You are on page 1of 5

Contraction Analysis of Time-Delayed Communications and Group Cooperation

Wei Wang
Nonlinear Systems Laboratory, MIT Cambridge, Massachusetts, 02139 Email: wangwei@mit.edu

Jean-Jacques E. Slotine
Nonlinear Systems Laboratory, MIT Cambridge, Massachusetts, 02139 Email: jjs@mit.edu

Abstract— We study stability of interacting nonlinear systems with time-delayed communications, using contraction theory and a simplified wave variable design inspired by robotic teleoperation. We show that contraction is preserved through specific time-delayed feedback communications, and that this property is independent of the values of the delays. The approach is then applied to group cooperation with linear protocols, where it is shown that synchronization can be made robust to arbitrary delays. NSL-041001, October 2004, revised December, 2005 To be published in I.E.E.E. Trans. Automatic Control

I. I NTRODUCTION In many engineering applications, communications delays between subsystems cannot be neglected. Such an example is bilateral teleoperation, where signals can experience significant transmission delays between local and remote sites. Throughout the last decade, both internet and wireless technologies have vastly extended practical communication distances. Information exchange and cooperation can now occur in very widely distributed systems, making the effect of time delays even more central. In the context of telerobotics, [2] proposed a control law for force-reflecting teleoperators which preserves passivity, and thus overcomes the instability caused by time delays. The idea was reformulated in [18] in terms of scattering or “wave” variables [6], [3]. Transmission of wave variables across communication channels ensures stability without knowledge of the time delay. Further extensions to internetbased applications were developed [16], [17], [4], in which communication delays are variable. Recently, [14], [23] extended the application of wave variables to a more general context by performing a nonlinear contraction analysis [13], [14], [1], [7] of the effect of timedelayed communications between contracting systems. This paper modifies the design of the wave variables proposed in [14], [23]. Specifically, a simplified form provides an effective analysis tool for interacting nonlinear systems with time-delayed feedback communications. For appropriate coupling terms, contraction as a generalized stability property is preserved regardless of the delay values. This also sheds a new light on the well-known fact in bilateral teleoperation that even small time-delays in feedback Proportional-Derivative (PD)

controllers may create stability problems for simple coupled second-order systems, which in turn motivated approaches based on passivity and wave variables [2], [18]. The approach is then applied to derive the paper’s main results on the group cooperation problem with delayed communications. We show that synchronization with linear protocols [19], [15] is robust to time delays and network connectivity without requiring the delays to be known or equal in all links. In a leaderless network, all the coupled elements reach a common state which depends on the initial conditions and the time delays, while in a leader-followers network the group agreement point is fixed by the leader. The approach is suitable to study both continuous-time and discrete-time models. II. C ONTRACTION A NALYSIS OF T IME -D ELAYED C OMMUNICATIONS This introductory section shows that a simplified form of the transmitted wave variables in [14] can be applied to analyze time-delayed feedback communications. The results which follow could be obtained using a variety of alternative techniques. For instance, they could be easily derived based on input-output analysis [31], although in an L2 sense rather than exponentially. The use of modified wave variables provides a unifying framework which will be central in the derivation of the paper’s main results in section III. Consider two interacting systems of possibly different dimensions, ˙ 1 = f1 (x1 , t) + G21 τ 21 x ˙ 2 = f2 (x2 , t) + G12 τ 12 x (1)

where x1 ∈ Rn1 , x2 ∈ Rn2 , τ 12 , τ 21 ∈ Rn , and G12 ∈ Rn1 ×n , G21 ∈ Rn2 ×n are two constant matrices. Inputs τ ij are computed by transmitting between the two systems simplified “wave” variables, defined as u21 u12 = GT 21 x1 + k21 τ 21 T = G12 x2 + k12 τ 12 v12 = GT 21 x1 T v21 = G12 x2

