CN109919742A

CN109919742A - Task crowdsourcing allocation processing method based on personalized competitive bidding excitation

Info

Publication number: CN109919742A
Application number: CN201910137835.4A
Authority: CN
Inventors: 向朝参; 陈超; 刘凯; 冯亮
Original assignee: Chongqing University
Current assignee: Chongqing University
Priority date: 2019-02-25
Filing date: 2019-02-25
Publication date: 2019-06-21

Abstract

The present invention provides a kind of task crowdsourcing allocation processing methods based on personalized competitive bidding excitation, in its description design for the task competitive bidding book for setting, user is allowed to carry out the competitive bidding distribution principle between each competitive bidding unit for including in personalized constraint task competitive bidding book using the competitive bidding relations of distribution；For the allocation processing of waiting task, the then competitive bidding corresponding relationship by establishing between waiting task and task competitive bidding book set by user, from can satisfy cost performance optimal one that selects competitive bidding unit in the task competitive bidding book that the competitive bidding relations of distribution require as acceptance of the bid competitive bidding book, it sets up full task packet and distributes to corresponding user's processing, the personalization distributed thus, it is possible to better meet user for task competitive bidding, flexibility demand, and it is designed by the description design to task competitive bidding book and to the Task-decomposing of task allocation processing, the validity and feasibility of function ease for use and task allocation processing are preferably taken into account.

Description

Task crowdsourcing allocation processing method based on personalized competitive bidding excitation

Technical field

The present invention relates to internet task allocation technique field, in particular to a kind of appointing based on personalized competitive bidding excitation Business crowdsourcing allocation processing method.

Background technique

With intelligent movable device-aware function is become stronger day by day and is popularized and internet and wireless communication technique it is fast Speed development, makes the crowdsourcing model of intelligent perception come into being, and receive more and more attention, also produces a large amount of gunz sense Know system and crowdsourcing service platform.Crowdsourcing model is completed by integrating the intelligent perception user of intelligent movable equipment and internet Task to be processed, oneself is widely used in information retrieval, artificial intelligence, knowledge excavation, urban transportation, man-machine friendship at present The mutually technological services field such as study.And task relevant for some research work, in order to improve the crowdsourcing distribution of intelligent perception Treatment effeciency is highly desirable design incentive mechanism to motivate the participation of user.

Auction theory can increase the competition between user, and therefore, it is widely used in current intelligent perception crowdsourcing model Exciting torque work in.User submits a tender to the task of publication according to their preference, then, is based on each use The competitive bidding at family, crowdsourcing service platform determine the distribution by each task according to certain task allocation rule and calculate reward.

In practical applications, since different user local environment, interest and when and where are different from, he Be often desirable to possess oneself diversified selection in competitive bidding task choosing.For example, traffic route work for example shown in Fig. 2 Making in task, user Bob can be completed at the same time task 1 and task 2 (task1 and task 2), and when user Lucy is due to the free time Between limited one but merely desired among processing task 1 and task 2 (task1 and task 2).But the research master of Most current Design objective and reward distribution mechanism are removed from the angle of platform, go to consider user's competitive bidding from user perspective to have ignored Diversified preference.The Bidding Strategiess of user individual preference are adapted to, referred to as personalized competitive bidding (Personalized Bidding, PB), it not only improves platform and is conducive to users again.For example, a professional investigation to 18 to 60 years old 1500 people Show that the crowd more than 56% more likes the service with individualized experience.Therefore, personalized competitive bidding is by meeting user's Personalization hobby improves the internal motivation in their psychology, so as to preferably motivate these intelligent perceptions user to participate in In the processing of crowdsourcing task.

Therefore, the personalized competitive bidding incentive mechanism of intelligent perception how is designed, the allocation processing of Lai Shixian crowdsourcing task is right It is particularly important for crowdsourcing service platform.

Summary of the invention

Aiming at the above shortcomings existing in the prior art, the purpose of the present invention is to provide one kind is swashed based on personalized competitive bidding The task crowdsourcing allocation processing method encouraged, to better meet personalization, the flexibility demand that user distributes task crowdsourcing, To help in the processing for motivating intelligent perception user to participate in crowdsourcing task, and take into account the reasonability of task distribution.

