[Paper Review] Necessary and Sufficient Conditions for Stable Synchronisation in Random Dynamical Systems

Julian Newman|arXiv (Cornell University)|Aug 24, 2014

Nonlinear Dynamics and Pattern Formation7 references3 citations

TL;DR

This paper establishes necessary and sufficient conditions for stable synchronization in random dynamical systems on compact spaces: synchronization occurs and is stable if and only if (1) there exists a smallest deterministic invariant set $K$, (2) any two points in $K$ can be brought closer together, and (3) $K$ supports asymptotically stable trajectories. The result generalizes Baxendale’s 1991 work by replacing complex vector field conditions with intrinsic geometric and dynamical properties of the system.

ABSTRACT

For a product of i.i.d. random maps or a memoryless stochastic flow on a compact space $X$, we find conditions under which the presence of locally asymptotically stable trajectories (e.g. as given by negative Lyapunov exponents) implies almost-sure mutual convergence of any given pair of trajectories ("synchronisation"). Namely, we find that synchronisation occurs and is stable if and only if the system exhibits the following properties: (i) there is a smallest deterministic invariant set $K \subset X$, (ii) any two points in $K$ are capable of being moved closer together, and (iii) $K$ admits asymptotically stable trajectories. Our first condition (for which unique ergodicity of the one-point transition probabilities is sufficient) replaces the intricate vector field conditions assumed in Baxendale's similar result of 1991, where (working on a compact manifold) sufficient conditions are given for synchronisation to occur in a SDE with negative Lyapunov exponents.

Motivation & Objective

• To identify minimal, intrinsic conditions under which locally asymptotically stable trajectories in random dynamical systems lead to almost-sure mutual convergence (synchronization).
• To replace the intricate vector field conditions in Baxendale’s 1991 result with more general, geometrically meaningful criteria applicable to memoryless stochastic flows.
• To establish a characterization of stable synchronization that is robust under perturbations of trajectories, not just parameter variations.
• To unify and generalize existing results on noise-induced synchronization by focusing on the structure of the minimal invariant set and local contraction properties.

Proposed method

• The analysis is conducted within the framework of filtered random dynamical systems (RDS) with memoryless noise, using time-homogeneous, stationary noise processes.
• The authors define and analyze key concepts: locally asymptotically stable trajectories, recurrent Lyapunov stability, and naïve attractiveness of pairs $(\omega,x)$.
• They introduce and apply the notion of $(\varepsilon,\delta)$-containment and use the Poincaré recurrence theorem to show that certain sets are null under the product measure $\mathbb{P} \otimes \rho$.
• The proof relies on technical lemmas concerning local stability in non-autonomous and random systems, particularly Lemma A3 and the recurrence properties of invariant sets.
• A critical step involves showing that the set of points where trajectories fail to converge is null, using the structure of the minimal invariant set $K$ and its contraction properties.
• The argument culminates in proving that asymptotic stability is equivalent to recurrent Lyapunov stability and naïve attractiveness, establishing the full characterization.

Experimental results

Research questions

• RQ1 Under what minimal conditions does the presence of locally asymptotically stable trajectories in a random dynamical system imply almost-sure mutual convergence of all trajectories?
• RQ2 Can the complex vector field conditions in Baxendale’s 1991 result be replaced by more geometric and intrinsic dynamical properties?
• RQ3 Is the existence of a smallest deterministic invariant set $K$ sufficient to ensure stable synchronization when combined with local contraction and stability?
• RQ4 How does the interplay between invariance, contraction, and stability in $K$ guarantee that all trajectories synchronize almost surely?
• RQ5 What is the precise relationship between Lyapunov stability, attractivity, and asymptotic stability in the context of random dynamical systems?

Key findings

• Synchronization occurs and is stable if and only if there exists a smallest deterministic invariant set $K \subset X$, any two points in $K$ can be moved closer together, and $K$ admits asymptotically stable trajectories.
• The condition of unique ergodicity of one-point transition probabilities is sufficient to ensure the existence of the smallest invariant set $K$, replacing the need for complex vector field assumptions.
• The system exhibits stable synchronization even without global contraction, provided the minimal invariant set $K$ satisfies the three stated conditions.
• The equivalence between asymptotic stability and the conjunction of recurrent Lyapunov stability and naïve attractiveness is established for $\mathbb{P} \otimes \rho$-almost every $ (\omega,x) $, under mild regularity conditions.
• The result generalizes Baxendale’s 1991 theorem by replacing differential-geometric conditions with topological and measure-theoretic properties of the system’s invariant sets.
• The proof shows that the set of trajectories failing to synchronize is a $\mathbb{P} \otimes \rho$-null set, confirming almost-sure convergence under the stated conditions.