Research On The Networked Evolutionary Game Theory Based On Diverse Agents

Cooperative behaviors are ubiquitous in nature and social systems.Networked evolutionary game theory provides an effective tools to understand the emergence of cooperative behaviors among selfish populations.Complex networks describe the topology of networks and game model depicts the interaction patterns between agents.Network structures and agents' behaviors have a variety of forms,and this diversity provides the basis for selection and is therefore critical to the emergence of cooperation.As a new theory in the repeated game,zero-determinant strategies can unilaterally guarantee that the long-term payoff of both sides meet the linear relationship.Extortion strategy,as an important subset of zero-determinant strategy,can ensure that its own payoff is not lower than that of its opponent,which has been widely studied by scholars recently.Under the framework of networked evolutionary game theory,this dissertation applies the theories and methods of network science,evolutionary game and networked evolutionary game to explore the influence of network structure and the diverse patterns of agents' behaviors on game dynamics.The main contents and innovations of this dissertation are as follows:Firstly,this dissertation studies the evolution of cooperation on different scale-free networks from the perspective of diverse agents' interaction.On the Barab(?)si-Albert(BA)scale-free network,the alliance between cooperation and extortion can emerge based on the replicator dynamic update rule under the normalized payoff framework.And extortion strategies can act as catalysts to promote the evolution of cooperation.Furthermore,the extortion factor plays a nonmonotonic role on the evolution of cooperation,which implies that proper values of the extortion factor can promote the cooperative level on networks.Subsequently,based on Fermi dynamic and mutation,this dissertation studies the interaction and evolution of cooperation,defection and extortion strategies on clustered scale-free networks.Our results show that with a small mutation rate,extortionate behaviors can emerge on the scale-free network with high clustering coefficient and promote the evolution of cooperative behaviors on the network.Through the micro analysis,it is shown that extortion strategy tends to occupy these hubs and induce more small-degree neighbors becoming cooperators.Furthermore,it is shown that the proper values of mutation is conducive to the emergence of cooperative behaviors on the scalefree network with high clustering coefficient.While the mutation inhibites cooperative behaviors on the scale-free network with low values of clustering coefficient.Secondly,this dissertation studies the influence of the diverse agents' extortionate behaviors on the evolution of cooperation.Based on the repeated Prisoner's Dilemma game model and myopic best response rule,agents' extortionate behaviors couple with their degrees to study the evolution of cooperative behaviors on the BA scale-free network.It is found that under both the accumulated payoff framework and the normalized payoff framework,when agents' extortion factors are negatively related with their degrees,diverse extortion strategies can act as catalysts to promote the level of cooperation on the network,while agents' extortion factors are positively related with their degrees,the effect of extortioners' catalysts is significantly weaken.Furthermore,it is shown that the rationality of agent plays a nonmonotonic role on the evolution of cooperation on the network,and some proper values of rationality can enhance the level of cooperation on the network.Then,this dissertation studies the role of time scale diversity caused by agents' centrality in the evolutionary game dynamic.Based on the weak Prisoner's Dilemma model,we associate the agents' collective influence with their strategy-updating time scales to investigate how the diverse collective influence of agents affects the evolution of cooperation on BA scale-free networks.The strategy-updating time scale can form feedback with its fitness and centrality on the network.It is found that influential cooperators locating at medium-or small-degrees are able to spread their behaviors among neighbors in a more efficient way than agents with large-degrees.Collective influence with proper path length can efficiently identify influencers and may promote the emergence of cooperation on heterogeneous networks.Furthermore,agents with high values of collective influence can hold cooperation strategy for longer time steps and diffuse their cooperative behaviors among neighbors through analyzing the learning patterns during the steady states.At the same time,heterogeneous hierarchical learning structures emerge on the game-learning skeleton.Subsequently,empirical analysis is carried out on real social networks and it is found that collective influence with proper depth length may promote the emergence of cooperation on real-world system.Then,results in donation game show that agents with high values of collective influence may hold cooperation for a longer time,thus,promote the cooperation level on the network.When introducing the extortion strategy,the the evolution of cooperation is almost unchanged at small values of benefit factor.While at large values of benefit factor,the extortion factor with large values can further promote the evolution of cooperation when considering collective influence with proper depth length.Under the accumulated payoff framework,the strategy-updating time scale mechanism inhibits the evolution of cooperation on the scale-free networks.Finally,the dissertation explores the influence of diverse strategy-updating time scale on the evolution of cooperation,defection and extortion strategies on a two-layer regular graph.Agents can adjust their frequencies of strategy-updating adaptively according to the fitness and the interlayer information between these two layers of networks.Based on the Fermi dynamics,it is shown that the information sharing between layers can effectively promote the cooperative behaviors on the lattice.When the extortion factor is the same,the trend of strategy evolution on the two layers is similar.On each lattice,cooperation-extortion alliance can be formed to defend the invasion of defection.In the case of small values of benefit factor,at small values of extortion factor,the rationality of agents with large values can promote the emergence of cooperation.While,at large values of extortion factor,the rationality of agents plays a non-monotonic role in the evolution of cooperation,and proper rationality of agent to promote the evolution of cooperation.When facing the high temptation to defect,the rationality of the agents has little effect on the evolution of strategies.At the same time,the optimal value of extortion factor makes the cooperation level on the network highest,and the cooperation level on two-layer lattice is higher than that of on the single-layer lattice.