsystems are disconnected and reconnected during the evolution of the systems in the state space. Dynamic changes of graphs have been subsequently described by integer-valued functions of integral arguments with operations defined by means of integer-valued recurrence relations [7]. Stochastic models of dynamic graphs have been introduced in the stability analysis of multi-controller schemes for reliability enhancement, where graphs representing failure and maintenance of controllers have appeared as states of a Markov process [8]. In control design, when input and output vertices are added to the system graph, the minimal requirement is that graph be input-reachable, that is, we want that each state variable of the system be reached by a path from at least one input. Either by accident or design, lines of communication and control may be disconnected, which raises the question of vulnerability [6]: Does a removal of an edge or vertex from the system graph destroy input-reachability? Graph-theoretic procedures have been developed [9] to identify the minimal number of edges which are essential for preserving input-reachability and structural controllability. Numerous methods and techniques have been developed to study static properties of complex control systems that go beyond reachability, including structural controllability, structurally fixed modes, and almost invariant subspaces, as well as structural decompositions of hierarchical and weak coupling variety [10]. Recently, there has been a growing interest in using system graphs in active roles to manipulate interconnections in a network of agents to achieve desired properties and performance of the overall multi-agent system. A dynamic adjacency matrix was added to a multi-agent system in [11] to set the interconnection levels among the agents that result in a preassigned equilibrium. Time- and state-dependent changes of adjacency matrices were used by Feddema et al. [12] to determine if and how communication failures affect input-reachability and stability of multiple cooperative robotic vehicles. Formation control graphs have been introduced by Tanner et al. [13] to model the control-related interconnections among the agents and achieve input-to-state stability in a string of leader-following agents. Fax and Murray [14] made the use of formation graphs to obtain information exchange strategies which are robust to changes in the communication structure. Assuming that state agreement among agents is reached if states of the agents converge to a common location in the state space, Lin et al. [15,16] have shown that such a state can be achieved by a switching control if the interconnection graph is sufficiently connected over time. The main objective of this paper is to initiate a study of graphs in a linear vector space, and describe a dynamic graph as a one-parameter group of transformations of the graph space into itself. By introducing a metric into the graph space, we define stability of equilibrium graphs in the sense of Lyapunov. Since numerous applications of graphs to natural and man-made systems require graphs to have positive weights, we define positive dynamic graphs that live in the nonnegative orthant of the graph space. In particular, we select the standard Lotka Volterra model to illustrate stability properties of positive graphs whose edges act as interacting species in a population model. We also show how a preassigned positive equilibrium can be inserted into a given model and then stabilized by feedback. When a nonlinear complex system is given as an interconnection of a number of subsystems (e.g., agents) having local inputs, we introduce the interconnection coordinator which is an add-on device that modifies the interconnections to maintain prespecified levels of coupling among the subsystems despite uncertain perturbations in the interconnection structure. The coordinator is modeled as a dynamic graph having the equilibrium that corresponds to the desired coupling strengths. By placing a region of attraction of the equilibrium graph inside the connective stability region, the proper functioning of the coordinator is ensured by the connective stability of the overall complex (multi-agent) system. Finally, by minimizing a suitably chosen functional in the graph space, we derive an adaptive coordination law to solve the equilibrium assignment problem in complex (multi-agent) systems. The coordinator applies the law to drive the coupling strengths among the subsystems (agents) until a prescribed constant state vector becomes the desired stable equilibrium of the overall complex system. In order to prevent interference of the coordinator dynamics with the dynamics of the subsystems (agents) during the on-line adaptation process, the complex system is set as a twotime-scale system; the coordinator, which contains a small parameter, moves by an order slower than the individual agents. Polytopic connective stability of the two-time-scale system is established using the concept of vector Lyapunov functions and the theory of M-matrices.

3 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 3 2. Dynamic graphs We consider a weighted directed graph (weighted digraph, or simply, graph) D = (V, E) which is an ordered pair, where V is a nonempty and finite set of N vertices (points) and E is a set of edges (lines). The vertices {v 1, v 2,..., v N } are connected by edges (v j, v i ), each edge being oriented from v j to v i, i, j N = {1, 2,..., N}. To each edge (v j, v i ) we assign a weight e i j if edge (v j, v i ) D, while e i j = 0 if (v j, v i ) D. With digraph D we associate the isomorphic concept of N N adjacency (interconnection) matrix E = (e i j ). In the following, we will take advantage of this isomorphism and use D and E interchangeably to fit the context. Our first objective is to define a space D of graphs with a fixed number N of vertices, as a linear space over the field F of real numbers. For any D 1, D 2 D, there is a unique graph D 1 + D 2 D called the sum of D 1 and D 2, and for any D D and any α F there is a unique graph αd D which is the multiplication of the graph D by a scalar α. When we choose in (2) α = 0, then αd = 0 is the zero graph, D = 0 D, which consists of N disconnected vertices V, that is, its set of edges is empty (E = ). The two operations that define D as a linear space, can be interpreted in the context of the linear space E of adjacency matrices. For any two N N matrices E 1 = (ei 1 j ), E 2 = (ei 2 j ) the sum is (e 1 i j ) + (e2 i j ) = (e1 i j + e2 i j ) E, and for any N N matrix E = (e i j ) E and a scalar α F, we recall the multiplication of the matrix E by a scalar α, α(e i j ) = (αe i j ) E. Finally, we note that the zero element of space E is the N N zero matrix E = 0 E. Since we are interested in motions of graphs and their stability, we need to add a norm to space D. We proceed formally and define a graph norm ν(d) by the following properties: (i) ν(d) > 0, D D (D 0), (ii) ν(αd) = α ν(d), D D, α F (iii) ν(d 1 + D 2 ) ν(d 1 ) + ν(d 2 ), D 1, D 2 D. Again, we can use the space of matrices E, which is isomorphic to the abstract space D, to interpret the properties of the norm ν(d). This is of interest in computations, because we can use the matrix norm ν : R N N R + in R N N defined as (i) ν(e) > 0, E R N N (E 0), (ii) ν(αe) = α ν(e), E R N N, α F (iii) ν(e 1 + E 2 ) ν(e 1 ) + ν(e 2 ), E 1, E 2 R N N. Once we choose a norm, we can define a metric in D, which provides a distance between any two graphs D 1 and D 2 in D, ρ(d 1, D 2 ) = ν(d 1 D 2 ), D 1, D 2 D. (7) For adjacency matrices we consider the matrix space E as R N N with the metric ρ(e 1, E 2 ) = ν(e 1 E 2 ), E 1, E 2 E. (8) We now use the recipe from [17] and propose an axiomatic definition of dynamic graphs as mappings of the abstract space D into itself. In this way, we provide a more general setting for analysis of dynamic graphs than it would be available by using differential equations as in [6,11]. (1) (2) (3) (4) (5) (6)

4 4 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) Let us now consider the space D and a family of mappings Φ(t, D), which to any graph D D and any parameter (time) t R assigns a graph Φ D. Definition 1. A dynamic graph D is a one-parameter mapping Φ : R D D of the space D into itself satisfying the following axioms: (i) Φ(t 0, D 0 ) = D 0, t 0 R, D 0 D (ii) Φ(t, D) is continuous t R, D D (iii) Φ(t 2, Φ(t 1, D)) = Φ(t 1 + t 2, D), t 1, t 2 R, D D. (9) The first axiom establishes D(t 0 ) = D 0 as the initial graph. The second axiom requires continuity of mapping Φ(t, D) with respect to all t and D, which includes continuity with respect to t 0 and D 0. Finally, the third axiom establishes dynamic graph D as a one-parameter group Φ(t, D) of transformations of the space D into itself. In applications of dynamic graphs, adjacency matrices play a prominent role, and we state: Definition 2. A dynamic adjacency matrix E is a one-parameter mapping Ψ : R R N N R N N of space R N N into itself satisfying the following axioms: (i) Ψ(t 0, E 0 ) = E 0, t 0 R, E 0 R N N (ii) Ψ(t, E) is continuous t R, E R N N (iii) Ψ(t 2, Ψ(t 1, E)) = Ψ(t 1 + t 2, E), t 1, t 2 R, E R N N. (10) In dynamic graph analysis, Φ(t, D) is called a motion of graph D, while Ψ(t, E) is a motion of the adjacency matrix E. Of particular interest will be the constant graph motions defined by Φ(t, D e ) = D e, t R. (11) A graph D e satisfying this condition is called the equilibrium graph. Similarly, an equilibrium adjacency matrix E e is defined by Ψ(t, E e ) = E e, t R. (12) In the following section, we consider stability of equilibrium graphs to determine if a given equilibrium graph is stable or unstable. An unstable equilibrium graph does not represent a viable steady-state structure that can emerge from graph motions in its neighborhood. 3. Stability of graphs The shape and character of motions around the equilibrium graphs and adjacency matrices are of interest primarily because they can tell us if a particular structure represented by a graph (or adjacency matrix) is going to persist in time. For this purpose, we introduce the notion of stability of dynamic graphs in the sense of Lyapunov. With a slight abuse of notation, we use D(t, D 0 ) instead of Φ(t, D 0 ) and state the following: Definition 3. An equilibrium graph D e D is: (i) stable if for every ε > 0 and t 0 R there is a δ > 0, such that implies ρ(d 0, D e ) < δ ρ(d(t, D 0 ), D e ) < ε, t t 0 ; (13) (14) (ii) asymptotically stable if it is stable and for every t 0 R there is an η > 0, such that ρ(d 0, D e ) < η (15)