where k12 and k21 are two strictly positive constants. Because of time delays, one has u12 (t) = v12 (t − T12 ) u21 (t) = v21 (t − T21 )

where T12 and T21 are two positive constants. Note that subscripts containing two numbers indicate the communication

a simple coordinate transformation yields x ˙1 y ˙1 x ˙2 y ˙2 = = ωy1 − bx1 −ωx1 ωy2 − bx2 −ωx2 + K( + K( x2 (t − T21 ) y2 (t − T21 ) x1 (t − T12 ) y1 (t − T12 ) − − x1 y1 x2 y2 ) ) k21 k12 1 δ xT δ xT V1. subscript “12” refers to communication from node 1 to 2. and V is bounded and tends to a limit. T21 = 4s. the kd 0 is neither symtransformed coupling gain K = kp 0 ω metric nor positive semi-definite for any kp = 0. [29] of spontaneous oscillation.g. t) + x 1 T k21 G21 (G12 x2 (t 1 T k12 G12 (G21 x1 (t ωy2 − bx2 ωy1 − bx1 are and f2 = −ωx2 −ωx1 both (semi-)contracting with identity metric [29]. However. Expanding system dynamics (1) yields ˙ V = k21 δ xT 1 ˙ 1 = f1 (x1 . are identical for the two plots. PD Control x 1 x 2 8 6 4 (b). Regardless of the values of the delays. This result does not contradict the well-known fact in teleoperation that even small time delays in bilateral PD controllers may create stability problems for coupled secondorder systems [2].. III. δ x2 . in which two or more identical biological cells. independent of the initial conditions. similarly to [14]. t) + x ˙ 2 = f2 (x2 . Initial conditions. If we assume further that x1 and x2 have the same dimension. [4]. Understanding natural aggregate motions as in bird flocks. ω 2 = 5.2 = t−T12 This yields ∂ f1 ∂ f2 δ x1 + k12 δ xT δ x2 2 ∂ x1 ∂ x2 2 2 k12 k21 δτ T δτ T − 21 δ τ 21 − 12 δ τ 12 2 2 If f1 and f2 are both contracting with identity metrics (i. and b > 0. [19]. then V ˙ tends lemma [22] in turn shows that if V to zero asymptotically. [20].direction. chosen randomly. Simulation results of two coupled mass-spring-damper systems with (a) PD control and (b) D control. it is straightforward to show that the results here apply recursively to feedback hierarchies of such systems. e. a synchronization condition was obtained for a group of coupled nonlinear systems. inert by themselves. a key condition for contraction to be preserved is that the coupling gains be symmetric positive semi-definite in the same metric as the subsystems. This notation will be helpful in Section III. [18]. [23]. the overall system is therefore contracting. central to contraction theory. Finally. Contraction cannot be preserved in this case. [17]. h2 = x ¨2 + bx ˙ 2 + ω 2 x2 . which implies that δ x1 . if ∂ f2 ∂ f1 ˙ ∂ x1 and ∂ x2 are both uniformly negative definite). t) + x 1 k21 1 k12 GGT ( x2 (t − T21 ) − x1 (t) ) GGT ( x1 (t − T12 ) − x2 (t) ) This implies that. where results will be extended to groups of interacting subsystems. fish schools. Consider. 