An Agent Deception Model Based On General Benefit Optimization

Agent optimization aims at improving the performance of Agent for efficiently achieving Agent's goal. The researches for Agent optimization are classified based on different object as follows: Agent individual optimization and optimization for Agents of collectivity, most of which are focused on the former. And the majority on optimization for Agents of collectivity is working at improving competence of cooperation between Agents using trust and reputation model.Deception, as the opposite of trust and reputation, always takes negative effect. So the purpose researching on deception is counter deception, aiming at preventing it from taking negative effect. While with the further researches, deception is to prove that it can be used as a method to take positive effect if used properly. The thesis brings forward a deception model based on general benefit optimization. And for further validating the model, a new test environment for testing deception between Agents called E-GOLEM are proposed, which is extended from GOLEM and implemented using ABLE, a plug-in of Eclipse. The main contents of the thesis can be summarized as follows:1. A deception model based on general benefit optimization is proposed. When the premise in the model are satisfied, the deception can be used for general benefit optimization, which can be used to different goals based on different situations.2. A new Agent deception test environment called E-GOLEM is proposed, which is extended from GOLEM, aiming at test the deception model proposed in the thesis.3. An E-GOLEM experimental environment is implemented by using ABLE (Agent Building and Learning Environment) which is a plug-in of Eclipse.