The Effect of Mobility and Emotion on Interactions in Multi-Agent Systems

摘要

Simulating emotions within a group of agents has been shown to support co-operation in the prisoner's dilemma game. Most work on simulating these emotions has focused on environments where the agents do not move, that is, they are static and their neighbours are fixed. However, it has also been shown in other work that when an agent is given the ability to move, then the type of the environment affects how co-operation evolves in the group of agents. In this paper, we investigate the combination of these two ideas in an experimental study that explores the effects on co-operation when autonomous agents that can show emotions are given the ability to move within structured environments. We observe that once mobility is introduced, different strategies become successful. Successful strategies respond quickly to defection, while not immediately reciprocating co-operation, regardless of the environment type. The further an agent travels, the higher its average payoff in a small world environment. The slower an agent is to copy another agent by imitating its strategy, the higher its increase in average payoff.