1. In the sequel we shall ¨ can indeed be bounded as a consequence of assume that V the boundedness of V . T21 > 0. G ROUP C OOPERATION WITH T IME -D ELAYED C OMMUNICATIONS We now present the paper’s main results.2 1 δ x1 + 2 δ x2 + 2 2 2 V1. synchronization or group agreement has been the object of extensive literature [11]. all solutions of system (1) converge to a single trajectory. tend to self-excited oscillations through diffusion interactions. D Control x 1 x 2 40 20 2 0 0 −2 −20 −4 −6 −40 −8 −60 0 20 40 t(sec) 60 −10 0 20 40 t(sec) 60 − T21 ) − GT 21 x1 (t)) − T12 ) − GT 12 x2 (t)) Fig. ω > 0. which makes the origin a stable equilibrium point. kd = 1. where the number of the elements . In both cases. the differential length V = where t t T δ v12 δ v12 dǫ + t−T21 T δ v21 δ v21 dǫ exponentially regardless of initial conditions. once we set kp = 0.1 Consider two identical second-order systems coupled through time-delayed feedback PD controllers h1 = kd (x ˙ 2 (t − T21 ) − x ˙ 1 (t)) + kp (x2 (t − T21 ) − x1 (t)) h2 = kd (x ˙ 1 (t − T12 ) − x ˙ 2 (t)) + kp (x1 (t − T12 ) − x2 (t)) where h1 = x ¨1 + bx ˙ 1 + ω 2 x1 .5. Applying Barbalat’s ¨ is bounded. and choose G12 = G21 = G. kp = 5 in (a) and kp = 0 in (b). [28]. In fact. which makes the transformed coupling gains lose positive semi-definiteness. then V ≤ 0. Parameters are b = 0. If T12 = T21 = 0. or animal herds may help achieve desired collective behaviors in artificial multi-agent systems. If T12 . δ τ 12 and δ τ 21 all tend to zero. 2 where f1 = 60 (a).. and the coupled systems turn out to be unstable for large enough delays as the simulation result in Figure 1(a) illustrates. which motivates approaches based on passivity and wave variables. Note that the relative simplicity with which both phenomena can be interpreted makes fundamental use of the notion of a metric. Recently. [21]. [29]. t) + x ˙ 2 = f2 (x2 . [12]. [27].e. While in Figure 1(b). In our previous work [24]. T12 = 2s. Example 2. x1 and x2 converge together The instability mechanism in the above example is actually very similar to that of the classical Smale model [26]. for appropriate coupling terms. contraction as a generalized stability property will be preserved regardless of the time delays and the delay values. This result has a useful interpretation. the instability is caused by a non-identity metric. the whole system is actually equivalent to two diffusively coupled subsystems ˙ 1 = f1 (x1 .

