Study Of Trust Problems Between Agents During Cooperation

In open environment there is no central control over agents' behaviors, and agents in such systems are driven by self-interests. In open environment traditional security methods such as encryption and identity authentication cannot always solve all the problems in the procedure of cooperation between agents, such as the collaborator's low capacity, deception and selfishness. In order to solve these problems, a new research area the trust between agents is attracting more and more attention.This dissertation does some research in trust of multi-agents system in open environment. Its background is cooperation between agents in open environment. And its focus is agent's trust description in quantity.First this dissertation comprehensively analyses the new problems aroused from the agents' cooperation in open environment. And this dissertation deems these problems can't be solved by traditional security methods.Second this dissertation develops a relation web model, and based on this model this dissertation develops a trust model between agents. This trust model begins from the definition of trust, and classifies trust from different views. Firstly this trust model gives some explaining of valuing direct trusts, and gives the computing formula of different aspects of direct trusts. Secondly this model explains the conception of valuing recommendation trusts and indirect trusts, and take the recommendation trusts as the experience adopting coefficient which is used to describe the different adopting degree of experience recommended by other agents with different trust levels, and define the indirect trusts formula which confine to the general facts of experience passing. Furthermore the trust model gives a solution to repetitive adoption problems of recommendation trusts in indirect trusts. In order to express different agents' characteristics when they evaluate the trust, this paper puts forward the conception of trust policy and improves the flexibility of trust valuing model with parameters configuration. Finally this dissertation brings forward an infrastructure design of cooperation trust model between agents based relation web, and gives a layered trust-reasoning model.As what have been said above, this paper aims at the trust security problems of multi-agents systems, and puts forward a trust model based relation web imitating the establish procedure of the trust and reputation system in human society. In this trust model the definition and reasoning of trust is based on quantity, so this model is not difficult to implement by software, and have some applied value in multi-agents systems.