Research on the optimal combination and scheduling method of crowdsourcing members in a cloud design platform

Abstract

Cloud design has become a potential emerging design service mode. More and more design service providers are clustered in a cloud design platform to become crowdsourcing members. Collaborative tasks published by the platform are usually performed by several crowdsourcing members, and the collaborative mode of crowdsourcing members is an asynchronous collaborative working mode. However, due to the variety of crowdsourcing members in terms of ability, time, and so on, how to combine crowdsourcing members into a rational crowdsourcing team has become a challenge. At the same time, due to the looseness and uncertainty of their collaborative mode, there is no effective guidance mechanism in the cooperation process. Therefore, we propose a method of crowdsourcing the optimal combination and scheduling of members. By evaluating the comprehensive abilities of crowdsourcing members, the method optimizes the selection of suitable crowdsourcing members for each subtask and combines them into a crowdsourcing team. At the same time, besides establishing a collaborative workflow simulation model to plan the timing of subtasks and the workflow of crowdsourcing members, the method also optimizes and schedules crowdsourcing members to process different subtasks in turn. Finally, in the case of the collaborative design task of a medical analgesia pump, this method is used to establish a rational crowdsourcing team for collaborative tasks and effectively plan the collaborative work process of crowdsourcing members. The results demonstrate that the proposed method is feasible and effective.

Keywords

Cloud design platform crowdsourcing service optimal combination optimal scheduling workflow simulation

Introduction

With the rapid development of computer technology and design service modes, new technologies, such as cloud service, Internet of things and big data, have been integrated into product design services. In addition, crowdsourcing and other innovative collaboration modes are also closely integrated with product design services.¹ Cloud design has become an innovative product design service mode. The cloud design platform not only is supported by new technology but also combines the models of innovation and cooperation. It is an extension and transformation of the former product design service model.^2,3

The cloud design platform supports not only the sharing of distributed and heterogeneous design resources but also the crowdsourcing members to complete the design task collaboratively. Crowdsourcing members can provide design services of the entire life cycle of product development, including product design, structural design, simulation analysis, production, sales, and maintenance. They gathered in the virtual platform through the network, and the cloud design platform publishes design tasks according to user requirements. The entire product development process is encapsulated as cloud services through task decomposition, crowdsourcing composition, and optimal scheduling. Then, the platform feedbacks the final solution to users, namely those who do not need to participate in the cooperative task of crowdsourcing members.² At the same time, the service content of the cloud design platform has diverse characteristics, which means it can provide users with overall services for product development and can also provide individual service according to user requirements, such as product structure design or quantity production. The innovation service model of the cloud design platform is illustrated in Figure 1.

Figure 1.

Innovation service model of the cloud design platform.

The crowdsourcing members who provide design services are clustered in the cloud design platform through the network, and the networked crowdsourcing team is built according to the task temporarily published by the platform. With the release and completion of the task, the group members organize and dissolve the team. By sharing knowledge and complementing advantages, the crowdsourcing members carry out the product research and development (R&D) tasks collaboratively.⁴ However, the crowdsourcing members are diverse and dispersed. Also, there are characteristics, such as asynchrony of working time, difference in professional ability, and flexibility in cooperation.⁵ When the cloud design platform releases its task, it needs to select the suitable members from a large number of crowdsourcing members to form a crowdsourcing team with them. In addition, it is necessary to manage the collaborative process effectively so as to improve the collaboration efficiency of the crowdsourcing team. Therefore, the optimal combination and scheduling method of crowdsourcing members is an important part of the innovative service mode of the cloud design platform and plays an important role in the implementation of service modes.

The rest of this article is organized as follows. In section “Related works,” the relevant literature are discussed. In section “Optimal combination method of crowdsourcing members,” the optimal combination method of crowdsourcing members is proposed, and the comprehensive ability model and calculation method are given. In section “Optimal scheduling method of crowdsourcing members,” the optimal scheduling method of crowdsourcing members is proposed, and the simulation model and steps are given. In section “Case study,” a case study is presented and discussed. In section “Conclusion and future works,” the conclusions and future work are presented.

Related works

Currently, various kinds of research are being conducted on the optimal combination and scheduling methods of network members. The existing research focuses on the following three aspects: optimization strategy, member optimization index system, and workflow modeling.

In the study of optimization strategy, there are mainly three strategies for implementing service composition as follows: local optimization strategy, global optimization strategy, and hybrid optimization strategy. First, regarding the aspect of local optimization strategy, Zeng et al. proposed a middleware platform to select Web services based on user satisfaction. In addition, Quality of Service (QOS) is used as the objective function to meet the composite service structure which the users set.⁶ Also, in previous studies, centrally optimized the basic service from candidate service sets corresponding to each subtasks and eventually formed a service portfolio that meets the constraints of each subtask.^7,8 Liu and Li⁹ constructed a multi-level resource virtualization framework for cloud services encapsulation and presented a multi-granularity resource classification algorithm, which can flexibly realize resource combination. This local optimization strategy is superior to other strategies for the optimization of time performance, but it has a defect when dealing with the overall QOS optimization. Second, regarding the aspect of global optimization strategy, Ardagna and Pernici introduced a new model to solve the choice problem of Web service. The model defines the Web service selection problem as a mixed integer linear programming problem. Considering the constraints of Web services, they used the method of cyclic stripping to realize the service composition in the process of optimization.¹⁰

In the study of the member optimization index system, the methods of selecting members in the existing literature are mainly from the angles of qualitative analysis and quantitative analysis. In terms of qualitative analysis, Wang et al.¹¹ put forward a mathematical model based on integer programming team organization. In the model, they took into consideration the technical, innovative, and management abilities of the developer. The team organization problem is decomposed into several independent subproblems, and the tabu search algorithm is used to solve these subproblems.¹¹ In the selection of members, Quélin¹² established the member optimization index system based on honor, ability, cooperation, and motivation. Regarding the quantitative analysis, Liu et al. proposed a member selection method based on the comprehensive performance information of the members. Considering the personal capabilities and collaborative capabilities of the members, they constructed a member information index system and a multi-objective optimization model with Pareto algorithm.¹³ Feng et al.¹⁴ developed a team member selection method based on collaborative information and used a quadratic programming method to solve the model. Papakostas et al.¹⁵ used specific criteria to select the most appropriate network partners in a dynamic manufacturing environment (DMN). The cost, time, and quality were important affecting factors in the selection process. Cao et al.¹⁶ established a manufacturing service ability evaluation model and quantitatively researched the types and capabilities of the service providers.

