Controlling Self-Organizing Dynamics on Networks Using Models that Self-Organize

Pierre‐André Noël, Charles D. Brummitt, Raissa M. D’SouzaPhysical 2013 · DOI · arXiv

Abstract

Controlling self-organizing systems is challenging because the system responds to the controller. Here, we develop a model that captures the essential self-organizing mechanisms of Bak-Tang-Wiesenfeld (BTW) sandpiles on networks, a self-organized critical (SOC) system. This model enables studying a simple control scheme that determines the frequency of cascades and that shapes systemic risk. We show that optimal strategies exist for generic cost functions and that controlling a subcritical system may drive it to criticality. This approach could enable controlling other self-organizing systems.

Introduction: The Control Paradox

Many complex systems—from power grids and financial markets to neural networks—exhibit Self-Organized Criticality (SOC). These systems naturally evolve toward a state where a minor disturbance can trigger a massive, system-wide cascade. Managing such systems presents a unique "Control Paradox": interventions intended to suppress small failures can cause the system to reorganize and accumulate stress, eventually leading to even larger, catastrophic collapses.

To manage these systems effectively, researchers need mechanistic models that understand the underlying rules of adaptation rather than just describing the current state. Without such models, a controller might inadvertently push a stable system into a critical or supercritical state.

Methodology

The study focuses on the Bak-Tang-Wiesenfeld (BTW) sandpile model on networks. In this model, "grains" (representing load) are added to nodes. When a node's capacity is reached, it "topples," distributing grains to neighbors and potentially triggering a cascade.

The researchers introduced two primary innovations:

  1. A Mechanistic Model ($M_{BTW}$): Utilizing a multitype branching process and self-consistency equations, this model predicts how a system will settle into a new steady state following a control intervention.
  2. A Control Parameter ($\mu$): This parameter represents the frequency of cascades. By selecting where the initial grain lands, a controller can influence how often cascades are triggered.

The model was validated using Monte Carlo simulations on random 3-regular graphs with up to one million nodes, providing a robust "ground truth" for cascade distributions.

Key Insights and Results

The core finding is that to control an SOC system, the model itself must be capable of self-organizing. A static model cannot account for how the system rearranges its internal stress in response to new rules.

The Danger of Avoidance

Reducing the frequency of cascades ($\mu$) to maintain short-term stability often moves the system closer to the critical line. As $\mu$ approaches zero, the probability of a massive failure increases significantly. The researchers successfully mapped these "supercritical," "subcritical," and "critical" regions across the parameter space.

Phase diagram showing the transition between subcritical and supercritical states as a function of control parameter mu.

Optimal Risk Shaping

There is no single "best" state; rather, there is a trade-off between the frequency and size of failures. Frequently triggering small cascades (high $\mu$) prevents massive ones but incurs constant minor costs. Conversely, avoiding cascades (low $\mu$) risks a "black swan" event. The study demonstrates that for generic cost functions, there exists a mathematically optimal frequency ($\mu^*$) that minimizes the total expected cost.

Expected cost as a function of control frequency mu, demonstrating the existence of an optimal strategy mu*.

Limitations and Future Work

While the use of 3-regular networks is mathematically convenient, real-world topologies are often more complex, featuring community structures or heavy-tailed degree distributions. Future research directions include:

  • Controlling Cascade Area: Developing methods to isolate damage to specific regions of a network.
  • Dynamic Time Scales: Modeling controllers that can intervene during an ongoing cascade.
  • Real-world Applications: Applying this framework to liquidity flows in banking systems or information dynamics in social networks.

This mechanistic approach provides a foundation for managing systems that "fight back," offering a path toward more resilient infrastructure and economic stability.

Summary updated

Abstract

Controlling self-organizing systems is challenging because the system responds to the controller. Here, we develop a model that captures the essential self-organizing mechanisms of Bak-Tang-Wiesenfeld (BTW) sandpiles on networks, a self-organized critical (SOC) system. This model enables studying a simple control scheme that determines the frequency of cascades and that shapes systemic risk. We show that optimal strategies exist for generic cost functions and that controlling a subcritical system may drive it to criticality. This approach could enable controlling other self-organizing systems.
BibTeX
@inproceedings{noel2013controlling,
  author = {Pierre‐André Noël and Charles D. Brummitt and Raissa M. D’Souza},
  title = {{Controlling Self-Organizing Dynamics on Networks Using Models that Self-Organize}},
  booktitle = {Physical},
  year = {2013},
  doi = {10.1103/physrevlett.111.078701},
}
Figure