we first assume that all the links are bidirectional with Kji = Kij . Equation (3) can be transformed to ˙i = x j ∈Ni δ xT i δ xi + i=1 1 2 t (j →i)∈N t−Tji T δ vji δ vji dǫ Gji τ ji where τ ji and correspondingly τ ij are defined through uji uij = GT ji xi + τ ji = GT ij xj n vij = GT ji xi vji = GT ij xj (4) and the rest of the proof is the same. and Kji is the coupling gain. δ x ˙ i does not necessarily imply that δ xi is convervanishing δ x gent. without requiring the delays to be known or equal in all links. According to Barbalat’s Thus we can say that V ˙ lemma. . the whole system will tend to reach a group agreement x1 (t) = · · · = xn (t) asymptotically if the network is connected. a Thus we know that ∀i. V (t) is bounded if initial states Since V are bounded. Therefore ˙ = −1 V 2 T (δ τ T ji δ τ ji + δ τ ij δ τ ij ) (i.can be arbitrary and the network structure can be very general. The case when both types of links are involved is similar. Define 1 V = 2 δ xT i δ xi i=1 1 + 2 Vi. The dynamics of the ith element is given as ˙i = x j ∈Ni where N = ∪n i=1 Ni denotes the set of all active links.e. δ uji (t) δ uij (t) T = GT ji δ xi (t) + δ τ ji (t) = Gij δ xj (t − Tji ) T = GT ij δ xj (t) + δ τ ij (t) = Gji δ xi (t − Tij ) Kji (xj − xi ) m (2) where i = 1. and each is chosen randomly around 0. in the state-space. . Simulation results are plotted in Figure 2. or unidirectional but formed in closed rings with identical gains.5 second. However it does in this case. we show that synchronization is robust to time delays both for the leaderless case and for the leader-followers case. j ) ∈ N . . and a two-way chain structure 1o /2o /3o /4o /5o /6 + τ ij with Gij = Gji > 0 and Kji = Kij = Gij GT ij . 2 Example 3. ∀(i. [29] based on partial contraction analysis. if ∀i. and Vi. Leaderless Group We first investigate a flocking model without group leader. and the coupling links are either bidirectional with Kji = Kij . any point inside the region x1 = · · · = xn is invariant to (3).j (i. In particular. The whole system will tend to reach a group agreement 1 x1 (t) = · · · = xn (t) = n (x1 (0) + · · · + xn (0)) exponentially if the network is connected.j is defined as in Section II for each link connecting two nodes i and j . Proof: For notational simplicity. This implies that δ τ ij (t) and δ x ˙ ij (t) since and so is δ τ ˙ ji (t) = GT ˙ j (t − Tji ) − GT ˙i δτ ij δ x ji δ x ¨ (t) is bounded. n = 6. δ τ ji and δ τ ij tend to zero asymptotically. and the coupling links are either bidirectional with Kji = Kij . Regardless of the explicit values of the delays. we notice that. the group agreement point in Theorem 2 generally does not equal the average value of the initial conditions. Theorem 1: Consider n coupled elements with linear protocol (2). because otherwise it would contradict the fact that δ xi tends to be periodic with constant period Tji + Tij . Thus. we study a simplified continuous-time model of schooling or flocking with time-delayed communications. any i. The dynamics of the ith element turns out to be ˙i = x j ∈Ni We can also conclude that. they will tend to a steady state δ x1 (t) = · · · = δ xn (t) = c where c is a constant vector whose value depends on the specific trajectories we analyze. One can easily show that δ xi (t) is bounded for ˙ i (t) are both bounded. Ni denotes the set of the active neighbors of element i. Tji = Tij . . Theorem 1 is derived in [24]. Assume now that time delays are non-negligible in communications. Consider the cooperative group (3) with one-dimensional xi . δ xi is convergent. ˙ i tends to zero. and generalize recent results in the literature [19]. In this section. but the time delays could be different along opposite directions. Now in general. V will then tend to zero asymptotically. all solutions of system (3) will tend to reach a group agreement x1 = · · · = xn asymptotically. and the result can be extended further to time-varying couplings (Kji = Kji (t)). which for instance can be defined as the set of the nearest neighbors within a certain distance around i. which implies that. switching networks (Ni = Ni (t)) and looser connectivity conditions. A. i.j )∈N ˙ (t) is non-positive. or unidirectional but formed in closed rings with identical gains. [15]. The delay values are different. Moreover. n and xi ∈ R .. Similar results are then derived for discrete-time models. we set V = 1 2 n Kji ( xj (t − Tji ) − xi (t) ) (3) Theorem 2: Consider n coupled elements (3) with timedelayed communications. 2 .j )∈N (5) The coupling gains are set to be identical with k = 5.1 Compared with Theorem 1. In the case that coupling links are unidirectional but form closed rings with identical coupling gains in each ring. By path integration this implies immediately that regardless of the delay values or the initial conditions. which is assumed to be symmetric and positive definite. but depends on the values of the time delays.