In the study of workflow modeling, Wang et al. studied the workflow from the perspective of consistency calculation. The conversion relationship based on event relationships was established through the analysis of the known event correspondences in the workflow, and different correspondences between events in the workflow were expressed by using the relational behavior matrix.¹⁷ In the research of design task decomposition and resource allocation, aiming at the defects in the priority assignment of design tasks, Yang et al.¹⁸ used a Petri net to realize the simulation modeling of resource allocation. Framinan et al.¹⁹ extensively reviewed and classified the studies of scheduling problems and gave some conclusions and directions for future research. Gyulai et al.²⁰ established a new scheduling simulation model and solved the scheduling scheme through constraint programming and genetic algorithm. Lee²¹ proposed a branch-and-bound algorithm to solve the two-stage assembly scheduling problem and developed four efficient heuristic algorithms to solve the optimal scheduling scheme. Cheng et al. introduced a general service scheduling model for cloud manufacturing resources, which is a comprehensive utility model. Their model considered energy consumption, cost, and risk for the provider, consumer, and operator.²² Lin et al.²³ proposed a multi-center management architecture based on a two-level scheduling strategy and provided an optimization model of manufacturing resources and capacity allocation.

In summary, the existing combination and scheduling methods of members are mainly applied in the traditional network environment. However, there are few related studies on the cloud design platform environment. Although the cloud design platform is also a network environment, the related studies by Mourtzis and Vlachou as well as those by others show that the cloud environment includes technologies, such as big data support, real-time operation, and configurability, which promote the development of the cloud platform architecture and services. It is an extension and transformation of the network environment.^1,2,24 Therefore, the differences between cloud environment and network environment are as follows: (1) cloud environment includes more new technologies and innovative service modes; (2) the cloud environment is transforming the conventional product-oriented design and manufacturing business model into a service-oriented one; and (3) cloud environment can provide users with product design/manufacturing services that are ready-to-use, on-demand, and of high quality and low cost. Accordingly, in the environment of cloud design platform, how to combine crowdsourcing members into a rational crowdsourcing team and rationally plan the collaborative process of crowdsourcing members has become a challenging task. Therefore, this study focuses on the comprehensive ability evaluation and collaborative workflow planning of crowdsourcing members and proposes a method of crowdsourcing members optimal combination and scheduling.

Optimal combination method of crowdsourcing members

Task analysis

Collaborative tasks are usually broken down into several subtasks to reduce the difficulty of crowdsourcing members working together. However, the number and type of subtasks will directly affect the collaboration efficiency of the crowdsourcing members. Thus, it is necessary to properly decompose the collaborative tasks. In this study, a previously reported method²⁵ is adopted to establish the collaborative task decomposition system. The system is integrated into the platform, and the collaborative task is decomposed by the platform according to its calculation principles. Task decomposition is an automatic process and involves the following steps: (1) according to the collaborative task demand information, the initial decomposition is performed by combining the function and structure; (2) the preliminarily obtained subtask type is determined; (3) the threshold range of subtask granularity, coupling, and equilibrium is set; and (4) if the granularity, coupling, and equilibrium data of the decomposed subtasks are not within the threshold range, the subtasks are decomposed. The decomposed subtasks usually have four dependencies,^26–28 as shown in Figure 2, which displays the dependencies between two subtasks, where the solid arrows represent the input and output of information between subtasks.

Figure 2.

Subtask dependencies: (a) serial mode, (b) serial coupled, (c) parallel mode, and (d) parallel coupled.

When the collaborative task is decomposed, the subtasks are assigned to the crowdsourcing members. In this study, a previously described method^29,30 is used to construct a subtask assignment system. The system is also integrated into the platform. According to its similarity calculation and semantic matching technology, the platform automatically assigns subtasks. The automatic subtask assignment process involves the following steps: (1) the crowdsourcing members complete their registration in the platform, and the registration information includes basic information, such as the types and professional skills of crowdsourcing members; (2) the platform database records the subtask information that the crowdsourcing members have provided in the past and the evaluation results of the completion status; and (3) through the similarity calculation and semantic matching method, the platform searches for crowdsourcing members with the relevant professional skills and experience.

At the same time, the subtasks have different priorities. If a crowdsourcing member performs two subtasks, the crowdsourcing member needs to first perform a subtask with a higher priority. This study adopts a quantitative method to determine the priority of subtasks and assigns priority to subtasks through the cloud design platform. The steps are as follows: (1) The priorities of subtasks are divided into five levels: {very high, high, moderate, low, very low}, the corresponding fraction is {1, 0.75, 0.5, 0.25, 0}. (2) In the cloud design platform, crowdsourcing members score the priority of subtasks through evaluation tools. The scoring process for crowdsourced members is performed online, and the platform will obtain the score data for subtask priority. (3) The platform builds a prioritized directed graph based on the final score data and uses it to determine the priority of the subtasks. Eventually, the priorities are assigned to the subtask by the platform. The mathematical expression of the prioritized directed graph is a binary group G

G = (S, E)

(1)

where S = (S₁, S₂, …, S_n) represents all the subtasks, E = (E₁, E₂, …, E_n) represents the priority of the subtasks, and the reachability of G is expressed as G = (g_ij), in which

g_{ij} = {\begin{matrix} 1 & If S_{i} can reach S_{j} \\ 0 & False \end{matrix}

(2)

in order to assess the connectivity of the prioritized directed graph.

