Research On Dynamic Task Scheduling Method In Cloud Manufacturing Environment Based On Complex Network

In the cloud manufacturing environment,a manufacturing task is usually decomposed into multiple sub-tasks and then completed by multiple manufacturing resources.Considering the huge number of manufacturing tasks,multiple sub-tasks may choose the same manufacturing resource,while the manufacturing resource providers will accept multiple subtasks if conditions allow.Each manufacturing resource provider is an independent individual with autonomous capabilities,which can independently arrange the processing sequence of sub-tasks but also be affected by the cooperative enterprises.Therefore,how to reasonably arrange the processing sequence of numerous sub-tasks in multiple manufacturing resources,while ensuring the processing efficiency of all manufacturing resources,is an urgent problem in the cloud manufacturing.Based on the complex network dynamics and its outstanding performance in network structure analysis,an effective solution for minimizing the total completion time of manufacturing tasks in the cloud manufacturing environment is proposed in this paper in the case of random arrival of manufacturing tasks.The main research contents are as follows:(1)Construct the complex network for scheduling object and abstract efficient rules from it.The complex network of all sub-tasks is constructed based on task constraints and resource constraints,then the task scheduling problem is transformed into the node traversal problem on the complex network.Analyze the node degree,node clustering coefficient and node task attribute,then propose dispatching rules based on task attributes,dispatching rules based on degree and task attributes,and dispatching rules based on clustering coefficients and task attributes.Compare the proposed three kinds of dispatching rules with the dispatching rule based on the complex network by experiments.(2)Based on the complex network,improve the ant colony optimization algorithm to make it suitable for solving the task scheduling problem in the cloud manufacturing environment.Firstly,according to the node degree,optimize the state transition rule of the ant colony optimization algorithm,and combine the node's degree with node's task attribute as node's heuristic information to guide the ant colony to search efficiently.Secondly,optimize the parallel strategy of ant colony optimization algorithm.Each ant in the same group is responsible for the arrangement of processing sequence for all sub-tasks of a manufacturing resource,and they have different search spaces but communicate with each other to simulate the autonomy of each manufacturing resource in the actual environment and the correlation between resources.Multiple groups of ant colonies search at the same time to improve the convergence rate.(3)Adopt predictive reactive dynamic scheduling model to decompose the dynamic scheduling in cloud manufacturing environment into multistage steady-state scheduling.Given manufacturing tasks' characteristic of random arrival,propose a hybrid rescheduling drive mechanism based on the number of new tasks to improve the stability of the hybrid driver when the disturbance is too frequent.Generate rescheduling scheme by fully rescheduling strategy and improve ant colony optimization algorithm.Finally,verify the feasibility and effectiveness of the rescheduling driving mechanism and rescheduling algorithm proposed in this paper in solving dynamic task scheduling problems in cloud manufacturing environment by simulation experiments.