5 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 5 implies lim ρ(d(t, D 0), D e ) = 0; t (iii) globally asymptotically stable if the limit holds for arbitrarily large (fixed) η, that is, if it holds for all D 0 D. Since stability properties of dynamic graphs are conveniently studied using adjacency matrices, we need the following variant of Definition 3: Definition 4. An equilibrium adjacency matrix E e R N N is: (i) stable if for every ε > 0 and t 0 R there is a δ > 0, such that (16) implies ρ(e 0, E e ) < δ ρ(e(t, E 0 ), E e ) < ε, t t 0 ; (17) (18) (ii) asymptotically stable if it is stable and for every t 0 R there is an ζ > 0, such that implies ρ(e 0, E e ) < ζ lim ρ(e(t, E 0), E e ) = 0; t (iii) globally asymptotically stable if the limit holds for arbitrarily large (fixed) ζ, that is, if it holds for all E 0 R N N. Time evolutions of dynamic graphs can be defined (abstractly) by the equation (19) (20) D : D = G(t, D), (21) where D represents a tendency of graph D to change in time, while G(t, D) is the transition function which defines how D depends on graph D(t) at time t. In terms of the corresponding adjacency matrix E, Eq. (21) can be written as E : d E = F(t, E), dt where we choose d/dt for E, but other choices, such as, E(t + 1) E(t), / t, etc., are possible. We assume that the matrix function F : R R N N R N N is sufficiently smooth, so that equation describing matrix E has a unique solution Ψ(t, E 0 ) for all (t 0, E 0 ) R R N N. In order to convert a matrix differential equation describing E into the standard vector form, we vectorize the N N matrix E = (e i j ) by stacking up the rows of E to form a vector differential equation (22) E : ė = g(t, e). (23) To show how the vector e R M and, thus, function g : R R M R M, are obtained from the matrix E, we first convert E to a binary matrix Ē = (ē i j ) R N N by the rule ē i j = 1, e i j 0 = 0, e i j 0. (24) We recall that the matrix Ē is the fundamental interconnection (adjacency) matrix, which serves to describe the nominal structure of a given graph D; it acknowledges the fact that some nodes of the graph D stay disconnected (one way or both ways) from other nodes throughout the dynamic process (for details see Šiljak [6]). Now, we line up all N rows of Ē to get a binary vector ē R N 2 defined as ē = (ē 11, ē 12,..., ē 1N ; ē 21, ē 22,..., ē 2N ;... ; ē N1, ē N2,..., ē N N ). (25)

6 6 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) Fig. 1. Dynamic graph. Then, we delete all fixed zeros from ē preserving the order of the nonzero elements, re-enumerate the remaining nonzero elements in natural order to get a new shorter vector ē R M and, finally, remove the overbars to get the vector e = (e 1, e 2,..., e M ) T, where superscript T denotes transpose. Once the vector e is set, the function g(t, e) follows from the matrix function F(t, E). Example 1. To illustrate the concept of dynamic graphs at this early stage of development, let us consider a simple example of a dynamic graph D shown in Fig. 1 having the adjacency matrix [ ] 0 e12 (t) E(t) =. (27) e 21 (t) 0 To vectorize this matrix we write the fundamental adjacency matrix [ ] [ ] 0 ē Ē = = ē and get the vector ē = (0, ē 12, ē 21, 0) T = (0, 1, 1, 0) T. Now, we remove the zeros from the vector ē and overbars from its elements, re-enumerate the nonzero elements as e 12 (t) = e 1 (t), e 21 (t) = e 2 (t), and get the state vector e = (e 1, e 2 ) T R 2, so that N 2 = 4, but M = 2. We assume that the functional dependence between the lines of graph D, that is, elements e 12 and e 21 of the matrix E, are such that Eq. (23), which describes the evolution in time of the dynamic matrix E, is given by two scalar equations E : ė 1 = 3e 1 + 4e 2 1 e e 1e 2 ė 2 = 2.1e 2 + e 1 e 2 (29) which can serve as a model for the predator prey interaction: e 1 a prey and e 2 a predator [18]. Although the interactions among the edges of a graph can be chosen in arbitrary ways, in selecting the predator pray model we want to indicate the possibility that interactions among the edges of a graph may mimic the interactions among species in a population community. A portrait of the motions of E is given in Fig. 2. To compute the equilibriums of E, we solve the algebraic equations, 3e 1 + 4e 2 1 e e 1e 2 = 0 2.1e 2 + e 1 e 2 = 0 and get the equilibrium states [ ] [ ] [ ] [ ] e1 e 0 =, e e =, e e =, e e = 0 which are denoted by stars in Fig. 2. We observe that the first two equilibria are asymptotically stable, while the remaining two are not. These facts can be verified by linearization of E at the equilibrium states. The corresponding equilibrium adjacency matrices E e 1 = [ ], E e 2 = [ ] [ ] [ ], E3 e 0 0 =, E e = 3 0 (26) (28) (30) (31) (32)

7 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 7 Fig. 2. State portrait. Fig. 3. Equilibrium graphs. have binary representations [ ] [ ] [ ] [ ] Ē1 e 0 0 =, Ē e =, Ē e =, Ē e =. (33) 1 0 Since the first two equilibria e1 e and ee 2 are asymptotically stable, the two equilibrium matrices Ee 1 and Ee 2 correspond to emerging interconnection structures from the dynamic process represented by adjacency matrices Ē1 e and Ē2 e or, equivalently, by the first and second graph, De 1 and De 2, in Fig. 3. Since connectedness is the fundamental feature of graphs [1], we observe that D e 1 is disconnected, while graph De 2 is strongly connected (i.e., strong). The third and forth graph, D e 3 and De 4, of Fig. 3, which cannot persist in a real world applications, are (strongly) unilateral. An interesting observation, which can be made in Fig. 2, is that dynamic

8 8 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) graph D can start as a strongly connected graph D0 1 at the initial condition e1 0 = (4, 3)T, but end as a disconnected equilibrium graph D e 1 at the origin ee 1 = 0. On the other hand, if graph D starts at D2 0 arbitrarily close to the equilibrium D e 4, which is an unilateral graph, it would end at equilibrium De 2 which is a strongly connected graph. Example 2. Let us now look at the Example 1 from a different point of view, which was proposed in [6]. We can use parametric differential equations S : ẋ 1 = 3x 1 + 4x 2 1 x e 21(t, x)x 2 ẋ 2 = 2.1x 2 + e 12 (t, x)x 1 (34) to define a dynamic adjacency matrix [ ] 0 e E(t, x) = 12 (t, x) e 21 (t, x) 0 where x R 2 is the parameter vector. If we choose e 12 (t, x) = x 1 and e 21 (t, x) = x 2, we would obtain the same diagram of Fig. 2 illustrating the motions of the graph D of Example 1. An interesting aspect of the form of S is that each variable e 12 (t, x) and e 21 (t, x) can be viewed as either input or output of S, as well as uncertain functions representing structural perturbations of S. 4. Positive graphs In a wide variety of applications, graphs are required to have nonnegative weights [1,6]. In the context of dynamic graphs this restriction means that if the graph trajectory starts with nonnegative weights it stays with nonnegative weights for all time. This further implies that the corresponding dynamic adjacency matrix E, as defined in Eq. (23), would be a positive dynamic system [19,20,6]. Recently, breaking with tradition, such systems were termed nonnegative systems [21,22]. We assume that function g(t, e) in (23) is continuous with respect to t and e, and is sufficiently smooth so that for each initial condition e 0 R M, there is a unique solution e(t; t 0, e 0 ) of (23) defined for all t t 0. We also recall the notation (35) R M + = {e RM : e i 0, i M} (36) of the positive (nonnegative) orthant in the state space R M of the dynamic adjacency matrix E, where M = {1, 2,..., M}. Then, Definition 5. Adjacency matrix E is a positive dynamic system if e 0 R + M implies e(t; t 0, e 0 ) R + M for all t t 0, that is, R + M is an invariant set of E. To characterize the invariance of R M + in terms of the function g(t, e), we state the following [19]: Theorem 1. If the function g(t, e) is nonnegative, that is, whenever g i (t, e 1,..., e i 1, 0, e i+1,..., e M ) 0, (t, e) R R M +, i M (37) e j 0, j = 1,..., i 1, i + 1,..., M, (38) then to every initial condition e 0 R M +, there corresponds a nonnegative solution e(t; t 0, e 0 ) R M +. Proof. Let us associate with dynamic matrix E the auxiliary system E ε : ė 1 = g 1 (t, e 1,..., e M ) + ε ė 2 = g 2 (t, e 1,..., e M ) + ε ė M = g M (t, e 1,..., e M ) + ε (39)

9 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 9 where ε > 0. Set R + M is invariant with respect to the auxiliary system E ε since at every point of intersection of a solution of E ε with the boundary of R + M, the components of the solution (which are zero) are increasing. When we let ε 0, the solutions of the auxiliary system E ε become solutions of the original system E. Since R + M is a closed set, the theorem follows. The dynamic adjacency matrix of Example 1, which is described by Eq. (29), is a positive system, as can be verified by Theorem 1 and directly from the motions in Fig. 2. Eq. (29), which describe a predator prey interaction between a prey species, density e 1, and its predator, density e 2, belongs to the class of Lotka Volterra models in population biology [23]. In terms of the graph in Fig. 1, which is represented by Eq. (29), this means that the rate at which the weight (prey) e 12 = e 1 is decreasing is proportional to the product e 21 e 12 of weights e 21 and e 12 (densities of predator and prey). For details regarding the interpretation of the sign and magnitude of population interactions (see Logofet [24] and references therein). We view the lines of a graph as interacting species in a population community, and their weights as size of the individual species. This type of interaction between the weights of the lines of a graph arises in modeling of a large variety of natural and man-made systems [6]. Now, we want to generalize Example 1 for graphs with M lines (edges), and choose the model E : ė = R(Pe + Qv + r) where, as before, e R M is the state of the dynamic matrix E. The system matrices R = diag{e 1, e 2,..., e M }, P = ( p i j ), and Q = (qi j ) are constant and have appropriate dimensions. The presence of inputs v R L and r R L represents a departure from the standard Lotka Volterra model. Constant vector r R L will be used to set equilibrium matrix E e at a desired location in the positive orthant R M + = {e RM : e i > 0, i M}, while input v will be used to stabilize it. We start the stability analysis of dynamic matrix E : ė = R(Pe + r). without the input v. Equilibriums of E are determined by the equation R(Pe + r) = 0. Since we are interested in equilibria e e within R + M, we follow [25] and represent ee as [ e e e e ] = α, e e β where e e α Rk +, 0 k M, and ee β = 0. We note that if Eq. (41) has a positive solution ee, the solution is unique and given by e e = P 1 r. To consider all equilibria in R + M, we introduce the vector q = Pe e + r and impose the decomposition [ ] [ rα Pαα r =, P = r β P βα ] P αβ, P ββ q = Eq. (42) implies q α = 0 and we get q β = P βα e e α + r β resulting in system equation E : ė = R[P(e e e ) + r], which is written explicitly in terms of the equilibrium e e we are interested in. [ qα (40) (41) (42) (43) (44) q β ]. (45) Remark 1. Since E is a positive system and R + M is an invariant space, the global asymptotic stability condition (iii) in Definition 4 is restricted to R+ M. In other words, the equilibrium ee R+ M is required to be stable and attractive with respect to R+ M, so that e 0 R+ M implies e(t; t 0, e 0 ) R+ M and e(t; t 0, e 0 ) e e as t. This is R+ M -stability [25]. (46) (47)