Collaborative process of crowdsourcing members

The service scope of the cloud design platform covers the whole product life cycle, which can be divided into three stages as follows: the pre-design stage, the mid-manufacturing stage, and the post-sales and maintenance stage. Among them, the pre-design stage includes demand analysis, conceptual design, appearance design, and prototype production. The mid-manufacturing stage includes mold design, engineering design, process planning, and test analysis. The stage of post-sales and maintenance includes sales and maintenance of products. Based on the subtasks, the platform selects the crowdsourcing members from each stage to form a team. In the collaborative working process, when the crowdsourcing members receive the input information of the subtasks, they need to submit the expected working time, complete the subtasks within the specified time, and output the feedback information to the platform. If several subtasks need to be completed at the same time, then the crowdsourcing members must complete them in turn according to the priority relationship of the subtasks. In general, the collaborative working mode of crowdsourcing members is the combination of online and offline work, and the transmission of information about the subtasks is completed on the platform, while the specific work content is completed offline.

Crowdsourcing members optimized selection strategy

The cloud design platform retrieves numerous crowdsourced members according to the subtask distribution system. The optimization strategy needs to evaluate the execution ability, innovation ability, busyness, and relative importance of the subtasks of crowdsourcing members, and after optimization the appropriate crowdsourcing members are selected through the comprehensive competency evaluation model, as shown in Figure 3.

Figure 3.

The optimization combination model for crowdsourcing members.

Based on the abovementioned analysis, the optimized selection strategy of the crowdsourcing members of the cloud design platform is as follows.

Construct the execution ability matrix of the crowdsourcing members and determine the execution ability of members for all subtasks.

Construct the degree of busyness matrix of crowdsourcing members to determine the effective working time of the crowdsourcing members in the execution of each subtask.

Construct the innovation ability matrix of the crowdsourcing members and determine the innovation ability of crowdsourcing members for all subtasks.

Construct the degree of relative importance matrix of subtasks to determine the relative importance of each subtask.

The crowdsourcing members are selected after optimization through the comprehensive ability evaluation model, and the crowdsourcing team is established.

Comprehensive ability evaluation mode

A comprehensive ability evaluation model is established according to the optimized selection strategy of crowdsourcing members. The model includes quantitative evaluation of the execution ability, busyness degree, innovation ability of crowdsourcing members, as well as the relative importance of the subtasks. The matrices of the execution ability, busyness degree, innovation ability, and subtask relative importance, respectively, are established.

Due to the differences in the types of services offered by crowdsourcing members and their abilities, it is necessary to evaluate the execution ability of the crowdsourcing members to perform all subtasks. The coefficient of execution ability is shown in Table 1. The coefficient acquisition mechanism is as follows: (1) After each subtask, the crowdsourcing members receive an evaluation of their execution ability. The evaluation level is {amateur, beginner, intermediate, advanced, expert}. (2) The score corresponding to the evaluation grade is {1, 2, 3, 4, 5}. The final score is a weighted average (integral part) of the total score, and the value of the final score represents the level of the executive ability. (3) Corresponding to Table 1, the crowdsourcing members receive the corresponding coefficient according to the execution ability level. Thus, establishing the execution ability matrix C of crowdsourcing members

C = [\begin{matrix} c_{11} & c_{12} & \dots & c_{1 j} \\ c_{21} & c_{22} & \dots & c_{2 j} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ c_{i 1} & c_{i 2} & \dots & c_{ij} \end{matrix}]

(3)

where 0 ≤ C _ij ≤ 1, when C _ij = 0, which means that the crowdsourcing member i does not have the ability to complete task j. When C _ij =1, the member i is an expert in task j.

Table 1.

Five-scale evaluation coefficients of the matrices.

Matrix	Five-scale evaluation coefficients
Matrix	Five-level	Four-level	Three-level	Two-level	One-level
C	1	0.8	0.6	0.4	0.1
B	1	0.8	0.6	0.4	0.2
H	0.9	0.7	0.5	0.3	0.1
E	1	0.8	0.5	0.4	0.3

Due to the flexibility and uncertainty of the effective working time of crowdsourcing members, the busyness degree represents the effective working time that crowdsourcing members may have in the future. At the same time, more subtasks are likely to be performed by capable crowdsourcing members. Therefore, the busyness degree of crowdsourcing members is ambiguous. It requires crowdsourcing members to self-evaluate and submit busyness degree status to the platform. In order to measure the busyness degree of the crowd members more accurately, the busyness degree is divided into five states and levels: {freest (1 level), freer (2 level), busy (3 level), busier (4 level), busiest (5 level)}. In order to standardize the calculation of the comprehensive ability of the crowdsourcing members, the values of each level of busyness degree are shown in Table 1. Thus, establishing the busyness degree matrix B of crowdsourcing members

B = [\begin{matrix} b_{1} \\ b_{2} \\ ⋮ \\ b_{i} \end{matrix}]

(4)

where 0 ≤ b_i ≤ 1, when b_i = 0, the crowdsourcing member b_i is the freest, when b_i = 1, the crowdsourcing member b_i is the busiest.

Crowdsourcing members have different innovation abilities for different subtasks, so it is necessary to evaluate the ability of crowdsourcing members to innovate for every subtask. The innovation ability coefficient is shown in Table 1. Its acquisition ability mechanism is similar to that of its execution ability. Thus, establishing the innovation ability matrix H of crowdsourcing members

H = [\begin{matrix} h_{11} & h_{12} & \dots & h_{1 j} \\ h_{21} & h_{22} & \dots & h_{2 j} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ h_{i 1} & h_{i 2} & \dots & h_{ij} \end{matrix}]

(5)

where 0 ≤ h_ij ≤ 1, when h_ij = 0, it means that the crowdsourcing member i does not have the innovation ability for task j. When h_ij =1, the member i is an innovative expert in task j.

