Research On Mechanism Design Of Network Evolutionary Game And Its Applications

It is well known that cooperation exists widely in many biological, social and economic sys-tems, and plays an important role in the functionality of these systems. Given the selfish nature of individuals, how the cooperation is performed has been a puzzle to researchers. It is still an open issue. Therefore, investigating the emergence and maintenance of coopera-tive behavior among selfish individuals is of vital importance. Meanwhile, along with the promotion of data processing ability, researchers found that many real systems can be seen as networks in which individuals and their connections are represented as nodes and edges. Then, researchers focus on utilizing the interaction between the network structures to explore the internal relations under real system.Therefore, in order to investigate the emerging mechanisms of cooperation generating among competitive individuals, network topology is applied to study the competing behaviors among individuals. Recently, such an king of research, named as network evolutionary game theory, has attracted extensive attention among researchers. In the field of evolutionary game theory, mechanism design aims at designing suitable rules, guidance and advice to induce selfish individuals’ behaviors to achieve the optimization of both their own interest and the overall system.Inspired by the current international research interests, in this dissertation, we focus on proposing several mechanisms design which could promote cooperation in the prisoner’s dilemma game. On the other hand, we investigate the application of game theory by combining it with pedestrian evacuation model which might help escape. The main work and contribution of this dissertation are summarized as follows.In spite of the achievements in recent years, the consideration of environment during the stage of strategy updating of a player and its neighbors has received relatively little at-tention till now. Although this kind of mechanism is almost ubiquitous in an individual’s decision-making process, only a few studies have addressed the effects of environment on evolutionary dynamics. Therefore, it is of future interest to inspect the dynamics of prison-er’s dilemma game with the consideration of environment. As a matter of fact, the strategy updating is a decision-making process. Every player has to decide either to keep its own strategy or to imitate its neighbor’s strategy. Some smart players would adopt a more clear strategy-updating decision by considering not only the payoff but also the environment of their neighbors. Accordingly, we propose a new mechanism design which takes both payoffs and the environments of individuals’ neighbors into consideration when making the strategy updating decision. We investigate how the cooperation emerges in the square lattice when individuals play the prisoner’s dilemma game by adopting different strategy updating meth-ods. First, we classify individuals into two classes at the strategy updating stage:stochastic players and smart players. The stochastic players are those who take the classical Fermi rule, while the smart players are those who make comprehensive evaluation for strategy updating. Second, we adopt Dempster-Shafer theory to combine smart players’ evaluations from payoff and environment aspects. Three different models have been proposed in this dissertation to describe the effect of comprehensive consideration on promoting cooperation. The first model considers global information, local information and payoff. In some situation, it is difficult to acquire global information; therefore, we propose the second model which just considers the local information and payoff. In the third model, a smart player only selects and imitates neighbors who own different strategy. It has to decide whether it want to imitate one of its neighbors or not. Simulation results reveal that all the three models have positive impact on the emergence of cooperation. It indicates that the consideration of environment not only benefit to individuals but also to the overall system.In the previous works, those who impose the punishment on the behavior of defect pay the corresponding cost. Instead, in this dissertation, we study the effect of punishment and reward in sustaining cooperation in prisoner’s dilemma game over square lattice from new aspect. By simulating the tax scheme to charge the public funds as a punishment for defect, we specify the scope and severity of punishment to better promote cooperation. We find that moderate scope and severity punishment works best. However, it is difficult to implement this kind of punishment in practical application. Thus, it is justifiable to choose wide scope and light punishment rather than small scope but severe sanction, because the former sustains higher level of cooperation compared with the latter. We also discuss how the evolutionary cooperation will be affected by the rewards that the public funds offer for cooperative be-haviors. Interestingly, the result is similar to the effect of punishment. It is wise to reward wide scope of cooperators rather than small numbers of them. Although the former means lower amount of incentives, it sustains higher level of cooperation. Furthermore, the method of comprehensive consideration in strategy updating stage is applied to the punishing and rewarding by public funds models. Then, two models—punishing by public funds based on local information and rewarding by public funds based on local information—are proposed. The simulation results indicate that the comprehensive consideration method could promote cooperation in both models. Meanwhile, the average payoff is also increased. Therefore, the effectiveness of comprehensive consideration of environment into strategy updating has been further verified. The research finding is pretty suggestive to the regulating of social wealth distribution.Finally, we investigate the application of game theory by combining it with pedestrian evacuation model to reveal the mechanism design which could decrease the escape time. Actually, accelerating escape time is another state of cooperation. First, we find that when the overall population owns appropriate fear index, it is useful for the individuals falling behind to keep a higher level of fear index, which can speed up the escape of the whole population. The finding has a guiding significance to the persons falling behind in emergency escape. Second, the comprehensive consideration of payoff and environment information in strategy updating is utilized to pedestrian evacuation model based on game theory. The results demonstrate that the more pedestrians utilize the comprehensive method, the more the escape time is decreased. Third, we propose a new pedestrian evacuation model based on game theory in which the pedestrian has priority direction. Based on this model, the best location of a door in rectangle room and square room is found. In order to help pedestrians escape in emergency, the best location of a door is in the middle of wall, which in consistent with the intuitive feeling of human. Meanwhile, it also indicates that this model successfully simulates the competing behavior in the process of escape. Last, we further study the escape effect when some pedestrians using learning activity. We name this kind of pedestrians as learners. In the proposed model, a learner would observe its neighbors’ strategy and its neighbors’ strategy updating probability. Then, it can choose an optimal strategy to get highest interest. With the involvement of learners, the escape time of the whole population can decrease. The research achievement suggests that it would be beneficial for some pedestrians to utilize the above-mentioned skills, because it can accelerate the escape time.In this dissertation, we further specify how cooperation emerge and evolve. Then, the method of punishing defectors and rewarding cooperators can offer guidance to decision mak-ers. Last, the proposed pedestrian evacuation model might help emergency escape in public places.