Hidden attractors in dynamical systems: systems with no-equilibria equilibria, coexistence of attractors and multistability Fundamental problems (16t Hilbert problem, Aizerman and Kalman conjectures) and engineering models (PLL, aircraft control systems, drilling systems, Chua circuits)
Nikolay Kuznetsov , Gennady Leonov, http://www.math.spbu.ru/user/nk/
tutorial: http://www.math.spbu.ru/user/nk/PDF/Hidden-attractor-coexistence-multistability.pdf
N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 1/26 Content
I Classical computation of self-excited attractors
I Hidden attractors: hidden periodic oscillations and hidden chaotic attractors, e.g., in the systems with no-equilibria or with the only stable equilibrium ( multistability and coexistence of attractors )
I Hidden attractors in fundamental problems and applied models
I 16th Hilbert problem on nested limit cycles in the plane
I Aizerman conjecture and Kalman conjecture on absolute stability of nonlinear control systems
I Phase-locked loop (PLL) and Costas loop circuits
I Aircrafts control systems (windup and antiwindup)
I Drilling systems, electrical machines and induction motors
I Hidden chaotic attractor in Chua circuit
I Secure (chaotic) communications: synchronization of coupled chaotic oscillators Survey: Leonov G.A.,Kuznetsov N.V., Hidden attractors in dynamical systems. From hidden oscillations in Hilbert-Kolmogorov, Aizerman, and Kalman problems to hidden chaotic attractor in Chua circuits, Int. J. Bif. and Chaos 23(1) 2013, art. no. 1330002, doi: 10.1142/S0218127413300024 N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 2/26 Engineering analysis of complex dynamical system x x˙ 1 f1(x1, x2, .., xn) 1 stable unstable x˙ 2 f2(x1, x2, .., xn) x˙ = = =F(x) ... ... x˙ f (x , x , .., x ) x n n 1 2 n 2 x 3 standard computational procedure: 1. Find equilibria xk : F(xk) = 0; 2. Computation of trajectories with local stability analysis initial data from neighborhoods of ˙ dF(x) unstable equilibria. y = dx x=xk y
3 Van der Pol Lorenz 2 45
40 . 1 . 35
x = y x = −s(x − y) 30 0 . 2 . 25 y = −x + e(1−x )y−1 y = rx − y − xz 20 . 15 −2 10 30 20 5 z = −bz + xy 10 −20 −15 0 −10 −5 −3 0 −10 5 10 −20 −4 −2 0 2 4 15 20 −30
only for localization of oscillations excited from unstable equilibria
N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 3/26 Computation of oscillations and attractors
Attractors classication: An attractor is called hidden attractor if its basin of attraction does not intersect with small neighborhoods of equilibria, otherwise it is called self-excited attractor. e.g. hidden attractors in the systems with no-equilibria or with the only stable equilibrium (special case of multistability& coexistence of attractors) standard computational procedure:[1. to nd equilibria; 15
2. trajectory, starting from a point of unstable manifold in a 10
5 neighborhood of unstable equilibrium, after transient process z 0 reaches an self-excited oscillation and localizes it] −5 5 −10 0 y allows to nd self-excited but not hidden oscillations. −15 −15 −10 −5 −5 x0 5 10 15 How to predict existence and choose initial data in attraction domain?