In collaborative tasks, the relative weight of each subtask is different. Thus, the platform needs to evaluate the relative weight of each subtask. The coefficient of the relative weights is shown in Table 1. Their acquisition mechanism is similar to that of the execution ability. The difference is that: (1) the evaluation level of the relative weight of the subtasks is {lower, low, middle, high, higher} and (2) the relative weight of the subtasks is evaluated by the crowdsourcing members participating in the collaborative task. Thus, establishing a subtask relative importance matrix E

E = [\begin{matrix} e_{1} \\ e_{2} \\ ⋮ \\ e_{i} \end{matrix}]

(6)

where 0 ≤ e_i ≤ 1. The e_i denotes the relative importance of the subtasks. When e_i is compared with e_j, the greater the relative coefficient, the higher the relative importance.

In the four matrices of the execution ability, busyness degree, innovation ability and mission importance of the crowdsourcing members, each element has fuzziness. Thus, the five-level scale is established for quantitative analysis, and the evaluation coefficient of each matrix is obtained as previously described,³¹ as shown in Table 1.

By evaluating the tasks completed by crowdsourcing members, we obtain the evaluation data of the executive capacity and innovation ability of crowdsourcing members. Then, the busyness degree of the crowdsourcing members is submitted by themselves to the cloud service platform. The relative importance ratings of subtasks are scored by all the crowdsourcing members participating in the task, and the evaluation data are fed back to the cloud design platform, and eventually the relative importance of each task is obtained based on the collected data. After obtaining the evaluation data of the execution ability, busyness degree, innovation ability, and relative importance of the modular tasks of the crowdsourcing members, the trend matrix TR is established

TR = [\begin{matrix} t r_{11} & t r_{12} & \dots & t r_{1 j} \\ t r_{21} & t r_{22} & \dots & t r_{2 j} \\ ⋮ & ⋮ & ⋱ & ⋮ \\ t r_{i 1} & t r_{i 2} & \dots & t r_{ij} \end{matrix}]

(7)

where tr_ij =c_ij – b_i + h_ij + e_i, which generate the matrix O with trend matrix TR , O = (o_ij)_m×n, matrix O _1j = Max(tr_1j). Then, we obtain all the elements of the entire matrix O and obtain the execution matrix D , D = (d_ij)_1×n, where d_1j = o₁_j. The optimal combination scheme of the group members is obtained by the execution matrix.

Optimal scheduling method of crowdsourcing members

The basic definition of time-colored Petri net

The basic Petri net consists of four structural elements as follows: place, transition, arc, and token. The place is generally represented as a circle, describing a state library, as a container of token, which is used to represent different states. The transition is typically represented as rectangle or a short line that describes the change from one state to another, which cannot usually be interrupted. The arc is represented as a section of a directed arc, from place to transition or transition to place. The arc can set the weights, namely the number of a one-time consumption of crowdsourcing members. The token is a network system of crowdsourcing members, where the number of tokens is the number of crowdsourcing members. In the network system, the crowdsourcing members change in continuous flow.¹⁸

The time-colored Petri net used in this study is defined as follows.

A time-colored Petri net is a multi-component group. TCPN = (∑, P, T, A, N, C, G, U, I, W, r₀), where

∑ is a finite set of time class or non-empty time class, that is a time or non-time color set;

P is a limited set of places;

T is a finite set of time or non-time transitions;

A is a finite set of arcs, P ∩ A=P∩T=T ∩ A = ∅;

N is a node function to define A to (P × T) ∪ (T × P);

C is a color function to define P to ∑;

G is a guard function to define T as ∀t ∈ T: (Type(G(t)) = B∧Type(Var(G(t))) ⊆ ∑), B ={true, false};

U is an arc expression function, which is a time or non-time expression function defined in A.

The formula is ∀a ∈ A: (Type(E(a)) = C(P(a)) ∧Type(Var(E(A))) ⊆ ∑), where P(a) is the place of N(a);

I denote an initialization function. It is the time or non-time function defined in P, the formula is ∀p ∈ P: (Type(I(p)) = C(P));

W is a set of time values, which is time identification, a set of non-negative values;

r₀ is the element of the time set of R values, which is the start time.

Simulation modeling of crowdsourcing team scheduling process

Based on the definition of the time-colored Petri net, a scheduling simulation model of crowdsourcing members is established. The basic principle is shown in Figure 4.

Figure 4.

Simulation model of crowdsourcing members scheduling based on the time-colored Petri net.

In the simulation model, the translation X_i is expressed as a subtask T_i scheduling of crowdsourcing members, and the crowdsourcing members perform the T_i subtask. When the T_i subtask is completed, the translation Y_i indicates that the crowdsourcing members have completed the implementation of the regression management library. In order to achieve a unified management and scheduling of the cloud design platform, a crowdsourcing member management repository is built, and all the schedulable crowdsourcing members exist in the management library. In the process of cooperation of crowdsourcing members in the cloud design platform, crowdsourcing members can be described as: color Res = (ID, Name, State, Capacity, Feature), where ID refers to the number of the crowdsourcing members, and the number is unique. Name indicates the name of the crowdsourcing members. State denotes the busyness state of the crowdsourcing members. Crowdsourcing members have both free and busy states. Capacity refers to the executive ability of members. Feature represents the features of the crowdsourcing members used, such as whether they have the innovation ability.

The scheduling of crowdsourcing members can be simplified to: R_i = k_i₁R₁ + k_i₂R₂ + … + k_imR_m (i = 1, 2,3, …, n). The steps of the cloud design platform to optimize the scheduling of crowdsourcing members are as follows.

Step 1. Subtask T_i proposes the requirement of a member application to the system package member of the management library.

Step 2. The system checks whether the scheduled members meet the application requirements of the crowdsourcing package member of the management library, which determines the busyness state of the crowdsourcing members. If the state is free, the system will execute the subtask T_i with the scheduling crowdsourcing members and proceed to Step 4; if the state is busy, the system will not dispatch crowdsourcing members, and the application will wait.

Step 3. If n subtasks simultaneously apply to the system crowdsourcing member of the management library for the same member, then we determine the priority relationship between n subtasks. The system first dispatches the crowdsourcing members with high priority for the subtask.

