Evolutionary Game Dynamics Based on Local Intervention in Multi-Agent Systems

How to model and control the networked game dynamics of collective behaviors in multi-agent systems remains a challenging problem. This brief proposes an interacting decision-making model for investigating the evolutionary dynamics of networked systems. The proposed model establishes a local intervention control scheme to promote and control evolution of cooperation. To drive the network dynamics, agents update according to the adaptive dynamics, under which they can adjust their actions along with the direction of optimal payoffs. An important factor for the system convergence is discussed here: asynchronous updating can guarantee the convergence of the local intervening gaming network, while full synchrony may never result in convergence. Under the asynchronous rule, local interventions of neighbors, which can be seen as a neighboring feedback information, contribute to enhance agents' preferences to cooperate. The results in this brief are expected to offer guidance on system performance optimizations, and help to develop new protocols to control collective decision-making in many real-world social and engineering systems.