Swarm Characteristics Classification Using Neural Networks

TL;DR

The paper addresses the problem of rapidly inferring swarm characteristics to inform counter-maneuvers in defense contexts. It proposes supervised neural network time series classification trained on simulated swarm-on-swarm data to predict two binary attributes (Comms and ProNav) and the resulting four tactics, leveraging a multihead architecture and sequence outputs. Key findings show high predictive accuracy (up to $97\%$) with a short observation window of $20$ time steps, robustness up to $50\%$ noise with about $80\%$ accuracy, and scalability from $N_A,N_D=10$ to $100$, with sequence-output transformers like TRS delivering strong performance. These results indicate NN TSC can provide real-time, behavior-aware defense intelligence under constrained sensing and communication, enabling faster and more informed countermeasures for autonomous swarm engagements.

Abstract

Understanding the characteristics of swarming autonomous agents is critical for defense and security applications. This article presents a study on using supervised neural network time series classification (NN TSC) to predict key attributes and tactics of swarming autonomous agents for military contexts. Specifically, NN TSC is applied to infer two binary attributes - communication and proportional navigation - which combine to define four mutually exclusive swarm tactics. We identify a gap in literature on using NNs for swarm classification and demonstrate the effectiveness of NN TSC in rapidly deducing intelligence about attacking swarms to inform counter-maneuvers. Through simulated swarm-vs-swarm engagements, we evaluate NN TSC performance in terms of observation window requirements, noise robustness, and scalability to swarm size. Key findings show NNs can predict swarm behaviors with 97% accuracy using short observation windows of 20 time steps, while also demonstrating graceful degradation down to 80% accuracy under 50% noise, as well as excellent scalability to swarm sizes from 10 to 100 agents. These capabilities are promising for real-time decision-making support in defense scenarios by rapidly inferring insights about swarm behavior.

Figures (16)

• Figure 1: Simulation example showing velocity vectors at initialization time = 0.
• Figure 2: Two binary attacker attributes combine to create four mutually exclusive swarm tactics.
• Figure 3: Simulation visualization showing all four tactics (Greedy, Greedy+, Auction, Auction+) with identical initialization at various engagement time steps; agent velocity vectors shown.
• Figure 4: Multihead classifier.
• Figure 5: Data generation parameters varied to develop baseline dataset for model comparison.