10 10 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) To establish R+ M -stability of equilibrium ee R+ M, we can use the Volterra-type Lyapunov function [26], V (e) = k δ i [e i ei e ei e ln(e i/ei e )] + i=1 M i=k+1 δ i e i, (48) where δ i s are all positive numbers. The function V : R+ M {ee } R + M is continuously differentiable and positive definite on R+ M, that is, V (e) > 0 for all e RM + {ee }, and V (e e ) = 0. Furthermore, to get R+ M -stability we note that V (e) + when, either e i 0 for some i = 1, 2,..., k, or e +. By computing the Lyapunov derivative V (e) E with respect to (47), we get V (e) E = k δ i (e i ei e )ė i/e i + i=1 M i=k+1 δ i ė i ( e e e) T Π ( e e e ) + e T β βe β (49) where e β = (e k+1, e k+2,..., e M ) T, β = diag{δ k+1, δ k+2,..., δ M }, Π = 1 2 (PT + P), (50) and = diag{δ 1, δ 2,..., δ k }. By applying the Lyapunov method, which requires that V (e e ), V (e) E > 0, we arrive at the following [25]: Theorem 2. The equilibrium e e of the dynamic adjacency matrix E is R+ M -stable if there exists a positive diagonal matrix, such that the matrix Π is positive definite, and q β is a nonpositive vector. Remark 2. We note that nonpositivity of vector q β in Theorem 2 takes place whenever the equilibrium e e is the solution of equation Pe = r, that is, e e = P 1 r. From Theorem 1 we conclude that if e e R + M, then all one needs for stability of the equilibrium e e is the existence of which makes Π positive definite. In order to test efficiently the stability condition of Theorem 2, we can use the theory of M-matrices (e.g., Kaszkurewicz and Bhaya [27]). We first recall the class of matrices Z M = {A R M M : a i j 0, i j; i, j M}. (52) The set M M of M-matrices is a subset of Z M, and we choose only a few characterizations of the set M M : Theorem 3. Let A Z M. Then, the following statements are equivalent: (i) A M M (ii) A is Hurwitz (iii) There exist numbers d i > 0, i M, such that d i a ii > (iv) All leading principal minors of A are positive (v) There exists a matrix = diag{δ 1, δ 2,..., δ M }, δ i > 0, i M, such that the matrix Π = A T + A is positive definite. M d j a i j, i M To establish a link between M-matrices of Theorem 3 and the stability Theorem 2, let us consider a case when the matrix P of the edge community of a given graph D is such that P Z M. This means that P has nonnegative off-diagonal elements and, therefore, is a Metzler matrix, which was introduced in modeling of the competitive equilibrium of multiple market systems (e.g., Arrow [28]). In terms of multiple market systems, we view the edges j i (51) (53)

11 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 11 of graph D as competing commodities and their weights as prices. Since the off-diagonal elements of a Metzler matrix are nonnegative, we have a gross-substitute case, implying that a weight increase (decrease) on an edge of the corresponding graph D would cause a weight increase (decrease) on all existing edges of D (for details see Šiljak [6]). From Theorem 3 we have the following corollary to Theorem 2: Corollary 1. Let P be a Metzler matrix, that is P Z M. Then, there exists a positive diagonal matrix, such that the matrix Π is positive definite if and only if P M M. If, in addition to P M M, we have that q β is a nonpositive vector, then the equilibrium e e of the dynamic adjacency matrix E is R M + -stable. We note that the defining matrix set Z M requires off-diagonal elements p i j to be nonpositive, which is too restrictive in modeling of multispecies communities with predation (as in Example 1 above), where an increase in the predator population causes a decrease in the prey species [23], requiring negative off-diagonal elements in the community matrix P. Negative off-diagonal elements in the context of competitive equilibrium imply the existence of complementary goods and appearance of the Morishima matrix [6,29]. In dynamic models of graphs, this fact means that an increase of weight e j causes a decrease of weight e i if p i j < 0. To capture the possibility of negative elements appearing in arbitrary patterns within the matrix P, we introduce the comparison matrix ˆP = ( ˆp i j ) R M M as ˆp i j = p i j, i = j = p i j, i j (54) which is the negative of McKenzie s diagonal form [6]. It has been shown by McKenzie that if matrix ˆP is quasidominant diagonal, that is, satisfies (iv) of Theorem 3, then P is Hurwitz. The connection of McKenzie s form and the Lyapunov part (v) of Theorem 3 was established by Moylan and Hill [30]. We re-state the result to fit our context: Theorem 4. If ˆP M M, then there exists a positive diagonal matrix such that the matrix Π = (P T + P), (55) is positive definite. Now, we go back to the original model (40) of E to explore the role of the input v in the stabilization of a positive equilibrium e e located in R+ M. At the end of this section we will show how the constant bias vector r can be used to set a stable equilibrium at a desired location. We start with the auxiliary linear system L : ė = Pe + Qv, (56) and seek a state feedback control law v = K e (57) which makes the function V (e) = e T e (58) with a diagonal matrix of (51) a Lyapunov function for the closed-loop system L K : ė = (P + QK )e. (59) Sufficient conditions for this kind of stability are well known, and can be expressed as a pair of matrix inequalities, > 0 (P + QK ) T + (P + QK ) > 0. (60) With the change of variables L = K Y, where Y = 1, the solution for the gain matrix K can be formulated as a Linear Matrix Inequalities (LMI) problem [31],

12 12 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) Y > 0 P T Y + Y P + QL + L T Q T < 0 [ κl I L T ] [ ] Y I < 0; > 0. L I I κ Y I The gain matrix is computed as (61) K = LY 1, and its norm is implicitly constrained by the last two inequalities above, which imply that K κ L κ Y. This constraint is necessary in order to prevent unacceptably high gains that an unconstrained optimization may otherwise produce (see Šiljak and Stipanović [32]). We should also note that in this context we can replace state feedback by static output feedback [33]. If we want the stronger stability result of Remark 1 with Metzler matrix (P + BK ), then proving R M + -stabilizability of system L K is equivalent to feasibility of the following LMI problem [31]: Y > 0 P T Y + Y P + QL + L T Q T < 0 (PY + QL) i j 0, i j. If we wish to limit the size of gain matrix K, we can add to this set of inequalities the last two inequalities of (61). Once we solve either of the two LMI problems above for the gain matrix K, we obtain a quadratic form (58) as a Lyapunov function for the system L K. This further means that the Lyapunov matrix equation P K T + P K = 2Π for the closed-loop matrix P K = P + QK has the solution as a positive diagonal matrix. By Theorem 2, this fact implies that the equilibrium e e of the closedloop Lotka Volterra equation E K : ė = R(P K e + r) is globally asymptotically stable with respect to the positive orthant R+ M. Finally, we note that we can use vector r to set the (stable) equilibrium e e of E K anywhere within the invariant region R+ M. In terms of the dynamic graph D K, which corresponds to the adjacency matrix E K we conclude that we can establish any set of fixed weights for the free edges of the graph D K as a stable equilibrium. This possibility will play an interesting role in the study of complex dynamic system carried out in the next section. There is a wealth of information about the behavior of positive systems that can be used in the context of positive graphs. Besides the theory of competitive equilibrium in economics and population biology discussed above, positive systems have been used as models in areas as diverse as interactions in social groups [34], arms race [35 37], compartmental and chemical systems [38 40], impulsive and hybrid systems [21], and biological and physiological models [22]. For a recent comprehensive treatment of positive systems, see [27]. 5. Complex dynamic systems At the outset of modeling complex systems, which are composed of interconnected subsystems, directed graphs have been introduced [5,6] to define and interpret the interconnection structure underlying the dynamics of the interacting subsystems. Subsystems were associated with vertices while interconnections with edges of the graph. In order to allow for accidental and intentional changes in the interconnection structure, which are always present in real world applications, the graph was assumed to vary as a function of time and the state of the system. To capture the effect of changing structure on the stability of large complex systems, the concept of connective stability was introduced as Lyapunov s stability under structural perturbations [5,6]. A system is considered connectively stable (62) (63) (64) (65) (66)