Step 4. The team members R_i = k_i₁R₁ + k_i₂R₂ + … + k_imR_m (i = 1, 2, 3, …m) are scheduled for the subtask T_i. The transition X_i schedules crowdsourcing members to perform subtasks, the member token in the crowdsourcing member management library decreases accordingly, which means that the crowdsourcing member token subtask T_i’s place P₁ increases.

Step 5. When the task T_i is completed after the execution, the transition T_i proposes the application of the member to the system management database. Transition Y_i returns the crowdsourcing members to the crowdsourcing team management library. The member token crowdsourcing package member management library increases correspondingly, while the crowdsourcing member token place P₁ decreases.

In the simulation model, when the multiple subtasks occupy the same crowdsourcing members, we first use a weighted directed graph to distribute subtasks with high priority to crowdsourcing members, and we need to ensure that all subtasks can be assigned to crowdsourcing members, so as to avoid that the low-priority subtasks plunder high-priority crowdsourcing members. In order to complete different tasks released by cloud design platform, different crowdsourcing members need to form temporary teams. By optimizing the management and control of the crowdsourcing team collaboration process, we can achieve optimal scheduling of the crowdsourcing members.

Case study

In this section, the collaborative design task of a medical analgesic pump is used as an example. The crowdsourcing members are selected after optimization for the collaborative task by the platform and form a crowdsourcing team. Simultaneously, a simulation model for the collaboration process is built by the platform, and the crowdsourcing members are optimally scheduled to handle different subtasks.

By leasing the cloud server provided by Alibaba and simulating the cloud design platform environment, a personal computer (PC) is used to simulate the heterogeneously distributed members of the package. The cloud design platform centrally manages the user needs, design tasks, crowdsourcing members, and collaborative workflow. In addition, crowdsourcing members complete case studies in the environment of network connectivity.

Optimal combination of crowdsourcing members

Users submit the design requirements of the medical analgesia pump to the cloud design platform. The design requirements of the medical analgesia pump mainly include some parts of the design stage and the manufacturing stage, as shown in Figure 3. Through the analysis of the task requirements and the method of task decomposition,²⁵ the cloud design platform decomposes the design task of the medical analgesic pump into 10 subtasks. The 10 decomposed subtasks are numbered, as shown in Table 2.

Table 2.

Subsets of subdivision for the design of the medical analgesic pump.

Number	Subtask	Number	Subtask
A ₁	Structural design	A ₆	Mechanical analysis
A ₂	Mold design	A ₇	Appearance design
A ₃	Reliability analysis	A ₈	Conceptual design
A ₄	Prototype	A ₉	Ergonomics design
A ₅	Process planning	A ₁₀	Usability test

When the cloud design platform completed the decomposition task, according to the function, requirement, and category of the subtask, we used the retrieval method in the document to retrieve the qualified crowdsourcing members. The cloud platform retrieved a total of 10 crowdsourcing members who met the conditions and evaluated the priority scoring of 10 subtasks by these 10 crowdsourcing members in the cloud design platform using priority evaluation tools. The cloud design platform received priority data among the 10 subtasks, in order to establish the prioritized directed graph among subtasks, as shown in Figure 5.

Figure 5.

The prioritized directed graph between subtasks of the medical analgesic pump design.

The cloud design platform searches for a collection of candidate crowdsourcing members and are recorded as R, R = (R₁, R₂, …, R₁₀). The 10 crowdsourcing members can provide services in the pre-design stage and the mid-manufacturing stage, and there are differences in abilities among the 10 crowdsourcing members. The cloud design platform needs to select a group of crowdsourcing members participating in the design task from 10 members. According to the proposed method, the matrices of execution ability, busyness degree, innovation ability, and the relative importance of the subtask are established. Among them, the busyness degree data are submitted by the members themselves, while the executive ability and the innovation ability data are acquired by the tasks of the members. The data of the executive ability, busyness degree, and innovation ability are evaluated by the evaluation coefficients listed in Table 1. The relative importance of the subtasks is scored by crowdsourcing members according to the coefficient shown in Table 1. Finally, the evaluation coefficients of crowdsourcing members are obtained (The data are available from the corresponding author).

The trend matrix TR is obtained by bringing the comprehensive ability and subtask evaluation data into the formulas (3)–(7)

T R = \begin{matrix} [\begin{matrix} A_{1} & A_{2} & A_{3} & A_{4} & A_{5} & A_{6} & A_{7} & A_{8} & A_{9} & A_{10} \\ R_{1} & 1.1 & 0.8 & 1.0 & 1.4 & 0.7 & 0.6 & 0.4 & 0.7 & 0.9 & 0.8 \\ R_{2} & 1.1 & 1.6 & 1.2 & 1.6 & 0.5 & 0.9 & 1.0 & 0.8 & 0.8 & 0.9 \\ R_{3} & 0.9 & 0.6 & 1.2 & 0.7 & 0.9 & 0.9 & 0.8 & 0.2 & 0.9 & 0.9 \\ R_{4} & 0.7 & 1.2 & 1.2 & 1.6 & 1.1 & 0.9 & 0.9 & 0.9 & 0.5 & 0.9 \\ R_{5} & 0.9 & 0.7 & 1.1 & 1.0 & 0.7 & 0.5 & 0.6 & 0.9 & 0.5 & 0.5 \\ R_{6} & 0.7 & 1.6 & 1.0 & 1.6 & 1.1 & 0.7 & 0.4 & 0.4 & 0.6 & 0.4 \\ R_{7} & 0.6 & 1.4 & 0.5 & 1.1 & 0.6 & 0.4 & 1.0 & 1.4 & 0.3 & 0.7 \\ R_{8} & 0.8 & 0.7 & 0.7 & 0.8 & 0.4 & 0.5 & 0.2 & 1.0 & 0.2 & 0.1 \\ R_{9} & 0.2 & 1.2 & 0.7 & 0.8 & 0.6 & 0.2 & 0.5 & 0.4 & 0.2 & 0.4 \\ R_{10} & 0.5 & 0.9 & 0.8 & 0.5 & 0.2 & 0.2 & 0.5 & 0.5 & 0.4 & 0.2 \end{matrix}] \end{matrix}

