Study On AGENT Based Simulation Of Queuing System

In recent years, along with the development of computer science, computer simulation receives the scholars' attention more and more as the new method of research. It may imitate and show various conditions of operation, give suggestion and reference information accordingly. Among so many methods, ABMS is an effective measure. Based on the mentality of Agent, ABMS simulates the complex systems, and by describing the actions and interaction among the agents receives the action of the complex systems.Based on ABMS a simulation model of queuing system was developed. The customer, queue and server were abstracted as different agents in the simulation model, and a management agent was advanced. The situation of queuing was simulated through the interaction among agents . The queue agent was the ligament between the customer agent and the management agent. The management agent managed the server agent, and in charge of arranging task. The server agent was supervised by the management agent, and providing service to the customer agent. In addition, the server agent decided whether to adjust it' s type of servers or not under the dispatch of management agent. The customer, queue and server were abstracted as different agents in the simulation model, and a management agent was advanced. The management agent not only reduced the interaction between the customer and the server but also coordinated the process of queuing effectively.Large queuing system contains many servers, and can supply several types of servers. Customer agents choose different queues to wait for services. Different tasks need different times to be completed. At the same time the numbers of asking for every service are same. So on some or other time, some queues are full some queues have no customer. In order to avoid the phenomenon, an algorithm for adjusting the type of servers dynamically was proposed in the model. The main ideology of the algorithm was so. After completing the current task, the server agent sent the type of itself to the management agent. Then the management agent informed the corresponding queue agent to compare the numbers of corresponding servers and the customers waiting in this queue. If the number of customers waiting was less than the number of corresponding servers, the management agent informed the server agent to prepare to adjust it' s type. At the same time the management agent checked out the type of the queue agent whose waiting customers were most and sent the type to the server agent to adjust. The algorithm can reduce the waiting time of customer and improve the work efficiency.The research model selected the Repast platform and simulated the bank queuing. The experiment and analysis show the system can simulate the situation of bank queuing actually, and it can supply references to reduce the waiting time of customer and improve the work effectively in bank. But the experiment supposed the server time of same type was uniform, the fact didn' t so. For example the time for drawing deposits 100 yuan was different to the time for drawing deposits 10000 yuan. So I will go on researching the simulation model which server time can be adjusted actively.