Two swarms moving in opposite directions collide. Each swarm has its own stable velocity — maintained by internal alignment and cohesion forces — before impact. After collision, three outcomes are possible: both swarms continue through each other, both reverse direction, or one absorbs the other. The outcome depends on whether a stable synchronized velocity exists in the composite system.
Note (2026-11-12): checked against the paper's full text. The central result matches: redirection occurs when a stable velocity-synchronized state of the combined swarm exists, and a rigid-body approximation gives scaling laws for the scatter-redirection transition, compared in the paper with particle simulations and wheeled-robot simulations. Two points are stronger in the essay. The paper distinguishes scattering from redirection (the swarms coalesce into one flock whose direction is set by the synchronized state, which can reverse one of the swarms); I did not find the essay's three outcomes of passing through, both reversing, or one absorbing the other. And the saddle-node bifurcation is the paper's statement for reciprocal interactions, where the stability matrix is symmetric with real eigenvalues; for non-reciprocal swarms the paper says only that fluctuations are more generally described, so 'the transition is a saddle-node bifurcation' is not claimed for all cases.
The scatter-redirection transition — the boundary between swarms passing through and swarms being redirected — can be computed from the composite system's stability landscape (arXiv:2603.12002). Using a rigid-body approximation where each swarm maintains its internal cohesion during collision, the researchers determine scaling laws for the transition across reciprocal, non-reciprocal, symmetric, and antagonistic configurations.
The key insight is that redirection occurs not through individual particle interactions but through a macroscopic stability transition. When two swarms collide, the question is whether the velocity of the combined system has a stable fixed point different from the pre-collision velocities. If yes, the collision redirects. If no, the swarms pass through.
This reframes swarm collision as a bifurcation problem. The collision parameters (relative velocity, swarm size, interaction strength) are bifurcation parameters, and the transition from scattering to redirection is a saddle-node bifurcation in the velocity space of the composite system. Individual particle dynamics are integrated out; the macroscopic transition is what matters.
Redirection is not the particles changing their minds. It is the composite system finding a new stable state that neither swarm had alone.