Improving coordination via emergent communication in cooperative multiagent systems: A Genetic Network Programming approach

The design and development of strategies to coordinate the actions of multiple agents is a central research issue in the field of multiagent systems. To address this issue, in our previously reported research we proposed a novel methodology based on genetic network programming (GNP) to allow agents in the pursuit domain to autonomously learn an effective coordination strategy in order to achieve group behavior. GNP is a newly developed evolutionary computation (EC) technique whose genome is network structure. In this paper, we extend our methodology by allowing agents in the pursuit domain to autonomously learn communication. We consider autonomous and independently learning agents, and we seek to obtain an optimal solution for the team as a whole. We design a novel methodology for the emergence of communication between agents in cooperative multiagent systems based on GNP and we seek to obtain more coordination. Through simulations, we demonstrate that the proposed approach is effective in evolving communicating agents and furthermore a comparison is made between agents with and without communication in order to show that the emergent communication among agents is beneficial i.e., improves their coordination. In addition, we show the robustness of generated programs which is achieved as a side-effect of the capability of communication.