13 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 13 if it remains stable despite perturbations whereby subsystems are disconnected and reconnected in uncertain ways during its motion. Our objective in this section is to initiate a systematic study of dynamic interconnections within complex systems using the proposed definition of dynamic graphs, which is rooted in the concept of connective stability. This will considerably broaden the modeling of complex systems by giving the dynamic interconnections the same level of importance that the dynamic subsystems have in shaping the behavior of complex dynamic systems. A dynamic graph (interconnection matrix) of a complex system can now be used as a basis for coordination of interconnection strength among the individual subsystems in a smooth fashion until interconnection levels reach preassigned steadystate values. A model for a complex system is now represented as a pair of coupled systems S : ẋ = f (t, x, u, E) C : ė = g(t, e, v, x) where x R n is the state of the system S, u R m is its input, and E = (e i j ) is an N N interconnection matrix. The vectors e R M and v R L are the state and input of the coordinator C. We assume that the infinitesimal transition functions f (t, x, u, E) and g(t, e, v, x) are sufficiently smooth, so that solution (x, e)(t; t 0, x 0, e 0, u, v) of interconnected system S&C exists and is unique for all initial conditions (t 0, x 0, e 0 ) and fixed piecewise continuous inputs u(t) and v(t). At this stage of development, it is necessary to add more structure to the system S&C. We consider S as an interconnected system S : ẋ i = a i (t, x i, u i ) + b i (t, x, E), i N (68) composed of N subsystems (67) S i : ẋ i = a i (t, x i, u i ), i N (69) with states x i R n i, so that the state x = (x1 T, xt 2,..., xt N )T R n of the system S is a concatenation of (disjoint, not overlapping) subsystems states. Each subsystem S i has its own input u i R m i, as a part of the overall input u = (u T 1, ut 2,..., ut N )T of system S, and N = {1, 2,..., N}. A standard role of the input u i is to stabilize the corresponding subsystem S i using only the locally available state (or output). There are well-known methods to make the decentralized control strategy produce a connectively stable overall system S [10,32]. We now want to discuss the less understood coordination task and assume that each subsystem S i : ẋ i = a i (t, x i ), i N (70) has an equilibrium at the origin (x e i = 0) which is stable, or has been stabilized by decentralized feedback. To assign the proper role to the elements e i j of the adjacency matrix E appearing in the interconnections b i (t, x, E) of S, we recall [6] the formulation of interconnections, b i (t, x, e) = b i (t, e i1 x 1, e i2 x 2,..., e i N x N ), i N. (71) According to this form, the elements e i j (t, x) are functions e i j : R R n [0, ê i j ], which are continuous in both arguments and belong to intervals [0, ê i j ], where ê i j are positive numbers. The elements represent the strength of coupling between the individual subsystems, including disconnections. Example 2 provides an illustration of the proposed form of system S. We also assume that interconnection functions b i (t, x, E) are such that the overall system S has the equilibrium at the origin (x e = 0), that is, b i (t, 0, E) = 0 for all t R and all E such that 0 E Ê with element-by-element inequalities (or, equivalently, E [0, Ê]). The infinitesimal transition function g(t, e, v, x) of the coordinator C has no special form; its construction is guided only by our desire to change the structure of the complex system S in a suitable way. We want to maintain interactions between subsystems S i at preassigned levels that are defined by elements e e i j of the equilibrium matrix Ee. We also want the equilibrium E e to be reached from all initial matrices E 0 within a permissible region containing E e, without disturbing the stability of the overall system S. Recalling the original definition of connective stability [6], we observe that the coordinator C produces dynamic structural perturbations of the interconnected system S. Our task is to define

14 14 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) a suitable concept of connective stability of equilibrium (0, e e ) of S&C, which can guarantee that the perturbations do not destroy its stability. We first assume that the input v has been utilized to set the equilibrium e e within its region of attraction Ω C R M +, where the equilibrium vector e e corresponds to the equilibrium matrix E e. Then, the function g(t, e, v, x), takes the form g(t, e, x) with g(t, e e, 0) = 0 for all t R. Finally the complex system S&C becomes S : ẋ i = a i (t, x i ) + b i (t, e i1 x 1, e i2 x 2,..., e i N x N ), C : ė = g(t, e, x). and we state i N (72) Definition 6. Complex system S is dynamically connectively stable if the equilibrium x e = 0 is globally asymptotically stable for all e 0 Ω C. To provide an interesting interpretation of the new idea of connective stability advanced by Definition 6, let us consider the joint system S&C described as S : ẋ i = a i (t, x i ) + b i (t, e i1 x 1, e i2 x 2,..., e i N x N ), C : ė = R(Pe + r). where system S remains unchanged, while the coordinator is described by a Lotka Volterra model with a stable matrix P (or stabilized by feedback). We first note the absence of the state x in the description of coordinator C, which implies a hierarchical structure of the joint system S&C [10]. Therefore, we can establish stability of the equilibrium e e R+ M of C and have a Volterra-type Lyapunov function to prove it. Now, what is left to do is to show stability of the equilibrium x e = 0 of S, which can withstand the structural perturbation caused by motion of the coordinator C restricted to the region of attraction Ω C R+ M ; this is the dynamic connective stability of S formulated in Definition 6. To derive conditions for the new kind of connective stability, we start by showing ordinary connective stability of the interconnected system S following the standard recipe [10], which is based on the Matrosov Bellman concept of vector Lyapunov functions [41,42]. Let us associate with each subsystem S i a scalar Lyapunov function V i : R R n i R +, such that V i (t, x i ) is a continuous function in both arguments and satisfies a Lipschitz condition in x i, that is, for some κ i > 0, V i (t, x i ) V i(t, x i ) κ i x i x i, i N We further assume that the function V i (t, x i ) satisfies the inequalities (73) t R x i, x i R n i. (74) φ 1i ( x i ) V i (t, x i ) φ 2i ( x i ) D + V i (t, x i ) Si φ 3i ( x i ), (t, x i ) R R n i (75) where φ 1i, φ 2i K, φ 3i K are Hahn s comparison functions (e.g., Šiljak [6]), and D + V i (t, x i ) Si is the Dini derivative computed with respect to S i. As for the interconnections b i (t, x, e), we impose the following bounds: b i (t, e i1 x 1, e i2 x 2,..., e i N x N ) N ê i j φ 3 j ( x i ), i N (76) j=1 where we deviated from the standard recipe [10] and, instead of the binary N N fundamental interconnection matrix Ē = (ē i j ), we use a more flexible N N bounding matrix Ê = (ê i j ), where ê i j are positive real numbers (ê i j 0 if and only if the corresponding element ē i j = 1). In order to state the condition for connective stability of system S, we define the constant N N test matrix Ŵ = (ŵ) as ŵ i j = 1 ê ii κ i i = j = ê i j κ i i j. (77)

15 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 15 Fig. 4. Stability regions. We note that matrix Ŵ has nonpositive off-diagonal elements, that is, Ŵ Z N and state without proof a straightforward modification of Theorem 2.5 of [10]: Theorem 5. The equilibrium x e = 0 of system S is connectively stable if matrix Ŵ is an M-matrix. To establish the dynamic connective stability of system S we first recall [10] that connective stability means that the equilibrium x e = 0 of S is (globally and asymptotically) stable for all E [0, Ê]. The next step is to convert this matrix statement to a statement involving state vector e of the coordinator C. By vectorizing the matrix Ê R N N, we obtain the vector ê R M. Then, connective stability implies that system S is stable for all e such that 0 e i ê i, i N. Now, Theorem 5 implies that the system S is connectively stable provided the uncertain vector e(t, x) is contained within the region (box), Ω S = {e R M : e i [0, ê i ], i N}. (78) The containment will obviously take place if a region of attraction Ω C of the equilibrium e e is itself contained within the connective stability region Ω S, and initial states e 0 are restricted to the region Ω C. Having the Volterra function V (e) available as a Lyapunov function for the coordinator C, we can compute a region of attraction as Ω C = {e R M + : V (e) < θ}, (79) where θ is a positive number. We finally arrive at the following: Theorem 6. The equilibrium x e = 0 of system S is dynamically connectively stable if Ω C Ω S. In Fig. 4, we illustrate the meaning of the inclusion Ω C Ω S, by borrowing the region of attraction Ω C of the equilibrium e e 2 = (2.10, 1.98)T from Example 1 (this region was computed by Parker and Chua in [18]). Now, if we establish the connective stability of an interconnected system S with respect to the rectangle Ω S specified by the fixed interconnection matrix Ê in Fig. 4, then the system is dynamically connectively stable with respect to region Ω C situated inside the rectangle Ω S. That is, e 0 Ω C implies e(t; t 0, e 0 ) Ω C Ω S for all t > t 0, and dynamic connective stability follows from the connective stability of system S. Remark 3. Theorem 6 poses an interesting numerical problem: How to maximize the sizes of regions Ω C and Ω S while retaining the inclusion condition Ω C Ω S. The larger the sizes of the two regions the larger the class of uncertain interconnections that can be neutralized by the coordinator. We should also note that over the years a large number of extensions and applications of the connective stability concept have been proposed. A wide variety of structural perturbations have been studied, starting as unknown but bounded functions of time and state [5,6]. Subsequently, modeling of uncertain interconnection matrices has (80)

16 16 D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) been broadened to include stochastic elements [8,43], expanding the size of complex systems [44,45], discontinuous interconnection matrices [46,47], impulsive and hybrid systems [48,21], singular perturbations [11], matrix Lyapunov functions [49,50], polytopic uncertainty sets [51], discrete systems [52], fuzzy sets [53], and hybrid systems [54, 55]. Of particular interest are the methods and techniques that have been invented for the construction of robust decentralized control laws which can produce connectively stable overall interconnected systems (see books by Šiljak [6,10], and Jamshidi [56]), as well as the recent papers [32,57]. 6. Multi-agent systems: The coordinator Let us now turn to the application of the concept of dynamic graphs to the control of multi-agent systems. One of the interesting distinctions of such systems within the general class of interconnected systems is the accessibility of interconnections to control action. We consider an interconnection of agents (scalar subsystems) which all have local control inputs and are nonlinearily connected to each other. The inputs are used to decentrally stabilize each individual agent. Then, a coordinator is attached to the system as an add-on control device to adaptively adjust the interconnection levels among the agents until a prescribed state of the multi-agent system becomes a stable equilibrium state. The principal difference between the coordinator role in this context and the role it played in the preceding section, is a departure from a passive action of maintaining a set of constant interaction levels among the subsystems to an active role of changing the interconnections among the agents until a desirable equilibrium is set for the overall multi-agent system Interconnected system Let us consider a multi-agent system n S : ẋ i = ax i + e i j h j (x j ) + u i, j=1 y i = h i (x i ) i N (81) which is composed of n agents (scalar subsystems) S i : ẋ i = a i x i + u i, i N y i = h i (x i ) (82) where x, u, y R n are the state, input, and output of S, a i is a positive number, e i j s are elements of the n n interconnection matrix E, and h i (x i ) is the sigmoid function h i : R ( 1, 1), which is a diffeomorphism having the following properties: (i) h i (0) = 0, 1 < h i ( ) < 1 (ii) h i ( ) is a monotonically increasing odd function (iii) h i ( ) and h 1 i ( ) are C 1 and dh 1 i (y i ) dy i ρ 0 and N = {1, 2,..., n}. We note that the subsystems are unstable and we use a decentralized state feedback u i = k i x i + r i (83) (84) to stabilize each subsystem independently. This is a popular approach in control of complex dynamic systems for at least two reasons. First, a decentralized feedback law uses the locally available state, which is suitable either for feasibility or economic reasons (for example, subsystems may be geographically far from each other). Second, the control strategy is shown to be robust with respect to uncertain interconnections among the subsystems [10]. We choose the feedback gain k i = 1 + a i and leave the reference input r i (bias) free but constant.

