The paper constructs multi-agent behavior decision model depending on the basic frame of cellular automata (CA) pedestrian flow. And Combine the passenger behavior with the decision of administrative department, basing on multi-agent simulation method of complex system theory, considering each advantage of CA and multi-agent, with the entrance of administrative departments and passenger behavior. In the passenger organization model involving in different kinds of agents such as passenger organization and passenger, which combines multi-agent with CA, the environment factor is expressed by CA. Individuals respond to the changes of environment around through the interaction between resource and the environment or negotiating with another individual. Through simulating the complex process of decision making of administrative departments, it brings forward the study way and the basic frame of the passenger organization optimization during rush hours of large-scale passenger station on complex adaptive system(CAS), constructs the optimization system of passenger organization during rush hours based on multi-agent, discusses the structure and the competing and cooperating relationship between each agent, and designs the optimization arithmetic based on evolutionary algorithm(EA) of multi-agent. At last, taking the optimization of passenger organization during rush hours at Guangzhou Station as the example to analysis and getting a good effect, this can offer the reference to the decision making of administrative departments.

The evolution of logistics industrial cluster reflects the self-organization mechanism and function of the complex system, which provides an analytical angle for the logistics industrial cluster. The multi-agent simulation method of the complex system theory is adopted to explore the general evolutional pattern of logistics industrial cluster and a basic idea for logistics industrial cluster based on multi-agent systems is proposed. This paper also inquires into the structure of each agent in this system and the corporation among all the agents, poses the simplified simulated model and studies the method to implement this model, offering a theoretical guide and a methodological basis to the study of logistics industrial cluster based on multi-agent systems. According to this theory, the simulated experimental system is constructed based on SWARM, and simulated results confirm the rationality of the system.

The problems of conditional entropy's definition and the formula to compute conditional entropy are analyzed from various perspectives, and the corrected computing formula is presented. Examples are given to prove the conclusion that conditional entropy never be increased is not absolute, thus the representation that information is to decrease uncertainty in the definition of information is not absolutely correct.