and γi = 0 or 1 represents the unidirectional links from the leader to the corresponding followers. Initial conditions. If the communication delays are nonnegligible. Applying Barbalat’s lemma shows that V tend to zero asymptotically. The delays are not equal. i. n = 6. 3. For each non-zero γi . we do not have to require Kij = Kji . although the agreement value is different. Recently. . each of which is chosen randomly around 0. Without Delays ∪n i=1 Ni ( xj (t − Tji ) − xi (t) ) 30 (b). group agreement to the leader value x0 is reached.e. Moreover. the whole system is asymptotically contracting. xi are the states of the followers. .1 without delays and with delays. 3. j ) ∈ N . Considering the same Lyapunov function V as (5). and structured by /2o /1o /3o /4o /5o /6 0 The state of the leader is constant with value x0 = 10. we can transform the equation (6) to ˙i = x j ∈Ni Gji τ ji + γi K0i (x0 − xi ) where τ ji and τ ij are defined the same as those in (4).e. i. All the coupling gains are set to be identical with k = 5. Regardless of the initial conditions. Moreover. where N = denotes the set of all active links among ˙ will the followers. which in this case is the point x1 (t) = · · · = xn (t) = x0 . but the result is limited by the assumptions that communication delays are equal in all links and that the selfresponse part in each coupling uses the same time delay. we get n t t −25 0 5 10 −25 0 5 10 15 ˙ = − V 20 γi δ xT i K0i δ xi − i=1 1 2 T (δ τ T ji δ τ ji + δ τ ij δ τ ij ) (i. chosen randomly. . as one of the main properties of contraction [13]. n. or unidirectional but formed in closed rings with identical gains. the dynamics of the ith element is given as ˙i = x j ∈Ni Kji (xj (t−Tji )−xi (t))+γi K0i (x0 (t−T0i )−xi (t)) the whole system is still asymptotically contracting according to exactly the same proof. are the same for each simulation. Proof: Exponential convergence of the leader-followers network (6) without delays has been shown in [24]. Simulation results for Example 3. Simulation results are plotted in Fig. The discretetime model studied in Section III-C is in this spirit.2 Consider a leader-followers network (6) with one-dimensional xi . Initial conditions. all solutions converge to a particular one. are the same for each simulation. [15] independently analyzed system (3) in the scalar case with the assumption that delays are equal in all links. if γi = 1. 2.2 without delays and with delays. With Delays x i xi 25 20 15 10 5 0 −5 −10 B. δ xi will tend to zero.. . chosen randomly. Leader-Followers Group Similar analysis can be applied to study coupled networks with group leaders.5 second. as well as δ τ ji and δ τ ij ∀(i. 2 Example 3. the virtual dynamics will converge to δ x1 (t) = · · · = δ xn (t) = 0 regardless of the initial conditions or the delay values. [29] using contraction theory. the dynamics of the ith element could be ˙ i = Ki x j ∈Ni 1 where Ki = k GGT and G is unique through the whole i network. Consider such a model ˙i = x j ∈Ni 25 20 15 10 5 Kji (xj (t − Tji ) − xi (t)) + γi K0i (x0 − xi ) (6) 0 −5 −10 −15 0 where i = 1. since δ τ ji (t) = GT ji ( δ xj (t − Tji ) − δ xi (t) ) we conclude that if the whole leader-followers network is connected. . and the coupling links among the followers are either bidirectional with Kji = Kij . Regardless of the explicit values of the delays. Ni indicate the neighborship among the followers. It implies that ∀i. All solutions will converge to a particular one. Note that even if x0 is not a constant. Instead. With Delays xi xi are bidirectional. the whole system will tend to reach a group agreement x1 (t) = · · · = xn (t) = x0 asymptotically if the whole network is connected. A similar condition was derived in [5] for a delayless swarm model. In both cases. Without Delays 25 20 15 10 5 0 −5 −10 −15 −20 (b). 2 30 (a). The proof is the same except that we incorporate ki into the wave variables and the function V . Such a design brings more flexibility to cooperation-law design. Note that the conditions on coupling gains can be relaxed. Theorem 3: Consider a leader-followers network (6) with time-delayed communications. if x0 is periodic.j )∈N Fig. all the followers’ states xi will tend to be periodic with the same period as x0 . Model (2) with delayed communications was also studied in [19]. x0 is the state of the leader. which we first assume to be a constant.25 20 15 10 5 0 −5 −10 −15 −20 (a). and assuming that all the links among the followers t 5 10 15 20 t −15 0 10 20 30 40 50 Fig. If the links are bidirectional.. Group agreement is reached in both cases. the coupling gain K0i is positive definite. Simulation results for Example 3. The proof is similar for unidirectional links in closed rings. which in this case depends on the dynamics of x0 and the explicit values of the delays.

