Research On Multi-agent Cooperation Mechanism Based On Expanding Learning Range And Donation Mechanism

The phenomenon of cooperation can be seen everywhere in real life.If all walks of life want to better realize the rational allocation and effective use of resources,it is particularly important to study the generation and maintenance of cooperation in the network.Evolutionary game theory is an effective tool for studying the cooperative behavior between groups.So far,scholars have proposed a variety of mechanisms to define the interaction between individuals,and proved that these mechanisms have a positive impact on promoting cooperation.In recent years,some researchers have proposed a method to increase the number of neighbors participating in the game,so that the range of neighbors increases in the form of a square ring.The incremental step size of this method is not a fixed value.The mechanism of expanding the learning range proposed in this paper is based on the traditional The von Neumann neighbor’s propagation method based on neighbor propagation is more theoretically meaningful.In this paper,the neighbor distribution of the agent under the learning ability of different transfer factors and the corresponding calculation formula are given.The mechanism is introduced in the three dilemma models.The results show that in the prisoner game,when the policy update rule is replicator dynamics,the transfer Factors play a particularly positive role in the influence of cooperation.It greatly enhances network reciprocity and makes it easier for the system to form stable cooperative clusters.In the snowdrift game,the influence of the transfer factor on cooperation presents two extremes.A certain fixed value of the betrayal factor r is the cut-off point.When the value is less than this value,it shows a hindrance to the cooperation.Active promotion.Under the unconditional strategy update rule,the deer hunting game,under the action of the transfer factor,the system is constantly evolving towards a cooperative equilibrium state.In order to further explore the promotion effect of the expansion of the learning scope mechanism on cooperation,and considering the heterogeneity of individual learning abilities,this paper conducted experiments with local random number nodes with the ability to expand the learning scope and related control experiments,the results obtained are consolidated The previous conclusion.This article proposed a donation mechanism in line with social public welfare behaviors,with a view to providing some new ideas for the evolutionary game dynamics.On the basis of the traditional situation,after each round of the game was completed,the player with the highest profit would allocate a part of his own revenue to the neighbor with the lowest revenue.The experimental results showed that the donation mechanism could effectively promote the maintenance of systematic cooperative behavior in the prisoner game.For the snowdrift games and deer hunting games under the dynamic update policy of the replicator,the donation mechanism had a non-monotonic effect on cooperation.The maintenance of be affected by the betrayal factor r and the donation coefficient e.The deer hunting game with unconditional rules was most strongly affected by the donation mechanism.With the introduction of the donation mechanism,the system cooperation rate increased sharply.When the donation coefficient e exceeded a certain value,the system moved to a purely strategic state of simultaneous cooperation.This paper based on mechanism for expanding the scope of learning and the donation mechanism,experiments were designed to compare the different values of learning ability and system cooperation based on donations.And considering the heterogeneity of individuals,an experiment was designed with random proportion of local nodes under the donation mechanism to have the same learning ability,in order to explore the comprehensive impact of these two mechanisms on system cooperation.Experimental results further confirmed the robustness of these two mechanisms.