The effect of randomness for dependency map on the robustness of interdependent lattices

Jing Yuan, Lixiang Li, Haipeng Peng, Jürgen Kurths, Jinghua Xiao, Yixian Yang

TL;DR

This paper investigates how the randomness of connections between interdependent networks affects their overall robustness. They find that a moderate level of randomness in these connections can lead to a more stable system, preventing catastrophic cascading failures that occur in highly random or perfectly ordered systems. This has implications for designing more resilient infrastructure networks.

What Problem Does This Paper Address?

Real-world complex systems, like power grids or communication networks, are rarely isolated. Instead, they are often interdependent, meaning the failure of one network can trigger failures in others. This interdependence can make systems extremely vulnerable, as a small initial failure can cascade into a widespread collapse. While previous research has studied interdependent networks with either perfectly ordered (identical) or completely random dependencies, real-world systems often fall somewhere in between. This paper addresses the gap by exploring how varying degrees of randomness in the dependency map between two spatially embedded networks influence their resilience to cascading failures.

Methods

The authors model two interdependent square lattice networks, A and B, where each node in A is linked to exactly one node in B. They introduce a parameter 'q' to control the randomness of these dependency links. When q=0, the dependency is identical (node i in A depends on node i in B). When q=1, the dependency is completely random. Intermediate values of 'q' represent a mix. They use Approximate Entropy (ApEn) to quantify the randomness of these dependency maps. The core of their analysis involves simulating cascade failures: nodes are randomly removed, and failures propagate through the interdependent links. They analyze the percolation threshold (the fraction of nodes that can be removed before the system collapses) and the number of iterations (NOI) it takes for the system to reach a steady state, which serves as a measure of the time scale of failures.

Datasets

Not discussed in the paper. The study uses simulated networks based on lattice structures, Watts-Strogatz networks, Erdős-Rényi networks, and scale-free networks.

Key Insights and Ideas

  • The "sweet spot" of randomness: Neither perfect order nor complete randomness in dependency links is optimal for network robustness. A moderate level of randomness appears to be most beneficial.
  • Phase transition control: The nature of the system's collapse (phase transition) shifts from continuous (second-order) to discontinuous (first-order) as dependency randomness increases beyond a critical point.
  • Percolation threshold dependency: The percolation threshold (pc) initially increases with randomness (q) but then plateaus, suggesting that beyond a certain point, further randomness doesn't significantly increase vulnerability to initial attacks.
  • Network topology matters: The critical level of randomness (qc) required to shift the phase transition varies significantly across different network topologies (lattice, scale-free, etc.), with lattice networks being the most vulnerable.
  • Time scale of failures: The time it takes for cascading failures to propagate is also influenced by randomness, showing distinct behaviors for different network types.

Key Results

  • A critical value of approximate entropy (ApEn) was identified. Below this critical ApEn, percolation is continuous; above it, it becomes a first-order transition.
  • The percolation threshold (pc) increases with dependency randomness (q) up to a certain point (q'c), after which it remains largely constant.
  • For interdependent square lattices, a critical q (qc ≈ 0.13) was found, below which the phase transition is second-order and above which it is first-order.
  • Different network topologies exhibit different critical values of q (qc): square lattices (0.13), Watts-Strogatz (0.52), Erdős-Rényi (0.61), and scale-free (0.87).
  • Interdependent square lattices were found to be the most vulnerable to random attacks, while scale-free networks were the most robust.
  • The number of iterations (NOI) for cascade failures shows rich properties depending on the degree of randomness and network type.

Limitations and Future Work

  • Limited dependency length: The paper briefly explores the influence of limited dependency length, suggesting it's less critical than randomness for robustness in linear dependency maps. However, a more thorough investigation into the interplay between dependency length and randomness could be beneficial.
  • Model Simplification: The model assumes a one-to-one dependency between nodes in the two networks. Real-world systems might have more complex dependency structures.
  • Future Directions: The authors suggest that their findings on the role of randomness in cascade failures could be extended to other dynamic processes in interdependent networks, such as synchronization and disease spreading. They also highlight the importance of understanding the time scale of cascade failures for disaster mitigation.

Summary updated

Abstract

The percolation for interdependent networks with identical dependency map follows a second-order phase transition which is exactly the same with percolation on a single network, while percolation for random dependency follows a first-order phase transition. In real networks, the dependency relations between networks are neither identical nor completely random. Thus in this paper, we study the influence of randomness for dependency maps on the robustness of interdependent lattice networks. We introduce approximate entropy(ApEn) as the measure of randomness of the dependency maps. We find that there is critical ApEnc below which the percolation is continuous, but for larger ApEn, it is a first-order transition. With the increment of ApEn, the pc increases until ApEn reaching ApEnc (') and then remains almost constant. The time scale of the system shows rich properties as ApEn increases. Our results uncover that randomness is one of the important factors that lead to cascading failures of spatially interdependent networks.
BibTeX
@inproceedings{yuan2016effect,
  author = {Jing Yuan and Lixiang Li and Haipeng Peng and Jürgen Kurths and Jinghua Xiao and Yixian Yang},
  title = {{The effect of randomness for dependency map on the robustness of interdependent lattices}},
  booktitle = {Chaos},
  year = {2016},
  doi = {10.1063/1.4939984},
}
Figure