17 We consider the closed-loop system ARTICLE IN PRESS D.D. Šiljak / Nonlinear Analysis: Hybrid Systems ( ) 17 S : ẋ i = x i + y i = h i (x i ) n e i j h j (x j ) + r i, j=1 C : ė i j = µg i j (e, y, y e ), i, j N i N (85) where we added a coordinator C with a coordination law g i j : R n2 R n R n R. We shall use the coordinator to drive the interconnection vector e until a desired fixed output y e is achieved as an asymptotically stable equilibrium of S. To accomplish the equilibrium assignments objective, we should establish dynamic connective stability of the interconnected two-time-scale system S&C for bounded interconnection weights e i j by choosing a suitable coordination rate (small parameter) µ (0 < µ 1). We note that we shall use matrix E R n n and vector e R n2 interchangeably to fit the context; they are equivalent, with corresponding components related by the formula e i j = e j+(n i). The system S&C was used in [11] as a model for artificial neural networks with learning, which relied on the standard notion of connective stability involving a unit hypercube of uncertain matrices E = (e i j ). By replacing the hypercube by a convex polytope, we use the new concept of polytopic connective stability [51] which can significantly increase the flexibility of coordinator C in reshaping interconnections of the system S. We bear in mind that the use of the broader concept is made possible by the two-time-scale property of the system, which allows us to consider interconnection matrices within a polytope as constant ( frozen ) in time. We begin our analysis by representing the system S in the output space, S : ẏ = B(y)[ h(y) + Ey + r] C : ė = µg(e, y, y e ) (86) where [ dh 1 i (y i ) B(y) = diag{b 1, b 2,..., b n }, b i = dy i h(y) = [h 1 1 (y 1), h 1 2 (y 2),..., h 1 n (y n)] T ] 1 and r = (r 1, r 2,..., r n ) T. For a sufficiently small coordination rate µ, the interconnection matrix E is considered frozen, and each equilibrium ȳ = (ȳ 1, ȳ 2,..., ȳ n ) of S satisfies the equation h 1 i (ȳ i ) + n e i j ȳ j + r i = 0, i N. (88) j=1 We need the following assumptions: (A 1 ) Each quasi steady-state ȳ(e) is a function of the frozen matrix E living in a convex polytope (87) P = conv {E k }, which is a convex hull of fixed vertex matrices E k, k V = {1, 2,..., v}. (A 2 ) For each E P, there exist isolated equilibria ȳ(e) of the system S. (89) As pointed out in [11], using the coordinator C to set an equilibrium of system S at a desired location y e may cause mutual interference of coordinator and system dynamics causing instability of the interconnected system S&C. A satisfactory separation of the two dynamics is achieved by choosing a sufficiently small coordination rate µ that guaranties the existence of an integral manifold [58] containing the slow dynamics and the desired equilibria.

On graph differential equations and its associated matrix differential equations

On graph differential equations and its associated matrix differential equations Malaya Journal of Matematik 1(1)(2013) 1 9 On graph differential equations and its associated matrix differential equations J. Vasundhara Devi, a, R.V.G. Ravi Kumar b and N. Giribabu c a,b,c GVP-Prof.V.Lakshmikantham

More information

An introduction to Mathematical Theory of Control

An introduction to Mathematical Theory of Control An introduction to Mathematical Theory of Control Vasile Staicu University of Aveiro UNICA, May 2018 Vasile Staicu (University of Aveiro) An introduction to Mathematical Theory of Control UNICA, May 2018

More information

Copyrighted Material. 1.1 Large-Scale Interconnected Dynamical Systems

Copyrighted Material. 1.1 Large-Scale Interconnected Dynamical Systems Chapter One Introduction 1.1 Large-Scale Interconnected Dynamical Systems Modern complex dynamical systems 1 are highly interconnected and mutually interdependent, both physically and through a multitude

More information

Multi-Robotic Systems

Multi-Robotic Systems CHAPTER 9 Multi-Robotic Systems The topic of multi-robotic systems is quite popular now. It is believed that such systems can have the following benefits: Improved performance ( winning by numbers ) Distributed

More information

Decentralized Stabilization of Heterogeneous Linear Multi-Agent Systems

Decentralized Stabilization of Heterogeneous Linear Multi-Agent Systems 1 Decentralized Stabilization of Heterogeneous Linear Multi-Agent Systems Mauro Franceschelli, Andrea Gasparri, Alessandro Giua, and Giovanni Ulivi Abstract In this paper the formation stabilization problem

More information

Robust Stability. Robust stability against time-invariant and time-varying uncertainties. Parameter dependent Lyapunov functions

Robust Stability. Robust stability against time-invariant and time-varying uncertainties. Parameter dependent Lyapunov functions Robust Stability Robust stability against time-invariant and time-varying uncertainties Parameter dependent Lyapunov functions Semi-infinite LMI problems From nominal to robust performance 1/24 Time-Invariant

More information

Chapter One. Introduction

Chapter One. Introduction Chapter One Introduction With the ever-increasing influence of mathematical modeling and engineering on biological, social, and medical sciences, it is not surprising that dynamical system theory has played

More information

A Decentralized Stabilization Scheme for Large-scale Interconnected Systems

A Decentralized Stabilization Scheme for Large-scale Interconnected Systems A Decentralized Stabilization Scheme for Large-scale Interconnected Systems OMID KHORSAND Master s Degree Project Stockholm, Sweden August 2010 XR-EE-RT 2010:015 Abstract This thesis considers the problem

More information

EN Nonlinear Control and Planning in Robotics Lecture 3: Stability February 4, 2015

EN Nonlinear Control and Planning in Robotics Lecture 3: Stability February 4, 2015 EN530.678 Nonlinear Control and Planning in Robotics Lecture 3: Stability February 4, 2015 Prof: Marin Kobilarov 0.1 Model prerequisites Consider ẋ = f(t, x). We will make the following basic assumptions

More information

Convergence Rate of Nonlinear Switched Systems

Convergence Rate of Nonlinear Switched Systems Convergence Rate of Nonlinear Switched Systems Philippe JOUAN and Saïd NACIRI arxiv:1511.01737v1 [math.oc] 5 Nov 2015 January 23, 2018 Abstract This paper is concerned with the convergence rate of the

More information

1 Lyapunov theory of stability

1 Lyapunov theory of stability M.Kawski, APM 581 Diff Equns Intro to Lyapunov theory. November 15, 29 1 1 Lyapunov theory of stability Introduction. Lyapunov s second (or direct) method provides tools for studying (asymptotic) stability

More information

Nonlinear Systems and Control Lecture # 12 Converse Lyapunov Functions & Time Varying Systems. p. 1/1

Nonlinear Systems and Control Lecture # 12 Converse Lyapunov Functions & Time Varying Systems. p. 1/1 Nonlinear Systems and Control Lecture # 12 Converse Lyapunov Functions & Time Varying Systems p. 1/1 p. 2/1 Converse Lyapunov Theorem Exponential Stability Let x = 0 be an exponentially stable equilibrium

More information

Computational Aspects of Aggregation in Biological Systems

Computational Aspects of Aggregation in Biological Systems Computational Aspects of Aggregation in Biological Systems Vladik Kreinovich and Max Shpak University of Texas at El Paso, El Paso, TX 79968, USA vladik@utep.edu, mshpak@utep.edu Summary. Many biologically

More information

Near-Potential Games: Geometry and Dynamics

Near-Potential Games: Geometry and Dynamics Near-Potential Games: Geometry and Dynamics Ozan Candogan, Asuman Ozdaglar and Pablo A. Parrilo January 29, 2012 Abstract Potential games are a special class of games for which many adaptive user dynamics

More information

Chap. 3. Controlled Systems, Controllability

Chap. 3. Controlled Systems, Controllability Chap. 3. Controlled Systems, Controllability 1. Controllability of Linear Systems 1.1. Kalman s Criterion Consider the linear system ẋ = Ax + Bu where x R n : state vector and u R m : input vector. A :

More information

Near-Potential Games: Geometry and Dynamics

Near-Potential Games: Geometry and Dynamics Near-Potential Games: Geometry and Dynamics Ozan Candogan, Asuman Ozdaglar and Pablo A. Parrilo September 6, 2011 Abstract Potential games are a special class of games for which many adaptive user dynamics

More information

Lecture 15 Perron-Frobenius Theory

Lecture 15 Perron-Frobenius Theory EE363 Winter 2005-06 Lecture 15 Perron-Frobenius Theory Positive and nonnegative matrices and vectors Perron-Frobenius theorems Markov chains Economic growth Population dynamics Max-min and min-max characterization

More information

Observations on the Stability Properties of Cooperative Systems

Observations on the Stability Properties of Cooperative Systems 1 Observations on the Stability Properties of Cooperative Systems Oliver Mason and Mark Verwoerd Abstract We extend two fundamental properties of positive linear time-invariant (LTI) systems to homogeneous

More information

Global Analysis of Piecewise Linear Systems Using Impact Maps and Surface Lyapunov Functions