Survey: Leonov G.A.,Kuznetsov N.V., Hidden attractors in dynamical systems. From hidden oscillations in Hilbert-Kolmogorov, Aizerman, and Kalman problems to hidden chaotic attractor in Chua circuits, Int. J. Bif. and Chaos, 23(1), 2013, art. no. 1330002, doi: 10.1142/S0218127413300024
N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 4/26 Poincare visualization problem Poincare visualization problem, 1881: the problem ... to construct the curves dened by dierential equations 16th Hilbert problem (second part) 1900: number and mutual disposition of limit cycles for 2 2 x˙ = Pn(x, y) = a1x + b1xy + c1y + α1x + β1y + ... 2 2 y˙ = Qn(x, y) = a2x + b2xy + c2y + α2x + β2y + ... V. Arnold (2005):To estimate the number of LCs of square vector elds on plane, A.N. Kolmogorov had distributed several hundreds of such elds (with randomly chosen coecients of quadratic expressions) among a few hundreds of students of Mech.&Math. Faculty of Moscow Univ. as a mathematical practice. Each student had to nd the number of LCs of his/her eld. The result of this experiment was absolutely unexpected: not a single eld had a LC!... Visualization problem: nested limit cycles are hidden oscillations
N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 5/26 Hidden oscillations in R2: nested limit cycles in QS x˙ = x2 + xy + y, y˙ = ax2 + bxy + cy2 + αx + βy c∈(1/3,1), α=−ε−1, bc<1, b>a + c, 2c(b−1)2, β =0 Theorem. For suciently small ε the system has three limit cycles: one to the left of line{x = −1} and two to the right of it. (Increase β and get four normal size limit cycles)
Kuznetsov N.V., Kuznetsova O.A., Leonov G.A., Visualization of four normal size limit cycles in two-dimensional polynomial quadratic system, Dierential equations and Dynamical systems, 21(1-2), 2013, pp. 29-34 (doi:10.1007/s12591-012-0118-6)
N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 6/26 Hidden oscillations in R3: Aizerman and Kalman conjectures ∗ if z˙ =Az+bkc z, is asympt. stable ∀ k ∈(k1, k2): z(t, z0)→0 ∀z0, then ∗ is x˙ =Ax+bϕ(σ), σ =c x, ϕ(0)=0, k1 <ϕ(σ)/σ k ϕ(σ) 0 1949 : k1 < ϕ(σ)/σ < k2 1957 : k1 < ϕ (σ) < k2 In general, conjectures are not true (Aizerman's in R2 , Kalman's in R4) Periodic solution can coexist with the only stable equilibria for nonlinearity from the sector of linear stability hidden oscillation Aizerman'sconj.: N.Krasovsky 1952: R2, x(t)→∞; V.Pliss 1958: R3, periodic x(t) Why such conjectures: by describing function method (DFM) there are no periodic solutions in the case of Aizerman's and Kalman's conditions N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 7/26 Kalman conjecture (Kalman problem) 1957 ∗ 0 x˙ =Ax+bϕ(σ), σ =c x, ϕ(0)=0, 0<ϕ (σ) I Bernat J. & Llibre J. 1996: analytical-numerical `counterexamples' 4 0 construction in R , ϕ(σ) `close' to sat(σ) (0≤ϕ (σ)) I Leonov G., Kuznetsov N.: some of Fitts counterexamples are true; smooth counterex. in 4: 0 ; R ϕ(σ) = tanh(σ): 0 N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 8/26 Synthesis of scenario of transition to deterministic and chaotic hidden oscillations: an example ˙ (1) x˙ 1 =Px1 +ϕ1(x1)(2) x˙ ε =Pxε +εϕε(xε) (3) xεj =Pxεj +εjϕεj (xεj) small ε allows one to determine analytically a stable nontrivial periodic solution xε(t) for system (2) oscillating attractor A0. 1. if all points of A0 are in the attraction domain of A1 (oscillating attractor of (3) with ), then solution can be determined numerically by starting a j = 1 xε1 (t) trajectory of (3) with j =1 from initial point xε(0). If in computational process is not fallen to equilibria and is not (on su. large ), then xε1 (t) → ∞ [0,T ] xε1 (t) computes attractor A1. Then perform similar procedure for (3) with j =2: by starting trajectory of (3) with from init. point (last point on xε2 (t) j = 2 xε1 (T ) previous step) A2 can be computed. And so on. Localization of attractor A in (1):numerically follow transformation of Aj with increasing j=0, ..,m (εm =1, Am=A) [special continuation method] 2. in the change from system (2) to (3) with j =1, it's observed loss of stability bifurcation and vanishing of attractor A0(or Aj−1 on j-th step). N.V.Kuznetsov, G.A.Leonov Hidden attractors: systems without equilibria, attractors coexistence & multistability 9/26 Synthesis of scenario of transition to hidden oscillations: counterexample to Aizerman conjecture ε = 0.3 30 0.4 0.3 x˙ =−x −10ϕ(σ), x˙ =x −10.1ϕ(σ) 20 1 2 2 1 0.2 10 0.1 x˙ 3 =x4, x˙ 4 =−x3 −x4 +ϕ(σ) ) 2 x 0 σ 0 ( 3