92(1). and Murray.. (1995) Novel Type of Phase Transition in a System of Self-Driven Particles. CO. Future work includes coupled networks with switching topologies and time-varying time-delays. in Beyond Webcams: An Introduction to Internet Telerobotics. both of which can be applied to study group cooperation problem with asynchronous updating instants [8]. T. 354-367. 16(1). 14(2). G.. IV. . R. Cambridge-Amsterdam. M. Spong. 34(6).J. and the Binding Problem: A View from Stability Theory. IEEE Transactions on Automatic Control. W.. Wang.A. J. and Slotine. [5] Chu. (2001) Observers for Euler-Bernoulli Beam with Hydraulic Drive. C. (2000) Control System Design for Mechanical Systems Using Contraction Theory. [31] Zames. and Mu. J. E. Assume now that time delays are non-negligible in communications. Cambridge.E.E. 11(3):465-476. J. J. L. (2001) Algebraic Graph Theory.. ni is a positive integer ∀i. San Francisco. [30] Wang. Springer. J. Belgium.J.pdf. Hirche. [3] Belevitch. [28]: xi (t + 1) = xi (t) + 1 1 + ni ( xj (t) − xi (t) ) j ∈Ni where i = 1. (2003) Modular Stability Tools for Distributed Computation and Control. J. L. N. MA. Belgium. and Nguyen. J.J. and the value of the neighborhood radius r is unique throughout the whole network. N. [27] Tanner. (2005) Information Consensus of Asynchronous Discrete-time Multi-agent Systems.. Cambridge. (2003) Coordination of Groups of Mobile Autonomous Agents Using Nearest Neighbor Rules.. [13] Lohmiller. IEEE J. International Journal of Adaptive Control and Signal Processing. (2001).W. and Johnson.J. Wang. Greece.e.W.E. Proceedings of the American Control Conference. Proof: See [30]. S. (2003) An Intrinsic Observer for a Class of Lagrangian Systems. A. Cambridge University Press. [6] Desoer. Consider the model of flocking or schooling studied in [11]. We assume: all elements update their states synchronously. Kristiansen. Modularity. [2] Anderson. [7] Egeland.E. of Oceanic Engineering. IEEE Mediterranian Conference on Control and Automation. W.E..M. Control.E. n. and Schools: a Distributed Behavioral Model. and Slotine. J. and El-Rifai. on Robotics and Automation. the time interval between any two updating steps is a constant. Holden-Day. Leuven. J. Portland. Prentice Hall. G. 17(6). [16] Niemeyer. C. with xi (t + 1) = xi (t) + j ∈Ni τji (t) and wave variables uji = xi + (1 + ni ) τji .R.. Springer-Verlag.J. which leads to the fact that all interactions are bidirectional. Control.E. submitted.C. Trans.. Neural Networks. V. and Slotine. [29] Wang. (2003) A Study of Synchronization and Group Cooperation Using Partial Contraction Theory.J. (1985) Matrix Analysis. Stable Adaptive Teleoperation. [22] Slotine. E. J. MA.. 34(5).. Control. J. (1966) On the Input-Output Stability of Time-Varying Nonlinear Feedback Systems. Kumar V. 48(6).. Artificial Potentials and Coordinated Control of Groups. M. G.S. [28] Vicsek. C. Aut. E. IEEE Trans.E..E. Electrical Science.. .E. (1998) On Contraction Analysis for Nonlinear Systems. and Pappas. Biological Cybernetics. T. 48:988-1001. NSL Report.. Sixteenth International Symposium on Mathematical Theory of Networks and Systems. The MIT Press. and Buss. eds. (1975) Feedback Systems: Input-Output Properties. IEEE Transactions On Automatic Control on Networked Control Systems. (1991) Applied Nonlinear Control. IEEE Conf.. vij = xi . N. [26] Smale. O. Calif. Dec.. [10] Horn. Jadbabaie. Computer