Combination Properties of Weakly Contracting Systems

Ian R. Manchester, Jean-Jacques SlotinearXiv 2014 · DOI · arXiv

Abstract

This note explores the property of weak contraction, which implies that all bounded solutions of a nonlinear system converge to a (possibly non-unique) equilibrium. We provide simple results regarding the interconnections of such systems, demonstrating how stability is preserved through parallel, feedback, and hierarchical structures.

Introduction

In nonlinear dynamical systems, contraction analysis is a robust framework used to prove that trajectories converge to a unique equilibrium or a common nominal trajectory. However, many real-world systems—including neural networks, biological circuits, and robotics—exhibit multi-stability, where multiple equilibria coexist.

Standard contraction is often too restrictive for these applications because it requires the system's Jacobian to be negative definite everywhere, effectively pulling every point toward a single center. This paper formalizes a "weak contraction" condition. Instead of requiring the distance between points to shrink (1D), weak contraction ensures that 2D areas in the state space shrink. This property is sufficient to guarantee that all bounded trajectories settle into an equilibrium, while allowing for the existence of multiple stable points.

Theoretical Framework

The authors integrate classical dimension theory with modern contraction analysis. The core technical requirement for weak contraction is that the sum of the top two eigenvalues of the system's Jacobian (or a generalized Jacobian under a Riemannian metric) is negative: $\lambda_1 + \lambda_2 < 0$.

Key elements of this framework include:

  • Eigenvalue Sums: The function $S_k(H)$, representing the sum of the $k$ largest eigenvalues of a symmetric matrix, is utilized. For weak contraction, the focus is on $k=2$.
  • Riemannian Metrics: By using coordinate transformations and metrics ($\Theta$), the theory applies to systems that may not appear contracting in standard Euclidean space.
  • Storage Functions: To extend the theory to non-autonomous (time-varying) systems, the authors introduce scalar storage functions $\gamma(t)$ to account for transient energy changes, reducing conservatism in stability criteria.

Compositional Stability and Interconnections

The primary contribution of this work is proving that weak contraction is a compositional property. This allows engineers to build complex, stable systems by interconnecting smaller, weakly contracting components. The paper establishes three primary interconnection theorems:

  1. Parallel Interconnection: A weighted sum of weakly contracting systems remains weakly contracting.
  2. Feedback Interconnection: A skew-symmetric feedback loop between a weakly contracting system and a strongly contracting system is guaranteed to be stable.
  3. Hierarchical (Triangular) Interconnection: If a weakly contracting system A drives system B (without feedback), and both are at least weakly contracting, the combined system remains weakly contracting.

This "snap-together" property is significant for modular design. It implies that as long as individual modules satisfy the weak contraction criteria, their combination will not result in limit cycles or chaotic attractors.

Geometric Intuition and Applications

The geometric interpretation of weak contraction is akin to a film of soapy water on a loop. As the system flow acts on the film, the area must shrink to zero, eventually forcing the loop to collapse into an equilibrium point. This prevents the formation of oscillations.

To demonstrate the theory, the authors analyze a simple damped pendulum. While a pendulum has multiple equilibria (at every $k\pi$), it satisfies the condition $\lambda_1 + \lambda_2 < 0$. This mathematically confirms why the pendulum must always settle into an equilibrium state rather than oscillating indefinitely.

From a computational standpoint, the eigenvalue sum function $S_k(H)$ is convex. This allows the conditions for weak contraction to be verified using Linear Matrix Inequalities (LMIs), which are efficiently solvable using standard convex optimization tools.

Limitations and Future Directions

While powerful, the current framework has specific constraints:

  • State Space Requirements: The guarantees rely on the system remaining within a compact, simply-connected, forward-invariant set.
  • Equilibrium Selection: The theory proves the system will reach an equilibrium but does not predict which specific equilibrium a multi-stable system will choose.
  • Metric Evolution: Future research may explore non-Riemannian metrics, such as $L_1$ or $L_\infty$ norms, which are often more appropriate for modeling chemical or biological reaction networks.

Summary updated

Abstract

A note on the property of weak contraction, which implies that all bounded solutions of a nonlinear system converge to a (possibly non-unique) equilibrium. We provide some simple results about interconnections of such systems, and a brief discussion.
BibTeX
@misc{manchester2014combination,
  author = {Ian R. Manchester and Jean-Jacques Slotine},
  title = {{Combination Properties of Weakly Contracting Systems}},
  year = {2014},
  eprint = {1408.5174},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1408.5174},
}
Figure