To achieve the above object, present invention employs following technical solutions:

Based on the task crowdsourcing allocation processing method of personalized competitive bidding excitation, include the following steps:

Waiting task in release tasks library；

User sets task competitive bidding book, and each each task competitive bidding school bag set by user includes one or more competitive bidding lists The competitive bidding relations of distribution between member and competitive bidding unit；Each competitive bidding unit corresponds to set by user one for recording it Or multiple expectation competitive bidding tasks and corresponding expectation are recompensed；

Statistics establishes the competitive bidding corresponding relationship between the waiting task in task library and task competitive bidding book set by user, And for each waiting task for having competitive bidding corresponding relationship is established, from the task competitive bidding that can satisfy the requirement of the competitive bidding relations of distribution Cost performance optimal one that competitive bidding unit is selected in book extracts waiting task as acceptance of the bid competitive bidding book from task library, Set up meet it is described acceptance of the bid competitive bidding book in the competitive bidding relations of distribution require and with the phase in the optimal competitive bidding unit of the cost performance Hope the task packet that matches of competitive bidding task, distribute to user corresponding to the acceptance of the bid competitive bidding book, thus realization have to foundation it is competing Mark the allocation processing of each waiting task of corresponding relationship.

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, specifically, for arbitrary i-th A user u_iThe task competitive bidding book of settingUpper right footmark * indicates that the competitive bidding relations of distribution mark, any k-th wherein included Competitive bidding unitIt indicates are as follows:

Wherein, N is total number of users,For the competitive bidding unit number for including in task competitive bidding book, T_i,k、P_i,kIt respectively indicates Competitive bidding unitTotal value is recompensed in the expectation competitive bidding task vector of middle record and expectation, and has T_i,k={ τ_i,k,h|h∈{1,..., H_i,k,τ_i,k,h、p_i,k,hRespectively indicate competitive bidding unitIn h-th expectation competitive bidding task and corresponding Expectation reward, h ∈ { 1 ..., H_i,k, H_i,kIndicate competitive bidding unitIn include expectation competitive bidding task total number, and it is expected Competitive bidding task vector T_i,kIt is the subset of task library T, that is, hasτ_i,k,h∈T。

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, specifically, the competitive bidding distributes Relationship includes the subset selection relations of distribution, and the subset selection relations of distribution indicate that expectation is assigned in task competitive bidding book comprising complete One among portion's competitive bidding unit or mutually without the multiple of intersection；That is, the task competitive bidding book of the subset selection relations of distribution indicates are as follows:

Wherein,Indicate i-th of user u_iThe task competitive bidding book of the set subset selection relations of distribution, upper right footmark O Indicate that the competitive bidding relations of distribution are that subset selects the relations of distribution, ∪ indicates that subset selects relations of distribution operator；Expression task is competing K-th of competitive bidding unit in bidding documents, For user u_iThe competitive bidding list for including in set task competitive bidding book First number.

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, specifically, the competitive bidding distributes Relationship includes selecting a selection relations of distribution, it is described select a selection relations of distribution and indicate that only expectation is assigned in task competitive bidding book include Any one among whole competitive bidding units；That is, selecting the task competitive bidding book of a selection relations of distribution indicates are as follows:

Wherein,Indicate i-th of user u_iThe set task competitive bidding book for selecting a selection relations of distribution, upper right footmark o Indicate that the competitive bidding relations of distribution are to select a selection relations of distribution,To select a selection relations of distribution operator, expression only distributes its company Connect any one among object；K-th of competitive bidding unit in expression task competitive bidding book, For user u_iThe competitive bidding unit number for including in set task competitive bidding book.

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, specifically, the competitive bidding distributes Relationship includes the mixing selection relations of distribution, and the mixing selection relations of distribution indicate that only expectation is assigned to set multiple competitive biddings Any one among task groups, includes one or more competitive bidding units in set each competitive bidding task groups, and each Subset selects the relations of distribution each other between each competitive bidding unit in competitive bidding task groups, i.e., each competitive bidding task groups indicate expectation distribution To one among whole competitive bidding units wherein included or mutually without the multiple of intersection；That is, the task of the mixing selection relations of distribution Competitive bidding book indicates are as follows:

Wherein,Indicate i-th of user u_iThe task competitive bidding book of the set mixing selection relations of distribution, upper right footmark XO indicates that the competitive bidding relations of distribution are the mixing selection relations of distribution,To select a selection relations of distribution operator, expression only distributes it Any one among connecting object, ∪ indicate that subset selects relations of distribution operator；Indicate the mixing selection relations of distribution Task competitive bidding bookIn set first of competitive bidding task groups, andL_iTask competitive bidding bookIn include competitive bidding task groups sum,Expression task competitive bidding bookIn in set first of competitive bidding task groups N-th of competitive bidding unit, n ∈ { 1 ..., K_i,l, and Expression task K-th of competitive bidding unit in competitive bidding book,For user u_iThe competitive bidding unit number for including in set task competitive bidding book.

