Bimodal Synchronization Performance: Why Noise and Sparse Connectivity Can Improve Collective Timing

Technical Summary: Bimodal Synchronization Performance in Firefly-Inspired Models

Problem Statement
Decentralized time coordination is a fundamental challenge in distributed systems, ranging from biological swarms (e.g., fireflies) to engineered networks (e.g., wireless sensor networks and swarm robotics). While classical pulse-coupled oscillator models suggest that increasing connectivity generally facilitates synchronization by providing agents with more timing information, this paper investigates a counterintuitive phenomenon: highly connected, noiseless systems can fail to synchronize. Specifically, the authors analyze a discrete-time, discrete-phase synchronization model to understand why systems sometimes exhibit "bimodal" performance—where identical parameter settings yield either near-perfect global synchrony or stable, suboptimal multi-cluster states.

Methodology
The authors propose and analyze a discrete-time, firefly-inspired synchronization model characterized by two primary mechanisms:

1. Quorum Threshold ($\theta$): An agent advances its phase only if the fraction of its flashing neighbors exceeds a predefined threshold.
2. Flash Duration ($f$): The fraction of the cycle during which an agent remains in a signaling (flashing) state.

The system consists of $N$ agents interacting over either $k$-regular bidirectional random graphs or random geometric graphs. Agents update their internal clocks discretely. If the fraction of flashing neighbors exceeds $\theta$, the agent advances its clock by an additional tick. The study quantifies synchronization success via the maximal amplitude ($A_{max}$), defined as the maximum fraction of agents flashing simultaneously during a simulation run. A run is considered successful if $A_{max} \geq 0.85$.

To investigate the causes of synchronization failure, the authors introduce two types of perturbations:

• Noise: Introducing stochasticity into the quorum decision rule (flipping the update decision with probability $\sigma$).
• Reduced Connectivity: Decreasing the number of communication links (reducing degree $k$ or range $r$) to break interaction symmetry.

Key Findings and Results
The analysis reveals a critical relationship between the quorum threshold ($\theta$) and the flash duration ($f$):

1. Critical Balance and Bimodality: High-performance synchronization occurs only in a narrow parameter region where $\theta \approx f$. Within this region, the system exhibits bimodal behavior. Independent simulation runs with identical parameters either achieve global synchrony or become trapped in stable multi-cluster states.
2. Symmetry-Induced Subgroup Locking: The authors identify the mechanism behind the low-performance states as "symmetry-induced subgroup locking." In highly connected, noiseless environments, agents self-organize into multiple subgroups with symmetric phase offsets. Because the quorum rule is satisfied by the presence of other subgroups (which provide the necessary fraction of flashing neighbors), these groups mutually reinforce their phase offsets. This creates a deadlock where the system remains partitioned into stable, phase-shifted clusters rather than converging to a single state.
3. The "Less is More" Effect: Contrary to the intuition that more connectivity improves synchronization, the study demonstrates that reducing connectivity or introducing noise can improve collective performance.
   • Noise: Introducing noise into the update rule ($\sigma > 0$) disrupts the precise symmetric phase relationships required to maintain subgroup locking. Moderate noise levels ($0.1 \leq \sigma \leq 0.4$) significantly increase the probability of global synchronization by breaking these locked states. However, excessive noise ($\sigma > 0.5$) destroys synchronization by inverting the update rule too frequently.
   • Connectivity: Reducing the number of links (increasing link removal $\sigma$) similarly breaks the symmetry of interactions perceived by different agents. This destabilizes the multi-cluster states, allowing the system to escape low-performance configurations. The effect is most pronounced when reducing connectivity from a fully connected state, though extremely sparse networks eventually fail due to lack of information spread.
4. Parity Effects: In sparse $k$-regular graphs, the authors observe that synchronization success depends on the parity of the degree $k$. Even degrees ($k=6, 8, \dots$) facilitate synchronization more effectively than odd degrees ($k=5, 7, \dots$) under specific threshold conditions, suggesting a subtle interaction between the quorum rule and neighborhood size.

Significance and Claims
The paper claims that the failure of highly connected systems to synchronize is not merely a result of physical congestion (as seen in some multi-robot traffic scenarios) but a dynamical phenomenon driven by symmetry-induced locking in phase space.

• Counter-Intuitive Optimization: The work challenges the assumption that perfect information and high connectivity are always optimal for decentralized coordination. Instead, it posits that "imperfections" such as noise or limited connectivity can be beneficial by preventing the system from settling into stable, low-performance symmetric configurations.
• Generalizability: The authors suggest that bimodal performance distributions may be a general feature of scalable distributed systems. Poor scalability may arise not only from physical obstruction but also from overly regular interaction structures that prevent symmetry breaking.
• Mechanism Distinction: While similar bimodal behaviors have been attributed to spatial congestion in other multi-robot studies, this paper identifies a distinct mechanism: the stabilization of symmetric, phase-offset subgroups through quorum-based mutual reinforcement. Unlike spatial congestion, this model does not exhibit hysteresis; once global synchronization is achieved, changing connectivity does not necessarily destroy it.

In conclusion, the study demonstrates that in discrete pulse-coupled oscillator systems, the path to global synchrony may require breaking the very symmetries that high-connectivity networks provide, highlighting the potential utility of noise and sparsity in collective timing protocols.