Inferring interaction domains of collectively moving agents with varying radius of influence

Abstract

Transfer entropy (TE) has proven to be an effective tool for determining the causal connection between two processes. For example, TE has been used to classify leader and follower agents in collective dynamics in the Vicsek model (VM). However, previous results have limited interaction radii, which are precisely the same among all agents, which is not realistic in practice. Here, we propose a modified version of the VM where the domains in which an agent can be influenced by others vary from agent-to-agent, which matches more closely with a real-life setting where not all agents have the same physical traits. We demonstrate that the TE vs cut-off technique is robust and efficacious in determining the maximum distance at which two interacting agents can transfer information in the system. We find that for two agents with different interaction domains, the derivative of the average inward TE can determine the individual agent’s interaction domain. In a system with numerous agents, the TE vs cut-off technique is shown to be effective in predicting the average interaction domain of all agents, where the interaction domain of each agent was randomly selected from a Gaussian distribution.