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, specifically, having for foundation competing The waiting task for marking corresponding relationship decomposes in each task competitive bidding book respectively if being corresponding with multiple tasks competitive bidding book and includes Competitive bidding unit, and according to decompose obtain each competitive bidding unit expectation competitive bidding task and expectation reward determine each competitive bidding respectively The cost performance of unit, the task competitive bidding book where selecting the optimal competitive bidding unit of cost performance is used as acceptance of the bid competitive bidding book, from task Extract corresponding waiting task in library, set up the competitive bidding relations of distribution met in the acceptance of the bid competitive bidding book require and with it is described The task packet that expectation competitive bidding task in the optimal competitive bidding unit of cost performance matches, is distributed to corresponding to the acceptance of the bid competitive bidding book User.

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, preferably, the competitive bidding The relations of distribution include the mixing selection relations of distribution, and it is set multiple that the mixing selection relations of distribution indicate that only expectation is assigned to Any one among competitive bidding task groups, includes one or more competitive bidding units in set each competitive bidding task groups, and Subset selects the relations of distribution each other between each competitive bidding unit in each competitive bidding task groups, i.e., each competitive bidding task groups indicate expectation Be assigned to whole competitive bidding unit wherein included one or more of；

When the competitive bidding unit for including in task resolution competitive bidding book, the task competitive bidding book of the relations of distribution is selected for mixing, Isolation specifically:

Mark is had at least one is added in each competitive bidding unit in the task competitive bidding book of the mixing selection relations of distribution The virtual task of note, it is ensured that the mutual not phase of the label of the virtual task added in each competitive bidding unit that each competitive bidding task groups include Together, and every two belong to all include in the competitive bidding units of different competitive bidding task groups same tag virtual task, it is each virtual Task is idle task, and identical virtual task is marked to be considered as the same idle task, and different virtual tasks is marked to be regarded For different idle tasks；Thus respectively by each competitive bidding cell translation be the competitive bidding selecting unit added with virtual task, simultaneously The competitive bidding relations of distribution between each competitive bidding selecting unit are determined as the subset selection relations of distribution, that is, it is complete to indicate that expectation is assigned to One among portion's competitive bidding selecting unit or mutually without the multiple of intersection；

User corresponding to task competitive bidding book for the mixing selection relations of distribution builds for each competitive bidding selecting unit correspondence Vertical one belongs to the Virtual User of the user；

It selects the task competitive bidding book of the relations of distribution to decompose mixing as a result, and is respectively corresponding with different virtual use in order to multiple The competitive bidding selecting unit of family and the relations of distribution of subset selection each other.

It is preferably, true in selection in the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation Surely when acceptance of the bid competitive bidding book, for the task competitive bidding book of the mixing selection relations of distribution, method of determination is selected are as follows: select according to each mixing The expectation competitive bidding task of each competitive bidding selecting unit and expectation reward, determine each competing respectively in the task competitive bidding book of the relations of distribution The cost performance for marking selecting unit, the task competitive bidding book where selecting the optimal competitive bidding selecting unit of cost performance are used as acceptance of the bid competitive bidding The user that the corresponding Virtual User of the optimal competitive bidding selecting unit of the cost performance is belonged to is determined as the user that gets the bid by book, from Corresponding waiting task is extracted in task library, is set up and is appointed with the expectation competitive bidding in the optimal competitive bidding selecting unit of the cost performance It is engaged in the task packet that matches, get the bid user described in dispensing.

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, preferably, competitive bidding is determined The concrete mode of the cost performance of unit are as follows: for arbitrary i-th of user u_iAny kth for including in the task competitive bidding book of setting A competitive bidding unitBy calculating its payment efficiencyDetermine corresponding competitive bidding unitCost performance, payment Efficiency ε_i,kValue it is more big, determine competitive bidding unitCost performance it is better；Wherein, T_i,kIndicate competitive bidding unitMiddle record It is expected that competitive bidding task vector, P_i,kIndicate competitive bidding unitTotal value is recompensed in the expectation of middle record, and has T_i,_k={ τ_i,k,h|h∈ {1,...,H_i,k,τ_i,k,h、p_i,k,hRespectively indicate competitive bidding unitIn h-th of expectation competitive bidding task With corresponding expectation reward, h ∈ { 1 ..., H_i,k, H_i,kIndicate competitive bidding unitIn include expectation competitive bidding task total number； |T_i,k| it indicates to calculate expectation competitive bidding task vector T_i,kIn include expectation competitive bidding task number, i.e., | T_i,k|=H_i,k。

In the above-mentioned task crowdsourcing allocation processing method based on personalized competitive bidding excitation, specifically, having for foundation competing The waiting task for marking corresponding relationship, if being only corresponding with a task competitive bidding book, using the task competitive bidding book as acceptance of the bid competitive bidding Book extracts corresponding waiting task from task library, distributes to user corresponding to the acceptance of the bid competitive bidding book.

