Exploring agent learning process by using mechanical features in agent-based simulation

In this paper, we explore the learning process of agents by finding and utilizing their mechanical features. Many researches show that agent's learning mechanism always presents phenomena against the world's natural tendency to disorder, which is known as the entropy law. This indicates that there exists energy transmission between agent and its environment. Thus by investigating energy features inside the agent's decision-maker, we successfully visualize and analyze the learning process of agent. Then we explore energy transmission, and give an explanation of the casual relationship between model of agent and behavior of agent-based system, so as to overcome the corresponding long-existing problems in construction and analysis of agent models, as well as agent-based systems. Application to a practical multi-agent system is also demonstrated to support our arguments.