Using formulas O = (o_ij), matrix O _1j = Max(tr_1j), the TR matrix is converted to the preferred matrix O

O = [\begin{matrix} 1, 2 & 2, 6 & 2, 3, 4 & 2, 4, 6 & 4, 6 & 2, 3, 4 & 2, 7 & 7 & 1, 3 & 2, 3, 4 \\ 5 & 7 & 5 & 1 & 3 & 6 & 4 & 8 & 2 & 1 \\ 3 & 4 & 1 & 7 & 5 & 1 & 3 & 5 & 6 & 7 \\ 8 & 9 & 6 & 5 & 1 & 8 & 5 & 4 & 4 & 5 \\ 6 & 10 & 10 & 9 & 9 & 5 & 9 & 2 & 5 & 9 \\ 4 & 1 & 8 & 8 & 7 & 7 & 10 & 1 & 10 & 6 \\ 7 & 5 & 9 & 3 & 2 & 10 & 6 & 10 & 7 & 10 \\ 10 & 8 & 7 & 10 & 10 & 9 & 1 & 9 & 9 & 8 \\ 9 & 3 & - & - & 8 & - & 8 & 6 & 8 & - \\ - & - & - & - & - & - & - & 3 & - & - \end{matrix}]

Using D = (d_ij)_1×n, where d_1j =o_1j, the execution matrix D is obtained

D = [\begin{matrix} 1, 2 & 2, 6 & 2, 3, 4 & 2, 4, 6 & 4, 6 & 2, 3, 4 & 2, 7 & 7 & 1, 3 & 2, 3, 4 \end{matrix}]

Through the execution matrix D , seven crowdsourcing members were ultimately selected to form the crowdsourcing team of the design of medical analgesia pump. The members of the crowdsourcing team were R₁–R₇. Crowdsourcing members R₈, R₉, and R₁₀ did not participate in the design of the medical analgesia pump.

Optimal scheduling of public crowdsourcing members

According to section “Optimal combination of crowdsourcing members,” the medical analgesia pump design task is decomposed into 10 subtasks, and the dependence analysis between subtasks revealed that all the subtask dependencies and workflow sequence are as shown in Figure 6, in which subtasks 7 and 9 are in parallel relationship. In addition, subtask 3 and subtasks 6 and 10 are also in parallel relationship.

Figure 6.

Subtask relationship and timing.

According to the execution matrix D , the mapping relationship between subtasks and crowdsourcing members are as shown in Table 3. Subtask 8 is executed by the public package member R₇. Subtask 7 is executed by the crowdsourcing members R₂ and R₇, and subtask 9 is executed by the crowdsourcing members R₁ and R₃. Subtask 1 is executed by the crowdsourcing members R₁ and R₂, and subtask 4 is executed by the crowdsourcing members R₂, R₄, and R₆. The crowdsourcing members R₂, R₃, and R₄ are responsible for subtasks 3, 6, and 10. Subtask 5 is executed by the crowdsourcing members R₄ and R₆, and subtask 2 is executed by the crowdsourcing members R₂ and R₆. At the same time, the crowdsourcing members submit their completion time of each subtask to the cloud design platform.

Table 3.

The mapping relationship between the design task of the analgesia pump and the crowdsourcing members.

Number	Subtasks	Finish time (h)	Crowdsourcing
A ₁	Structural design	21	R₁ + R₂
A ₂	Mold design	16	R₂ + R₆
A ₃	Reliability analysis	24	R₂ + R₃ + R₄
A ₄	Prototype	40	R₂ + R₄ + R₆
A ₅	Process planning	18	R₄ + R₆
A ₆	Mechanical analysis	10	R₂ + R₃ + R₄
A ₇	Appearance design	16	R₂ + R₇
A ₈	Conceptual design	24	R ₇
A ₉	Ergonomics design	18	R₁ + R₃
A ₁₀	Usability test	32	R₂ + R₃ + R₄

According to sections “The basic definition of time-colored Petri net” and “Simulation modeling of crowdsourcing team scheduling process,” a simulation model of collaboration process is established based on the time-colored Petri net, as shown in Figure 7. In this model, crowdsourcing members are first stored in the management library. According to the requirement of the submission of the medical analgesia pump design, the system calls the application for the request of the crowdsourcing members to the management library. When the subtask is completed, the crowdsourcing members are returned to the management library.

Figure 7.

The simulation model for the collaborative process of the crowdsourcing members.

In Figure 7, subtasks 3, 6, and 10 are in parallel relationships, and crowdsourcing members R₂, R₃, and R₄ are all required to participate in their execution. The priority relation of the three subtasks can be determined according to the weighted directed graph of the subtasks in Figure 6. Among them, the priority of subtask 6 is the highest, so the group members R₂, R₃, and R₄ will give priority to subtask 6. Subtasks 3 and 10 enter the waiting state after completion of subtask 6 and return of crowdsourcing members. As subtasks 3 and 10 have the same priority, crowdsourcing members R₂, R₃, and R₄ will execute the two subtasks at the same time. In Figure 7, “@+” represents the time required to complete the subtasks, according to the simulation model based on the time-colored Petri net, and the total (R&D) time for the design task of the medical analgesic pump is 183 h.

Comparative analysis

In order to further evaluate the effectiveness of the method proposed in this article, in this section, we further analyze and compare it with other methods proposed in the literature.¹³ In a previously proposed method,¹³ eight crowdsourcing members were selected from the candidate members, namely R₁, R₂, R₃, R₄, R₅, R₇, R₉, and R₁₀, and the total R&D time was 219 h; the mapping relationship between the design task and the crowdsourcing members is shown in Table 4.

Table 4.

