The Evolution Of Cooperation In Ad Hoc Networks

In Ad Hoc networks, packets need to be forwarded by other nodes if there is more than one hop between source node and destiny node. Since every node belongs to different domain, and there is no central authority involved, cooperation is essential in this environment of absence of infrastructure. However, there is a non negligible cost of energy and bandwidth for each node when they engage themselves to helping the networks operate. Nodes in Ad Hoc networks as a rational economic man will always try to maximize their own interests, so they all have a motivation to refuse to serve others or not to participate routing scheme.The node which is reluctant to rely other’s packets is called selfish node, they can seriously damage the regular function of networks and even paralyze the whole system.How to design the incentive mechanism so as to promote the prevalence of cooperation has been a hot spot in this field for recent years. My innovation lies in exploring the way to boost cooperation by virtue of evolutionary game instead of making use of external entities like reputation or virtual currency. The forwarding process has been modeled as games in mean-field population structure, and we conduct a series ecological analysis to compare the robustness of various strategies. As a result, strategy Gradual is the most efficient one in terms of promoting the emergence of cooperation under the circumstances of noiseless, but this strategy require to record all the history of game which surely will aggravate the load and delay of the networks. After adding several levels of noise, GTFT strategy takes the spotlight for its top notch and stable performance against other strategies. Since noise is inevitable in real world GTFT is the most appropriate strategy to apply to Ad Hoc networks. Even though TF2T also contains the nature of generosity, according to the outcome of our simulation, strategy with randomness is preferred by nature selection. Grim trigger as a subgame perfect Nash equilibrium can’t stand the pressure of evolution, this also illustrates that deduction isn’t able to substitute the affect of simulation with regard to cooperation theory. Pavlov which can withstand the negative influence of noise is easy to be exploited by ALLD, therefore it doesn’t have sufficient robustness. Then we extend my research objection to the realm of static wireless sensor network which is modeling as a spatial iterated prisoner’s dilemma with neighbor structure of Moore, we simulate the evolution process of five strategies, and come to the conclusion that the reason why communication which is confined to local relationship can enhance the level of cooperation is that cooperative strategies are able to cluster, inside the cluster they cooperate with each other to receive higher payoff, this payoff is high enough to compensate their lost from defending themselves against defective intruders. In the spatial circumstance of a little noise and mutation, Pavlov is proved to be the optimal strategy which not only is able to correct a mistake rapidly by his own, but also not troubled by neutral drift like TFT.