Research On Long/Short-term Utility-aware Multi-task Scheduling Optimization In Cloud Manufacturing

With the continuous development and integration of emerging information technology and advanced manufacturing technology,the traditional manufacturing mode is gradually transformed into intelligent and convenient networked manufacturing mode.As a new customer-centric,service-oriented,demand-driven,and knowledge-based manufacturing network,cloud manufacturing comes to being with this background.In the cloud manufacturing platform,large-scale and distributed manufacturing resources offered by different providers are encapsulated into cloud services through virtualization technologies.When customers submit their individual requirements as manufacturing tasks to a cloud manufacturing platform,the centralized management and operation of manufacturing services will enable the platform to deal with multiple tasks at the same time.In most cases,these tasks have a complicated structure that must be decomposed into several subtasks.Each subtask has its own candidate service set with different quality of service values.Therefore,how to schedule multiple manufacturing tasks,i.e.,to allocate available manufacturing services to fulfill a set of specific tasks,is a key issue in cloud manufacturing.In the past decade,the cloud manufacturing multi-task scheduling problem has attracted increasing attention from researchers.However,most of existing studies have concentrated on achieving optimal system performance by dealing with all tasks indistinguishably,while neglecting the individualized requirements of different customers.In fact,with larger scale manufacturing services and their varied configurability,the requirements of customers in a cloud manufacturing platform are more and more individualized.The requirements submitted by different customers vary in terms of the objectives,preferences,and constraints.Moreover,the existing studies only have focused on satisfying the requirements of customers,but neglected the interests of providers.This may result in the dissatisfaction and disappointment of providers over time.In the worst-case scenario,they might even withdraw from the platform.To overcome the above-mentioned limitations,this study investigates the long/short-term utility-aware multi-task scheduling method in cloud manufacturing from the perspective of cloud manufacturing platform development.Considering the preference of the platform at different stages,this study also incorporates two concepts of short-term utility and long-term utility.Firstly,a short-term utility-aware multi-task scheduling model aiming to maximize the benefits of individual tasks simultaneously is proposed,which contributes to cultivating a certain customer group for the early platform development,so as to maximize the short-term utility of the platform and promote the rapid development of the platform in the short term.When the cloud manufacturing platform has a stable customer group,to maximize the long-term utility of the platform and promote the long-term stable development of the platform,a longterm utility-aware multi-task scheduling model aiming to balance and sustain the interests of all stakeholders involved to encourage their continuous participation is proposed.The main contributions of this study can be conducted as follows:(1)A short-term utility-aware multi-task scheduling model is proposed.In the proposed model,three non-functional indexes of quality of service(QoS),i.e.time,cost,and reliability,are used as the optimization objectives for individual tasks.The different preferences and constraints defined by each customer for the objectives are also taken into account.This model aiming to maximize the benefits of individual tasks simultaneously is formulated as a multi-task optimization problem,which contributes to cultivating a certain user group for the early platform development,so as to maximize the short-term utility of the platform and promote the rapid development of the platform in the short term.An extended multifactorial evolutionary algorithm(EMFEA)that uses a unified random key representation scheme with a new decoding method,integrates new genetic mechanism,and employs level-based selection mechanism and fast nondominated sorting is proposed to solve this model.The results obtained from the simulation experiments confirm the effectiveness of the proposed model as well as the proposed algorithm in solving the model.(2)A long-term utility-aware multi-task scheduling model is proposed.In the proposed model,three non-functional indexes of QoS,i.e.time,cost,and reliability,are the criteria to evaluate the overall interest of customers,and the workload balancing index is used to evaluate the overall interest of providers.This model aiming to balance and sustain the interests of all stakeholders involved to encourage their continuous participation,which contributes to maximizing the long-term utility of the platform and promoting the long-term stable development of the platform.In addition,an extended non-dominated sorting genetic algorithm-II with a new three-dimensional representation scheme,a new crossover operator,and a new local search strategy is proposed to solve the presented model.Game theory is employed to recommend an optimal solution to the cloud manufacturing platform from the approximate optimal Pareto solution set.Simulation experiments are conducted to verify the effectiveness of the proposed algorithm by comparing it with three baseline multi-objective evolutionary algorithms.