Mapping relationship between the design task and the crowdsourcing members (in a previously reported method in the literature).¹³

Number	Completion time (h)	Crowdsourcing
A ₁	30	R₂ + R₄ + R₉
A ₂	14	R₆ + R₁₀
A ₃	25	R₁ + R₂ + R₅
A ₄	50	R₉ + R₁₀
A ₅	15	R₄ + R₆ + R₁₀
A ₆	8	R₄ + R₁₀
A ₇	24	R₂ + R₇ + R₉
A ₈	15	R₁ + R₂
A ₉	20	R ₃
A ₁₀	18	R₄ + R₅

Due to differences between the two methods in the specific solution process, it is not easy to directly compare and analyze them. Therefore, we used a previously reported evaluation method.³² Taking service quality, task cycle, and cooperation satisfaction of the members as objective functions, these two methods were indirectly evaluated by solving the optimal value of the objective function. The initial parameters of the experiment were set as follows: the population size was 60, the mutation rate was 0.05, the crossover rate was 0.7, the maximum number of iterations was 100, and the number of experiments was 50. In the MATLAB environment, the genetic algorithm was used to compare and analyze the two methods, and the results are shown in Figure 8.

Figure 8.

Optimal value of objective function.

The data presented in Figure 8 reveal that the two methods converged to the same objective function value in 50 experiments, so these two methods are feasible and effective. By comparing the optimal value of the objective function, the method proposed in this article has higher initial values, faster convergence, and the final value is closer to 1. These findings show that the method proposed in this article is more advantageous in the case solving process.

In addition, further analysis of the two methods reveals that the previously proposed method¹³ only considers personal ability and collaborative ability as optimization indices when selecting members and does not consider certain factors, such as relative importance of subtasks, task cycle, and busyness of crowdsourcing members. In the process of member selection, this method is prone to fall into a local optimum. Although the convergence result can be obtained earlier, the consideration of global optimization is inadequate. In the process of scheduling members, the scheduling work is mainly completed according to the time sequence of the subtask workflow. However, in view of the parallel tasks, the corresponding scheduling strategy and basis are absent.

The method proposed in this article fully considers the comprehensive ability of crowdsourcing members, relative importance of subtasks, and task cycle. According to the comprehensive evaluation results, on the one hand, more rational members can be selected after optimization to form a team. On the other hand, the evaluation results provide a decision and basis for the collaborative simulation model. This method can rationally plan the collaborative process of crowdsourcing members and improve the efficiency of collaboration. Therefore, the method proposed in this article is more practical and effective in the solution process of the case.

Case discussion

According to the case study, when faced with 10 crowdsourcing candidate sets with diversity and difference, and so on, using the comprehensive ability evaluation model, the proposed method optimized and selected seven crowdsourcing members who met the subtask requirements and grouped them into a crowdsourcing team. Simultaneously, the workflow timing and cycle of collaborative design tasks were clearly planned, and the effective participation of the crowdsourcing members in the coordinated task was optimized. Therefore, the practical significance of this study is as follows: (1) in the cloud design platform, this method can effectively optimize and select the crowdsourcing members who meet the requirements of subtasks and improve the rationality of the construction of the crowdsourcing team; (2) this method can optimally schedule crowdsourcing members to process different subtasks in turn to improve the efficiency of collaborative work.

This study is based on the analysis of the innovative service model of the cloud design platform, the characteristics of collaborative tasks, and the collaborative process of crowdsourcing members. Simultaneously, the proposed method closely combines the rules and characteristics of product life cycle R&D. Therefore, for other cases, although they are different in subtasks, crowdsourcing members, and so on, they share the same process of optimal combination and scheduling, so that the proposed method has a certain universality for different cases. Evidently, more case studies are needed in future work.

There are still additional limitations and uncertainties in this case study. First, the case study is performed in an ideal mode, where there is no design conflict or iterative design in the collaborative process of the crowdsourcing members. However, various types of conflicts and iterative processes are inevitable in the collaborative process. Therefore, further research on conflict resolution is needed in the later stage to improve the proposed method. Second, in this case study, it is uncertain whether the crowdsourcing members are willing to participate in the collaborative design task. Therefore, when facing a new collaborative design platform, an incentive system should be adopted to encourage and attract crowdsourcing members to participate in the collaborative tasks in the platform. Thus, although there are some deficiencies about the proposed methods, the overall train of thought is feasible.

Conclusion and future works

In general, the research method proposed in this article solves the problems existing in the combination and scheduling of crowdsourcing members to support the construction of a cloud design platform system. By analyzing the collaborative design process of crowdsourcing members in the cloud environment, the following two problems were identified: (1) Crowdsourcing members have the characteristics of diversity of subjects, geographical dispersion, asynchronous working hours, differences in professional abilities, and flexibility in cooperation. It is difficult for the platform to find appropriate crowdsourcing members with comprehensive capabilities and to combine them into a crowdsourcing team. (2) The working mode of crowdsourcing members is asynchronous and loose. In addition, it lacks a scientific mechanism for overall planning and management of collaborative workflows.

In order to solve these problems, this article proposes a method for optimal combination and scheduling of crowdsourcing members. The proposed method can effectively form a reasonable crowdsourcing team. In view of the choice of crowdsourcing members with the characteristics of magnanimity, difference, and diversity, using the comprehensive ability evaluation model and calculation method, the crowdsourcing members who meet the subtask requirements can be optimized and selected. Different types of crowdsourcing members can form a rational crowdsourcing team. In addition, the proposed method provides an effective mechanism to guide crowdsourcing members to participate in collaborative tasks. By establishing a collaborative workflow simulation model, the asynchronous collaborative work process of crowdsourcing members is rationally planned and managed. It also optimizes and schedules crowdsourcing members to handle the subtask whose priority is different from the workflow timing. Finally, through a case study of the collaborative design task of a medical analgesia pump, the effectiveness and feasibility of this method are demonstrated.

