Collective Motion of Predictive Swarms

TL;DR

This paper explores how predictive behavior influences swarm dynamics, revealing that limited foresight and small predictor populations enhance resource consumption and overall efficiency.

Contribution

It introduces a model of predictive agents in swarms, analyzing how anticipation affects collective motion and resource gathering compared to non-predictive agents.

Findings

Predictive agents perform better when their foresight is limited.

Predictors outperform non-predictors only when they are few in number.

Long-term prediction attempts reduce resource intake.

Abstract

Theoretical models of populations and swarms typically start with the assumption that the motion of agents is governed by the local stimuli. However, an intelligent agent, with some understanding of the laws that govern its habitat, can anticipate the future, and make predictions to gather resources more efficiently. Here we study a specific model of this kind, where agents aim to maximize their consumption of a diffusing resource, by attempting to predict the future of a resource field and the actions of other agents. Once the agents make a prediction, they are attracted to move towards regions that have, and will have, denser resources. We find that the further the agents attempt to see into the future, the more their attempts at prediction fail, and the less resources they consume. We also study the case where predictive agents compete against non-predictive agents and find the predictors perform better than the non-predictors only when their relative numbers are very small. We conclude that predictivity pays off either when the predictors do not see too far into the future or the number of predictors is small.