Bio-inspired AI swarm simulation platform

Cosmos Institute granted Alfaxad Eyembe on September 10, 2024 as part of the inaugural Cosmos Ventures cohort, with funding in the $1K to $10K range. Cosmos Institute described the project in its announcement as "creating ALFA, an open-source simulation platform for bio-inspired AI swarms."

ALFA is an artificial-life simulation platform written entirely in Rust and published openly on GitHub at Alfaxad/ALFA. The system simulates heterogeneous swarms in which multiple agent species coexist, each with its own morphology, interaction radius, and behavioral rules. The dynamics are physics-based: attraction and repulsion forces governed by per-species coefficient matrices let populations self-organize around objectives like cluster cohesion or formation achievement. Built-in evolutionary mechanisms then optimize the interaction coefficients over generations, which lets researchers study how agent diversity influences collective stability and emergent behavior rather than running only homogeneous Boid-style simulations. The rendering and UI layer is built on Bevy, the Rust game engine, with bevy_egui for controls and bevy_prototype_lyon for geometric primitives.

Public material on ALFA frames it as a research substrate rather than a deployed product: the design goal is to expose population-level levers (viscosity, gravity, wall repulsion, species ratios, morphological diversity) so researchers can probe the conditions under which decentralized agent populations negotiate goals, share information, and resolve conflicts without a central controller. That places the project at the agents layer with evaluation as secondary attribution, since swarm simulators function both as research tools and as benchmark environments for multi-agent reasoning.