For future work, first, since the optimal combination process of crowdsourcing members does not take the incentive problem into account, that is, whether the crowdsourcing members will receive the expected compensation after providing the service, it is necessary to further improve the business model of the cloud design platform to increase the attractiveness of collaborative tasks to crowdsourcing members. Simultaneously, depending on the specific business model, we will consider compensation as one of the evaluation factors for the comprehensive ability of crowdsourcing members. Second, a larger case study is needed to further validate and improve the method proposed in this article.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the National Key R&D Program of China (No. 2017YFB1104205) and the National 111 Project (No. B13044).

ORCID iDs

Jianjie Chu

Jiashuang Fan

References

Rosen

Wang

, et al. Cloud-based design and manufacturing: a new paradigm in digital manufacturing and design innovation. Comput-Aided Des 2015; 59(C): 1–14.

Zhang

Ren

, et al. Further discussion on cloud manufacturing. Comput Int Manuf Syst 2011(3): 449–457.

Tao

Zhang

Venkatesh

, et al. Cloud manufacturing: a computing and service-oriented manufacturing model. Proc IMechE, Part B: J Engineering Manufacture 2011; 225(10): 1969–1976.

Valilai

Houshmand

. Depicting additive manufacturing from a global perspective; using Cloud manufacturing paradigm for integration and collaboration. Proc IMechE, Part B: J Engineering Manufacture 2014; 229(12): 2216–2237.

Chen

Chu

, et al. Modular restructuring and distribution method of collaborative task in industrial design cloud platform. Comput Int Manuf Syst 2018; 24(3): 720–730.

Zeng

Benatallah

Ngu

AHH

, et al. QoS-aware middleware for Web services composition. IEEE Trans Software Eng 2004; 30(5): 311–327.

Menascé Daniel

. QoS issues in Web services. IEEE Internet Comput 2002; 6(6): 72–75.

Ardagna

Pernici

. Global and local QoS guarantee in Web service selection. In: Proceedings of the BMP international conference on business process management, Berlin; Heidelberg, 5 September 2005, pp.32–46. Berlin; Heidelberg: Springer.

Liu

. Granulation-based resource classification in Cloud Manufacturing. Proc IMechE, Part B: J Engineering Manufacture 2015; 229(7): 1258–1270.

10.

Ardagna

Pernici

. Adaptive service composition in flexible processes. IEEE Trans Software Eng 2007; 33(6): 369–384.

11.

Wang

Yan

. A quantitative approach to the organisation of cross-functional teams in concurrent engineering. Int J Adv Manuf Technol 2003; 21(10–11): 879–888.

12.

Quélin

. Core competencies, R&D management and partnerships. Euro Manage J 2000; 18(5): 476–487.

13.

Liu

Chu

, et al. Research on member optimal selection of network team. Comput Int Manuf Syst 2017; 23(6): 1205–1215.

14.

Feng

Jiang

Fan

, et al. A method for member selection of cross-functional teams using the individual and collaborative performances. Euro J Operat Res 2010; 203(3): 652–661.

15.

Papakostas

Georgoulias

Koukas

, et al. Organisation and operation of dynamic manufacturing networks. Int J Comput Int Manuf 2015; 28(8): 893–901.

16.

Cao

Jiang

. Demand-based manufacturing service capability estimation of a manufacturing system in a social manufacturing environment. Proc IMechE, Part B: J Engineering Manufacture 2015; 38(4): 242–252.

17.

Wang

Liu

Zhao

, et al. Behavior consistency computation for workflow nets with unknown correspondence. IEEE/CAA J Automat Sinica 2018; 5(1): 281–291.

18.

Yang

Fei

, et al. Task decomposition and resources allocation of product collaboration innovative design. J Chongqing Univ 2014; 37(1): 31–38.

19.

Framinan

Perez-Gonzalez

Fernandez-Viagas

. Deterministic assembly scheduling problems: a review and classification of concurrent-type scheduling models and solution procedures. Euro J Operat Res 2018; 273(2): 401–417.

20.

Gyulai

Kádár

Monostori

. Scheduling and operator control in reconfigurable assembly systems. Procedia Cirp 2017; 63: 459–464.

21.

Lee

. Minimizing total completion time in the assembly scheduling problem. Comput Ind Eng 2018; 122: 211–218.

22.

Cheng

Tao

Liu

, et al. Energy-aware resource service scheduling based on utility evaluation in cloud manufacturing system. Proc IMechE, Part B: J Engineering Manufacture 2013; 227(12): 1901–1915.

23.

Lin

Yang

Zhuang

, et al. Multi-centric management and optimized allocation of manufacturing resource and capability in cloud manufacturing system. Proc IMechE, Part B: J Engineering Manufacture 2017; 231(12): 2159–2172.

24.

Mourtzis

Vlachou

. Cloud-based cyber-physical systems and quality of services. TQM J 2016; 28(5): 704–733.

25.

Bao

Yang

, et al. Decomposition model in product customization collaborative development task. Comput Int Manuf Syst 2014; 20(7): 1537–1545.

26.

Tong

, et al. Modular decomposition design process based on fuzzy design structure matrix. Comput Int Manuf Syst 2010; 3(16): 476–483.

27.

Stone

Veloso

. Task decomposition, dynamic role assignment, and low-bandwidth communication for real-time strategic teamwork. Artif Intell 1999; 110(2): 241–273.

28.

Kam

. Manufacturing network for rapid tool/die making. Int J Comput Int Manuf 2006; 19(1): 79–89.

29.

Liu

. Agent-based resource discovery architecture for environmental emergency management. Exp Syst Appl 2004; 27(1): 77–95.

30.

Yang

Wang

Xiao

, et al. Research on resource discovery in networked manufacturing environment. Comput Int Manuf Syst 2003; 9(1): 47–51.

31.

Meng

. Theory and method on collaborative design support environment and conflict resolution. Nanjing, China: Southeast University Press, 2010.

32.

Xiong

Wang

, et al. Virtual resource scheduling method of cloud manufacturing oriented to multi-objective optimization. Comput Int Manuf Syst 2015; 21(11): 3079–3087.