Study Of Path Planning Method For Migrating Instance Based On QoS And Q-Learning

Agent refers to automatic software or hardware entity in artificial intelligence domain,it’s an entity with autonomy,reactive,proactive and social characteristics.With the deep and rapid development of Internet application and computer technology,the concept of mobile agent is proposed and widely used. The mobile agent is a special kind of agent,and mobility is its most important characteristics. It can move form one computer to another through the Internet and execute the tasks it carries using the local services or resources,child agent which has the same characteristics as its parent can be derived if needed.The other main superiorities of mobile agent are heterogeneous and asynchronic,distributivity and parallelism,lower network communication cost,intelligent routing and so on.Mobile agent is widely used in domain of information retrieval,distributed computing and electronic commerce,users expect to get the best service over the entire Internet and make the Internet as a whole.According to the characteristics of mobile agent,migrating workflow technic is the combination of mobile agent technic and workflow technic.Migrating workflow is an important application of mobile agent.The mobile agent in migrating workflow is called migrating instance.The workflow technic can achieve automaticly executing workflow entirely or partially.The path planning of migrating instance completing given business process is a important research direction in migrating workflow,it is also the critical to improve the performance of migrating workflow system., and it can improve the flexibility of workflow management system.The path planning problem of migrating workflow within QoS constrain is a kind of constrained path planning,migrating instance selects workplace according to the QoS constrain.The path planning method can be divided into static planning and dynamic planning,static planning method requires entire environment model,the migrating path is given before dispatching mobile agent,and it only need to migrate according to the path.Mobile agent have to adopt the dynamic path planning method in uncertain environment,and don’t decide the migrating path before dispatch.This paper proposes the dynamic path planning method based on reinforcement learning and QoS,mobile agent acquire the environment information by on-line learning based Q-learning algorithm.The path planning of mobile agent problem is the process of selecting workplace and utilizing its local service,that is to say,it is a problem of workplace selection.This paper combines the web service selection model with mobile agent’s path planning problem,and model the path planning problem as workplace service selection,the workplace evaluation method is given based on QoS of workplace service.For the business process with QoS constrain,this paper proposes the solution model and method,converting this problem into workplace service selection with QoS constrain. The path planning problem can be completed by the service selection method based on QoS and Q-learning,so the solution of business process can be obtained.