Graphics. Contr. http://www.. [23] Slotine. and Wang. Proceedings of the 2005 American Control Conference. and Vidyasagar. and Lohmiller. Rev. (2004) On Partial Contraction Analysis for Coupled Nonlinear Oscillators.. and Slotine. (1989) Bilateral Control of Teleoperators. and Slotine... C ONCLUSION This paper introduces modified wave variables in the context of contraction analysis. (2003) Coordination of Multiple Autonomous Vehicles. Regardless of the explicit values of the delays. W. Lett. IEEE Trans.J. W. (1998) Towards Force-Reflecting Teleoperation Over the Internet. and shows that they yield effective and simple tools for analyzing interacting systems and group synchronization with time-delayed feedback communications.mit. i. G. Rouchon. As in previous sections. and Fiorelli. H. [20] Reynolds. [21] Slotine.. the whole system will tend to reach a group agreement x1 (t) = · · · = xn (t) asymptotically. The 16th International Symposium on Mathematical Theory of Networks and Systems. (1968) Classical network theory. W. S. MIT. (2003) Bilateral Teleoperation over the Internet: the Time Varying Delay Problem. S. Tji could be different for different communication links. Oregon [9] Godsil. 21:25-34.. J. W. Goldberg and R. and Morse. [19] Olfati-Saber. I. W. IEEE Transactions on Automatic Control. [8] Fang. P. . G.J. Spinger-Verlag. Czir´ ok. Ben-Jacob.E.. R. Block Island Workshop on Cooperative Control..E. [14] Lohmiller. and Antsaklis P. As proved in [11].E. and Cohen I. and Spong. (2004) Contraction Analysis of TimeDelayed Communications using Simplified Wave Variables. The update law of the ith element changes to xi (t +1) = xi (t) + 1 1 + ni ( xj (t − Tji ) − xi (t) ) (7) j ∈Ni where the delay value Tji is an integer based on the number of updating steps. Lin. [24] Slotine... [12] Leonard..J.. Siegwart. A. [15] Moreau. 40th IEEE Conference on Decision and Control. K.. (2004) Contraction Analysis of Synchronization in Networks of Nonlinearly Coupled Oscillators. L (2004) Stability of Continuous-Time Distributed Consensus Algorithms. 75:1226-1229.E.A. Birds. (2001) Towards Bilateral Internet Teleoperation.edu/∼nsl/timedelay. [25] Slotine.494-501. Belgium. Academic Press. Automatica. and Royle.. [4] Chopra. Denver.. (2004) Collective Behavior Analysis of an Anisotropic Swarm Model. [17] Niemeyer.J. (1987) Flocks.J. the network structure is connected and fixed. M. A similar analysis leads to the same result as Theorem 3 for a discrete-time model with a leader-followers structure. London. A. Evolution. New York.. Conf.J.E. R EFERENCES [1] Aghannan. and Slotine.J. G. 45(5).. R.E. J..E. Phys. [18] Niemeyer. or even different along opposite directions on the same link. Aut. Discrete-Time Models Simplified wave variables can also be applied to study timedelayed communications in discrete-time models.D. W. and Slotine.. and ni equals the number of the neighbors of element i. K. Rhodes. I. Aut. . (1991). (2001) Virtual Leaders.J..E... and Li. C. A. (1976) A Mathematical Model of Two Cells via Turing’s Equation. T. (2004) Consensus Problems in Networks of Agents with Switching Topology and Time-Delays. .E. the whole system will tend to reach a group agreement if the network is connected in a very loose sense. Theorem 4: Consider n coupled elements (7) with timedelayed communications. in The Hopf Bifurcation and Its Applications. Editor. [11] Jadbabaie. R.. M.