[Paper Review] Asynchrony promotes polarized collective motion in attraction based models

Daniel Strömbom, Tasnia Hassan+2 more|arXiv (Cornell University)|Oct 6, 2017

Distributed Control Multi-Agent Systems17 references3 citations

TL;DR

This study demonstrates that asynchronous update mechanisms—where particles update their headings at random times—induce polarized collective motion in attraction-based self-propelled particle models, whereas synchronous updates (all particles update simultaneously) fail to produce polarization. The key finding is that update timing, not just interaction rules, critically determines emergent group behavior.

ABSTRACT

Animal groups frequently move in a highly organized manner, as represented by flocks of birds and schools of fish. Despite being an everyday occurrence, we do not yet fully understand how this works. What type of social interactions between animals gives rise to the overall flock structure and behavior we observe? This question is often investigated using self-propelled particle models where particles represent the individual animals. These models differ in the social interactions used, individual particle properties, and various technical assumptions. One particular technical assumption relates to whether all particles update their headings and positions at exactly the same time (synchronous update) or not (asynchronous update). Here we investigate the causal effects of this assumption in a specific model and find that it has a dramatic impact. In particular, polarized groups do not form when synchronous update is used, but are always produced with asynchronous updates. We also show that full asynchrony is not required for polarized groups to form and quantify time to polarized group formation. Since many important models in the literature have been implemented with synchronous update only, we speculate that our understanding of these models, or rather the social interactions on which they are based, may be incomplete. Perhaps a range of previously unobserved dynamic phenomena will emerge if other potentially more realistic update schemes are chosen.

Motivation & Objective

• Investigate the causal impact of update timing (synchronous vs. asynchronous) on collective motion in self-propelled particle (SPP) models.
• Address a long-ignored technical assumption in SPP modeling: whether particle updates occur simultaneously or staggered.
• Determine whether the choice of update scheme affects the emergence of polarized groups, mills, or swarms in minimal models.
• Challenge the assumption that synchronous updates are sufficient or realistic for modeling collective animal motion.
• Highlight that overlooked update schemes may lead to missing dynamic phenomena in existing SPP models.

Proposed method

• Implement the Local Attraction Model (LAM), a minimal SPP model with only attraction-based interactions and constant speed.
• Compare two update schemes: fully synchronous (all particles update at once) and fully asynchronous (randomized, independent update times).
• Use periodic boundary conditions and track particle positions and headings over time steps.
• Define polarization (α) as the magnitude of the average heading vector; scaled size (σ) measures group cohesion.
• Apply termination criteria: α > 0.995 for polarized groups, σ < 0.01 for cohesive groups, and α < 0.02 with 0.01 < σ < 0.25 for large mills.
• Measure mean α and σ over the last 50 timesteps and record time to group formation (τ).

Experimental results

Research questions

• RQ1 Does the choice between synchronous and asynchronous update schemes affect the emergence of polarized collective motion in attraction-based SPP models?
• RQ2 Can polarized groups form under synchronous update in the Local Attraction Model (LAM)?
• RQ3 Is full asynchrony necessary for polarization, or can partial asynchrony suffice?
• RQ4 How does the time to formation of polarized groups vary between synchronous and asynchronous update schemes?
• RQ5 To what extent does the update scheme influence the validity and completeness of SPP models in explaining real-world collective motion?

Key findings

• Polarized groups fail to form under fully synchronous update in the LAM, despite the model being known to produce polarization under other conditions.
• Under fully asynchronous update, polarized groups consistently form and remain stable, indicating that asynchrony is a necessary condition for polarization in this model.
• Polarized group formation is significantly faster under asynchronous update, with formation times measured in hundreds of timesteps, compared to over 10^8 timesteps under synchronous update.
• Partial asynchrony—where only a fraction of particles update asynchronously—still enables polarization, showing that full asynchrony is not strictly required.
• The study identifies that the update scheme is not a neutral technical detail but a critical factor that can determine whether key collective behaviors emerge.
• The results suggest that many existing SPP models may be incomplete or misrepresentative due to the exclusive use of synchronous updates, potentially obscuring emergent dynamics.