Global Analysis of Piecewise Linear Systems Using Impact Maps and Surface Lyapunov Functions IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL 48, NO 12, DECEMBER 2003 2089 Global Analysis of Piecewise Linear Systems Using Impact Maps and Surface Lyapunov Functions Jorge M Gonçalves, Alexandre Megretski,

More information

Gramians based model reduction for hybrid switched systems

Gramians based model reduction for hybrid switched systems Gramians based model reduction for hybrid switched systems Y. Chahlaoui Younes.Chahlaoui@manchester.ac.uk Centre for Interdisciplinary Computational and Dynamical Analysis (CICADA) School of Mathematics

More information

Chapter One. Introduction

Chapter One. Introduction Chapter One Introduction A system is a combination of components or parts that is perceived as a single entity. The parts making up the system may be clearly or vaguely defined. These parts are related

More information

Graph and Controller Design for Disturbance Attenuation in Consensus Networks

Graph and Controller Design for Disturbance Attenuation in Consensus Networks 203 3th International Conference on Control, Automation and Systems (ICCAS 203) Oct. 20-23, 203 in Kimdaejung Convention Center, Gwangju, Korea Graph and Controller Design for Disturbance Attenuation in

More information

Distributed Receding Horizon Control of Cost Coupled Systems

Distributed Receding Horizon Control of Cost Coupled Systems Distributed Receding Horizon Control of Cost Coupled Systems William B. Dunbar Abstract This paper considers the problem of distributed control of dynamically decoupled systems that are subject to decoupled

More information

Topic # /31 Feedback Control Systems. Analysis of Nonlinear Systems Lyapunov Stability Analysis

Topic # /31 Feedback Control Systems. Analysis of Nonlinear Systems Lyapunov Stability Analysis Topic # 16.30/31 Feedback Control Systems Analysis of Nonlinear Systems Lyapunov Stability Analysis Fall 010 16.30/31 Lyapunov Stability Analysis Very general method to prove (or disprove) stability of

More information

MCE/EEC 647/747: Robot Dynamics and Control. Lecture 12: Multivariable Control of Robotic Manipulators Part II

MCE/EEC 647/747: Robot Dynamics and Control. Lecture 12: Multivariable Control of Robotic Manipulators Part II MCE/EEC 647/747: Robot Dynamics and Control Lecture 12: Multivariable Control of Robotic Manipulators Part II Reading: SHV Ch.8 Mechanical Engineering Hanz Richter, PhD MCE647 p.1/14 Robust vs. Adaptive

More information

1 Differentiable manifolds and smooth maps

1 Differentiable manifolds and smooth maps 1 Differentiable manifolds and smooth maps Last updated: April 14, 2011. 1.1 Examples and definitions Roughly, manifolds are sets where one can introduce coordinates. An n-dimensional manifold is a set

More information

Disturbance Attenuation Properties for Discrete-Time Uncertain Switched Linear Systems

Disturbance Attenuation Properties for Discrete-Time Uncertain Switched Linear Systems Disturbance Attenuation Properties for Discrete-Time Uncertain Switched Linear Systems Hai Lin Department of Electrical Engineering University of Notre Dame Notre Dame, IN 46556, USA Panos J. Antsaklis

More information

Stabilization of fixed modes in expansions of LTI systems

Stabilization of fixed modes in expansions of LTI systems Systems & Control Letters 57 (28) 365 37 www.elsevier.com/locate/sysconle Stabilization of fixed modes in expansions of LTI systems Srdjan S. Stanković a,, Dragoslav D. Šiljak b a Faculty of Electrical

More information

ON COST MATRICES WITH TWO AND THREE DISTINCT VALUES OF HAMILTONIAN PATHS AND CYCLES

ON COST MATRICES WITH TWO AND THREE DISTINCT VALUES OF HAMILTONIAN PATHS AND CYCLES ON COST MATRICES WITH TWO AND THREE DISTINCT VALUES OF HAMILTONIAN PATHS AND CYCLES SANTOSH N. KABADI AND ABRAHAM P. PUNNEN Abstract. Polynomially testable characterization of cost matrices associated

More information

Impulsive Stabilization and Application to a Population Growth Model*

Impulsive Stabilization and Application to a Population Growth Model* Nonlinear Dynamics and Systems Theory, 2(2) (2002) 173 184 Impulsive Stabilization and Application to a Population Growth Model* Xinzhi Liu 1 and Xuemin Shen 2 1 Department of Applied Mathematics, University

More information

Hybrid decentralized maximum entropy control for large-scale dynamical systems

Hybrid decentralized maximum entropy control for large-scale dynamical systems Nonlinear Analysis: Hybrid Systems 1 (2007) 244 263 www.elsevier.com/locate/nahs Hybrid decentralized maximum entropy control for large-scale dynamical systems Wassim M. Haddad a,, Qing Hui a,1, VijaySekhar

More information

Feedback stabilisation with positive control of dissipative compartmental systems

Feedback stabilisation with positive control of dissipative compartmental systems Feedback stabilisation with positive control of dissipative compartmental systems G. Bastin and A. Provost Centre for Systems Engineering and Applied Mechanics (CESAME Université Catholique de Louvain

More information

Lecture 4. Chapter 4: Lyapunov Stability. Eugenio Schuster. Mechanical Engineering and Mechanics Lehigh University.

Lecture 4. Chapter 4: Lyapunov Stability. Eugenio Schuster. Mechanical Engineering and Mechanics Lehigh University. Lecture 4 Chapter 4: Lyapunov Stability Eugenio Schuster schuster@lehigh.edu Mechanical Engineering and Mechanics Lehigh University Lecture 4 p. 1/86 Autonomous Systems Consider the autonomous system ẋ

More information

On the simultaneous diagonal stability of a pair of positive linear systems

On the simultaneous diagonal stability of a pair of positive linear systems On the simultaneous diagonal stability of a pair of positive linear systems Oliver Mason Hamilton Institute NUI Maynooth Ireland Robert Shorten Hamilton Institute NUI Maynooth Ireland Abstract In this

More information

MULTI-AGENT TRACKING OF A HIGH-DIMENSIONAL ACTIVE LEADER WITH SWITCHING TOPOLOGY

MULTI-AGENT TRACKING OF A HIGH-DIMENSIONAL ACTIVE LEADER WITH SWITCHING TOPOLOGY Jrl Syst Sci & Complexity (2009) 22: 722 731 MULTI-AGENT TRACKING OF A HIGH-DIMENSIONAL ACTIVE LEADER WITH SWITCHING TOPOLOGY Yiguang HONG Xiaoli WANG Received: 11 May 2009 / Revised: 16 June 2009 c 2009

More information

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games Alberto Bressan ) and Khai T. Nguyen ) *) Department of Mathematics, Penn State University **) Department of Mathematics,

More information

Price of Stability in Survivable Network Design

Price of Stability in Survivable Network Design Noname manuscript No. (will be inserted by the editor) Price of Stability in Survivable Network Design Elliot Anshelevich Bugra Caskurlu Received: January 2010 / Accepted: Abstract We study the survivable

More information

Navigation and Obstacle Avoidance via Backstepping for Mechanical Systems with Drift in the Closed Loop

Navigation and Obstacle Avoidance via Backstepping for Mechanical Systems with Drift in the Closed Loop Navigation and Obstacle Avoidance via Backstepping for Mechanical Systems with Drift in the Closed Loop Jan Maximilian Montenbruck, Mathias Bürger, Frank Allgöwer Abstract We study backstepping controllers

More information

Chapter 7 Interconnected Systems and Feedback: Well-Posedness, Stability, and Performance 7. Introduction Feedback control is a powerful approach to o

Chapter 7 Interconnected Systems and Feedback: Well-Posedness, Stability, and Performance 7. Introduction Feedback control is a powerful approach to o Lectures on Dynamic Systems and Control Mohammed Dahleh Munther A. Dahleh George Verghese Department of Electrical Engineering and Computer Science Massachuasetts Institute of Technology c Chapter 7 Interconnected

More information

Prof. Krstic Nonlinear Systems MAE281A Homework set 1 Linearization & phase portrait

Prof. Krstic Nonlinear Systems MAE281A Homework set 1 Linearization & phase portrait Prof. Krstic Nonlinear Systems MAE28A Homework set Linearization & phase portrait. For each of the following systems, find all equilibrium points and determine the type of each isolated equilibrium. Use

More information

Solution of a Distributed Linear System Stabilisation Problem

Solution of a Distributed Linear System Stabilisation Problem Solution of a Distributed Linear System Stabilisation Problem NICTA Linear Systems Workshop ANU Brian Anderson (ANU/NICTA) with contributions from: Brad Yu, Baris Fidan, Soura Dasgupta, Steve Morse Overview

More information

Tracking control for multi-agent consensus with an active leader and variable topology

Tracking control for multi-agent consensus with an active leader and variable topology Automatica 42 (2006) 1177 1182 wwwelseviercom/locate/automatica Brief paper Tracking control for multi-agent consensus with an active leader and variable topology Yiguang Hong a,, Jiangping Hu a, Linxin

More information

MCE693/793: Analysis and Control of Nonlinear Systems

MCE693/793: Analysis and Control of Nonlinear Systems MCE693/793: Analysis and Control of Nonlinear Systems Lyapunov Stability - I Hanz Richter Mechanical Engineering Department Cleveland State University Definition of Stability - Lyapunov Sense Lyapunov

More information

The general programming problem is the nonlinear programming problem where a given function is maximized subject to a set of inequality constraints.

The general programming problem is the nonlinear programming problem where a given function is maximized subject to a set of inequality constraints. 1 Optimization Mathematical programming refers to the basic mathematical problem of finding a maximum to a function, f, subject to some constraints. 1 In other words, the objective is to find a point,

More information

2. Dual space is essential for the concept of gradient which, in turn, leads to the variational analysis of Lagrange multipliers.

