An agent architecture for competitive application environment

Multi-agent technology has recently become one of the promising technologies used in various complex applications. The use of the concept of multi-agent system can increase productivity of organization both in cooperative and competitive environments. In complex environments, there could be a large number of agents working together. When working with different types of agents, the agents may not know whether they are dealing with teammates or opponents. For the agents, error may easily occur unexpectedly in such a complex situation. This is simply because an unsecured or equivocated environment may be intentionally generated by a type of agent referred to as "a lying agent". This is especially true in the competitive environment where the lying agent needs to achieve its goal and resources are very limited. With such possibility, there is a clear potential benefit for ones who intend to apply the technology to understand how the lying agents could behave. In this paper, by using a special kind of auction existing in the real-world as our problem domain, we exploit an agent architecture and also modeling technique used by the lying agent. The experimental result has shown that the lying agents can efficiently achieve their goals while leaving others to only complete their tasks, but not with what the others had intended to do originally.