SIGMAS: Second-Order Interaction-based Grouping for Overlapping Multi-Agent Swarms

Researchers introduce SIGMAS, a self-supervised AI framework for identifying group structures in multi-agent swarms like drone fleets without ground-truth supervision. The system uses second-order interactions to infer latent group memberships from agent trajectories, demonstrating robust performance across diverse synthetic swarm scenarios.

• →SIGMAS addresses group prediction in overlapping multi-agent swarms through self-supervised learning without requiring labeled training data.
• →The framework models second-order interactions between agents rather than just direct pairwise relationships for more accurate group inference.
• →A learnable gating mechanism adaptively balances individual and collective dynamics in swarm behavior analysis.
• →Experiments show the system accurately recovers latent group structures across diverse synthetic swarm scenarios.
• →The research establishes both a new benchmark task and modeling framework for understanding complex swarm dynamics.