2. Dual space is essential for the concept of gradient which, in turn, leads to the variational analysis of Lagrange multipliers. Chapter 3 Duality in Banach Space Modern optimization theory largely centers around the interplay of a normed vector space and its corresponding dual. The notion of duality is important for the following

More information

The Rationale for Second Level Adaptation

The Rationale for Second Level Adaptation The Rationale for Second Level Adaptation Kumpati S. Narendra, Yu Wang and Wei Chen Center for Systems Science, Yale University arxiv:1510.04989v1 [cs.sy] 16 Oct 2015 Abstract Recently, a new approach

More information

Topic 6: Projected Dynamical Systems

Topic 6: Projected Dynamical Systems Topic 6: Projected Dynamical Systems John F. Smith Memorial Professor and Director Virtual Center for Supernetworks Isenberg School of Management University of Massachusetts Amherst, Massachusetts 01003

More information

Introduction to Nonlinear Control Lecture # 3 Time-Varying and Perturbed Systems

Introduction to Nonlinear Control Lecture # 3 Time-Varying and Perturbed Systems p. 1/5 Introduction to Nonlinear Control Lecture # 3 Time-Varying and Perturbed Systems p. 2/5 Time-varying Systems ẋ = f(t, x) f(t, x) is piecewise continuous in t and locally Lipschitz in x for all t

More information

Energy-based Swing-up of the Acrobot and Time-optimal Motion

Energy-based Swing-up of the Acrobot and Time-optimal Motion Energy-based Swing-up of the Acrobot and Time-optimal Motion Ravi N. Banavar Systems and Control Engineering Indian Institute of Technology, Bombay Mumbai-476, India Email: banavar@ee.iitb.ac.in Telephone:(91)-(22)

More information

EML5311 Lyapunov Stability & Robust Control Design

EML5311 Lyapunov Stability & Robust Control Design EML5311 Lyapunov Stability & Robust Control Design 1 Lyapunov Stability criterion In Robust control design of nonlinear uncertain systems, stability theory plays an important role in engineering systems.

More information

DECENTRALIZED CONTROL DESIGN USING LMI MODEL REDUCTION

DECENTRALIZED CONTROL DESIGN USING LMI MODEL REDUCTION Journal of ELECTRICAL ENGINEERING, VOL. 58, NO. 6, 2007, 307 312 DECENTRALIZED CONTROL DESIGN USING LMI MODEL REDUCTION Szabolcs Dorák Danica Rosinová Decentralized control design approach based on partial

More information

Lecture Introduction. 2 Brief Recap of Lecture 10. CS-621 Theory Gems October 24, 2012

Lecture Introduction. 2 Brief Recap of Lecture 10. CS-621 Theory Gems October 24, 2012 CS-62 Theory Gems October 24, 202 Lecture Lecturer: Aleksander Mądry Scribes: Carsten Moldenhauer and Robin Scheibler Introduction In Lecture 0, we introduced a fundamental object of spectral graph theory:

More information

Alternative Characterization of Ergodicity for Doubly Stochastic Chains

Alternative Characterization of Ergodicity for Doubly Stochastic Chains Alternative Characterization of Ergodicity for Doubly Stochastic Chains Behrouz Touri and Angelia Nedić Abstract In this paper we discuss the ergodicity of stochastic and doubly stochastic chains. We define

More information

Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers

Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers 28 American Control Conference Westin Seattle Hotel, Seattle, Washington, USA June 11-13, 28 WeC15.1 Robust Stabilization of Non-Minimum Phase Nonlinear Systems Using Extended High Gain Observers Shahid

More information

On Linear Copositive Lyapunov Functions and the Stability of Switched Positive Linear Systems

On Linear Copositive Lyapunov Functions and the Stability of Switched Positive Linear Systems 1 On Linear Copositive Lyapunov Functions and the Stability of Switched Positive Linear Systems O. Mason and R. Shorten Abstract We consider the problem of common linear copositive function existence for

More information

MCE/EEC 647/747: Robot Dynamics and Control. Lecture 8: Basic Lyapunov Stability Theory

MCE/EEC 647/747: Robot Dynamics and Control. Lecture 8: Basic Lyapunov Stability Theory MCE/EEC 647/747: Robot Dynamics and Control Lecture 8: Basic Lyapunov Stability Theory Reading: SHV Appendix Mechanical Engineering Hanz Richter, PhD MCE503 p.1/17 Stability in the sense of Lyapunov A

More information

Stability of linear time-varying systems through quadratically parameter-dependent Lyapunov functions

Stability of linear time-varying systems through quadratically parameter-dependent Lyapunov functions Stability of linear time-varying systems through quadratically parameter-dependent Lyapunov functions Vinícius F. Montagner Department of Telematics Pedro L. D. Peres School of Electrical and Computer

More information

arzelier

arzelier COURSE ON LMI OPTIMIZATION WITH APPLICATIONS IN CONTROL PART II.1 LMIs IN SYSTEMS CONTROL STATE-SPACE METHODS STABILITY ANALYSIS Didier HENRION www.laas.fr/ henrion henrion@laas.fr Denis ARZELIER www.laas.fr/

More information

ADAPTIVE control of uncertain time-varying plants is a

ADAPTIVE control of uncertain time-varying plants is a IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 56, NO. 1, JANUARY 2011 27 Supervisory Control of Uncertain Linear Time-Varying Systems Linh Vu, Member, IEEE, Daniel Liberzon, Senior Member, IEEE Abstract

More information

Output Input Stability and Minimum-Phase Nonlinear Systems

Output Input Stability and Minimum-Phase Nonlinear Systems 422 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 47, NO. 3, MARCH 2002 Output Input Stability and Minimum-Phase Nonlinear Systems Daniel Liberzon, Member, IEEE, A. Stephen Morse, Fellow, IEEE, and Eduardo

More information

An asymptotic ratio characterization of input-to-state stability

An asymptotic ratio characterization of input-to-state stability 1 An asymptotic ratio characterization of input-to-state stability Daniel Liberzon and Hyungbo Shim Abstract For continuous-time nonlinear systems with inputs, we introduce the notion of an asymptotic

More information

arxiv: v3 [math.ds] 22 Feb 2012

arxiv: v3 [math.ds] 22 Feb 2012 Stability of interconnected impulsive systems with and without time-delays using Lyapunov methods arxiv:1011.2865v3 [math.ds] 22 Feb 2012 Sergey Dashkovskiy a, Michael Kosmykov b, Andrii Mironchenko b,

More information

Nonlinear Control. Nonlinear Control Lecture # 8 Time Varying and Perturbed Systems

Nonlinear Control. Nonlinear Control Lecture # 8 Time Varying and Perturbed Systems Nonlinear Control Lecture # 8 Time Varying and Perturbed Systems Time-varying Systems ẋ = f(t,x) f(t,x) is piecewise continuous in t and locally Lipschitz in x for all t 0 and all x D, (0 D). The origin

More information

Detailed Proof of The PerronFrobenius Theorem

Detailed Proof of The PerronFrobenius Theorem Detailed Proof of The PerronFrobenius Theorem Arseny M Shur Ural Federal University October 30, 2016 1 Introduction This famous theorem has numerous applications, but to apply it you should understand

More information

IMPROVED MPC DESIGN BASED ON SATURATING CONTROL LAWS

IMPROVED MPC DESIGN BASED ON SATURATING CONTROL LAWS IMPROVED MPC DESIGN BASED ON SATURATING CONTROL LAWS D. Limon, J.M. Gomes da Silva Jr., T. Alamo and E.F. Camacho Dpto. de Ingenieria de Sistemas y Automática. Universidad de Sevilla Camino de los Descubrimientos

More information

Quadratic and Copositive Lyapunov Functions and the Stability of Positive Switched Linear Systems

Quadratic and Copositive Lyapunov Functions and the Stability of Positive Switched Linear Systems Proceedings of the 2007 American Control Conference Marriott Marquis Hotel at Times Square New York City, USA, July 11-13, 2007 WeA20.1 Quadratic and Copositive Lyapunov Functions and the Stability of

More information

Dynamic Coalitional TU Games: Distributed Bargaining among Players Neighbors

Dynamic Coalitional TU Games: Distributed Bargaining among Players Neighbors Dynamic Coalitional TU Games: Distributed Bargaining among Players Neighbors Dario Bauso and Angelia Nedić January 20, 2011 Abstract We consider a sequence of transferable utility (TU) games where, at

More information

1 The Observability Canonical Form

1 The Observability Canonical Form NONLINEAR OBSERVERS AND SEPARATION PRINCIPLE 1 The Observability Canonical Form In this Chapter we discuss the design of observers for nonlinear systems modelled by equations of the form ẋ = f(x, u) (1)

More information

Event-Triggered Decentralized Dynamic Output Feedback Control for LTI Systems

Event-Triggered Decentralized Dynamic Output Feedback Control for LTI Systems Event-Triggered Decentralized Dynamic Output Feedback Control for LTI Systems Pavankumar Tallapragada Nikhil Chopra Department of Mechanical Engineering, University of Maryland, College Park, 2742 MD,

More information

Stability of Dynamical Systems on a Graph

Stability of Dynamical Systems on a Graph Stability of Dynamical Systems on a Graph Mohammad Pirani, Thilan Costa and Shreyas Sundaram Abstract We study the stability of largescale discretetime dynamical systems that are composed of interconnected

More information

Modeling and Stability Analysis of a Communication Network System

Modeling and Stability Analysis of a Communication Network System Modeling and Stability Analysis of a Communication Network System Zvi Retchkiman Königsberg Instituto Politecnico Nacional e-mail: mzvi@cic.ipn.mx Abstract In this work, the modeling and stability problem

More information

Nonlinear Control. Nonlinear Control Lecture # 3 Stability of Equilibrium Points

Nonlinear Control. Nonlinear Control Lecture # 3 Stability of Equilibrium Points Nonlinear Control Lecture # 3 Stability of Equilibrium Points The Invariance Principle Definitions Let x(t) be a solution of ẋ = f(x) A point p is a positive limit point of x(t) if there is a sequence

More information

Newtonian Mechanics. Chapter Classical space-time

Newtonian Mechanics. Chapter Classical space-time Chapter 1 Newtonian Mechanics In these notes classical mechanics will be viewed as a mathematical model for the description of physical systems consisting of a certain (generally finite) number of particles

More information

L 2 -induced Gains of Switched Systems and Classes of Switching Signals

L 2 -induced Gains of Switched Systems and Classes of Switching Signals L 2 -induced Gains of Switched Systems and Classes of Switching Signals Kenji Hirata and João P. Hespanha Abstract This paper addresses the L 2-induced gain analysis for switched linear systems. We exploit

More information

APPROXIMATE SIMULATION RELATIONS FOR HYBRID SYSTEMS 1. Antoine Girard A. Agung Julius George J. Pappas

APPROXIMATE SIMULATION RELATIONS FOR HYBRID SYSTEMS 1. Antoine Girard A. Agung Julius George J. Pappas APPROXIMATE SIMULATION RELATIONS FOR HYBRID SYSTEMS 1 Antoine Girard A. Agung Julius George J. Pappas Department of Electrical and Systems Engineering University of Pennsylvania Philadelphia, PA 1914 {agirard,agung,pappasg}@seas.upenn.edu

More information

Theory in Model Predictive Control :" Constraint Satisfaction and Stability!

Theory in Model Predictive Control : Constraint Satisfaction and Stability! Theory in Model Predictive Control :" Constraint Satisfaction and Stability Colin Jones, Melanie Zeilinger Automatic Control Laboratory, EPFL Example: Cessna Citation Aircraft Linearized continuous-time

More information

Research Article Convex Polyhedron Method to Stability of Continuous Systems with Two Additive Time-Varying Delay Components

Research Article Convex Polyhedron Method to Stability of Continuous Systems with Two Additive Time-Varying Delay Components Applied Mathematics Volume 202, Article ID 689820, 3 pages doi:0.55/202/689820 Research Article Convex Polyhedron Method to Stability of Continuous Systems with Two Additive Time-Varying Delay Components

More information

CDS Solutions to the Midterm Exam

CDS Solutions to the Midterm Exam CDS 22 - Solutions to the Midterm Exam Instructor: Danielle C. Tarraf November 6, 27 Problem (a) Recall that the H norm of a transfer function is time-delay invariant. Hence: ( ) Ĝ(s) = s + a = sup /2

More information

Floor Control (kn) Time (sec) Floor 5. Displacement (mm) Time (sec) Floor 5.

Floor Control (kn) Time (sec) Floor 5. Displacement (mm) Time (sec) Floor 5. DECENTRALIZED ROBUST H CONTROL OF MECHANICAL STRUCTURES. Introduction L. Bakule and J. Böhm Institute of Information Theory and Automation Academy of Sciences of the Czech Republic The results contributed

More information

Discrete Applied Mathematics

Discrete Applied Mathematics Discrete Applied Mathematics 194 (015) 37 59 Contents lists available at ScienceDirect Discrete Applied Mathematics journal homepage: wwwelseviercom/locate/dam Loopy, Hankel, and combinatorially skew-hankel

More information

On the Scalability in Cooperative Control. Zhongkui Li. Peking University

On the Scalability in Cooperative Control. Zhongkui Li. Peking University On the Scalability in Cooperative Control Zhongkui Li Email: zhongkli@pku.edu.cn Peking University June 25, 2016 Zhongkui Li (PKU) Scalability June 25, 2016 1 / 28 Background Cooperative control is to

More information

Putzer s Algorithm. Norman Lebovitz. September 8, 2016

Putzer s Algorithm. Norman Lebovitz. September 8, 2016 Putzer s Algorithm Norman Lebovitz September 8, 2016 1 Putzer s algorithm The differential equation dx = Ax, (1) dt where A is an n n matrix of constants, possesses the fundamental matrix solution exp(at),

More information

Robust distributed linear programming

Robust distributed linear programming Robust distributed linear programming Dean Richert Jorge Cortés Abstract This paper presents a robust, distributed algorithm to solve general linear programs. The algorithm design builds on the characterization

More information

Global Analysis of Piecewise Linear Systems Using Impact Maps and Quadratic Surface Lyapunov Functions

Global Analysis of Piecewise Linear Systems Using Impact Maps and Quadratic Surface Lyapunov Functions Global Analysis of Piecewise Linear Systems Using Impact Maps and Quadratic Surface Lyapunov Functions Jorge M. Gonçalves, Alexandre Megretski, Munther A. Dahleh Department of EECS, Room 35-41 MIT, Cambridge,

More information

Review of Controllability Results of Dynamical System

Review of Controllability Results of Dynamical System IOSR Journal of Mathematics (IOSR-JM) e-issn: 2278-5728, p-issn: 2319-765X. Volume 13, Issue 4 Ver. II (Jul. Aug. 2017), PP 01-05 www.iosrjournals.org Review of Controllability Results of Dynamical System

More information

x 2 F 1 = 0 K 2 v 2 E 1 E 2 F 2 = 0 v 1 K 1 x 1

x 2 F 1 = 0 K 2 v 2 E 1 E 2 F 2 = 0 v 1 K 1 x 1 ROCKY MOUNTAIN JOURNAL OF MATHEMATICS Volume 20, Number 4, Fall 1990 ON THE STABILITY OF ONE-PREDATOR TWO-PREY SYSTEMS M. FARKAS 1. Introduction. The MacArthur-Rosenzweig graphical criterion" of stability

More information

Dynamical Systems & Lyapunov Stability

Dynamical Systems & Lyapunov Stability Dynamical Systems & Lyapunov Stability Harry G. Kwatny Department of Mechanical Engineering & Mechanics Drexel University Outline Ordinary Differential Equations Existence & uniqueness Continuous dependence

More information

Network Flows that Solve Linear Equations

Network Flows that Solve Linear Equations Network Flows that Solve Linear Equations Guodong Shi, Brian D. O. Anderson and Uwe Helmke Abstract We study distributed network flows as solvers in continuous time for the linear algebraic equation arxiv:1510.05176v3

More information

CDS 101/110a: Lecture 2.1 Dynamic Behavior

CDS 101/110a: Lecture 2.1 Dynamic Behavior CDS 11/11a: Lecture.1 Dynamic Behavior Richard M. Murray 6 October 8 Goals: Learn to use phase portraits to visualize behavior of dynamical systems Understand different types of stability for an equilibrium

More information

ENERGY DECAY ESTIMATES FOR LIENARD S EQUATION WITH QUADRATIC VISCOUS FEEDBACK

ENERGY DECAY ESTIMATES FOR LIENARD S EQUATION WITH QUADRATIC VISCOUS FEEDBACK Electronic Journal of Differential Equations, Vol. 00(00, No. 70, pp. 1 1. ISSN: 107-6691. URL: http://ejde.math.swt.edu or http://ejde.math.unt.edu ftp ejde.math.swt.edu (login: ftp ENERGY DECAY ESTIMATES

More information

EE C128 / ME C134 Feedback Control Systems

EE C128 / ME C134 Feedback Control Systems EE C128 / ME C134 Feedback Control Systems Lecture Additional Material Introduction to Model Predictive Control Maximilian Balandat Department of Electrical Engineering & Computer Science University of

More information

Stochastic Processes

Stochastic Processes qmc082.tex. Version of 30 September 2010. Lecture Notes on Quantum Mechanics No. 8 R. B. Griffiths References: Stochastic Processes CQT = R. B. Griffiths, Consistent Quantum Theory (Cambridge, 2002) DeGroot

More information

Stability and Disturbance Propagation in Autonomous Vehicle Formations : A Graph Laplacian Approach

Stability and Disturbance Propagation in Autonomous Vehicle Formations : A Graph Laplacian Approach Stability and Disturbance Propagation in Autonomous Vehicle Formations : A Graph Laplacian Approach Francesco Borrelli*, Kingsley Fregene, Datta Godbole, Gary Balas* *Department of Aerospace Engineering

More information

8.1 Bifurcations of Equilibria

8.1 Bifurcations of Equilibria 1 81 Bifurcations of Equilibria Bifurcation theory studies qualitative changes in solutions as a parameter varies In general one could study the bifurcation theory of ODEs PDEs integro-differential equations

More information

Minimum-Phase Property of Nonlinear Systems in Terms of a Dissipation Inequality

Minimum-Phase Property of Nonlinear Systems in Terms of a Dissipation Inequality Minimum-Phase Property of Nonlinear Systems in Terms of a Dissipation Inequality Christian Ebenbauer Institute for Systems Theory in Engineering, University of Stuttgart, 70550 Stuttgart, Germany ce@ist.uni-stuttgart.de

More information

Some Background Material

Some Background Material Chapter 1 Some Background Material In the first chapter, we present a quick review of elementary - but important - material as a way of dipping our toes in the water. This chapter also introduces important

More information

Positive Markov Jump Linear Systems (PMJLS) with applications

Positive Markov Jump Linear Systems (PMJLS) with applications Positive Markov Jump Linear Systems (PMJLS) with applications P. Bolzern, P. Colaneri DEIB, Politecnico di Milano - Italy December 12, 2015 Summary Positive Markov Jump Linear Systems Mean stability Input-output

More information

STRUCTURED SPATIAL DISCRETIZATION OF DYNAMICAL SYSTEMS

STRUCTURED SPATIAL DISCRETIZATION OF DYNAMICAL SYSTEMS ECCOMAS Congress 2016 VII European Congress on Computational Methods in Applied Sciences and Engineering M. Papadrakakis, V. Papadopoulos, G. Stefanou, V. Plevris (eds. Crete Island, Greece, 5 10 June

More information

TOPOLOGICAL EQUIVALENCE OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS

TOPOLOGICAL EQUIVALENCE OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS TOPOLOGICAL EQUIVALENCE OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS ALEX HUMMELS Abstract. This paper proves a theorem that gives conditions for the topological equivalence of linear ordinary differential

More information

On Backward Product of Stochastic Matrices

On Backward Product of Stochastic Matrices On Backward Product of Stochastic Matrices Behrouz Touri and Angelia Nedić 1 Abstract We study the ergodicity of backward product of stochastic and doubly stochastic matrices by introducing the concept

More information