Compared with the prior art, the invention has the following advantages:

1, the present invention is based on the task crowdsourcing allocation processing method of personalized competitive bidding excitation, the designs of task distribution principle There are corresponding relationships for description method design with the task competitive bidding book for setting；For retouching for the task competitive bidding book for setting It states in design, user is allowed to carry out each competitive bidding unit for including in personalized constraint task competitive bidding book using the competitive bidding relations of distribution Between competitive bidding distribution principle；For the allocation processing of waiting task, then by establish waiting task with it is set by user Competitive bidding corresponding relationship between task competitive bidding book judges that can platform meet the competitive bidding relations of distribution of task competitive bidding book requirement, and The optimal conduct of cost performance of competitive bidding unit is selected from the task competitive bidding book that can satisfy the requirement of the competitive bidding relations of distribution Get the bid competitive bidding book, extract waiting task set up meet it is described acceptance of the bid competitive bidding book in the competitive bidding relations of distribution require and with it is described The task packet that expectation competitive bidding task in the optimal competitive bidding unit of cost performance matches distributes to user's processing, so that wait locate The distribution of reason task is dependent on the task competitive bidding book for setting, distributes so as to better meet user for task competitive bidding Personalization, flexibility demand, facilitate motivate intelligent perception user participate in the processing of crowdsourcing task.

2, the present invention is based on the task crowdsourcing allocation processing method of personalized competitive bidding excitation, task distribution principle is will Enough meet an acceptance of the bid competitive bidding book for having the optimal competitive bidding unit of cost performance in the task competitive bidding book of competitive bidding relations of distribution requirement, It is distributed to execute task, that is, introduces cost performance contrast mechanism, it is competing to avoid excessive, blindness the raising its expectation reward of the task The interests of platform and task supplier are got the bid and influenced to bidding documents, has preferably taken into account the reasonability of task distribution.

3, the present invention is based in the task crowdsourcing allocation processing method of personalized competitive bidding excitation, XOR-of-OR has also been devised Competitive bidding describes the competitive bidding relations of distribution that method removes description user task competitive bidding book, can either be used to indicate the personalization that user is all Competitive bidding describes demand, meets personalization, flexibility demand that user distributes task competitive bidding well, while in turn avoiding competing Too long, the excessively complicated problem of the logical description of the relations of distribution is marked, it is user-friendly, to successfully realize user's expression Property and user use the compromise between friendly.

4, in the task crowdsourcing allocation processing method motivated the present invention is based on personalized competitive bidding, for XOR-of-OR competitive bidding The task competitive bidding book of the relations of distribution, in task allocation processing by jointly considering the XOR-of-OR competitive bidding of task competitive bidding book Relationship in the relations of distribution between three kinds of logical operators, be broken down into it is multiple be respectively corresponding with different Virtual User and The competitive bidding selecting unit of the relations of distribution of subset selection each other, that will there is complicated task competitive bidding relations of distribution dependency graph in this way Property competitive bidding dismantling at multiple independent, simple single idea competitive biddings so that task distribution principle is from complicated XOR-of-OR The conversion of the competitive bidding relations of distribution reduces the complexity and operand that task distribution calculates for relatively simple OR dependence, Make task distribution more easily handle, thus realize user's expressivity, user friendly, calculate validity three between folding Inner feelings.

Detailed description of the invention

Fig. 1 is the flow chart of the task crowdsourcing allocation processing method motivated the present invention is based on personalized competitive bidding.

Fig. 2 is a traffic route task exemplary diagram.

Fig. 3 is the description dimensional space signal of the competitive bidding description based on two-dimensional space and the competitive bidding description based on three-dimensional space Figure.

Fig. 4 is dismantling and the regrouping process schematic diagram of user individual competitive bidding task dependence graph.

Fig. 5 is the real trace datagram of the user vehicle of experiment.

Fig. 6 is that the present invention is based on the task crowdsourcing allocation processing method of personalized competitive bidding excitation and existing task are many in experiment The performance of packet allocation processing method compares figure.

English information involved in figure is described as follows:

In Fig. 2, Task 1~5 respectively indicates task 1~5；Jack, Alice, Lucy, Bob respectively indicate participating user's Title.

In Fig. 3, the expression descriptive power of Expressiveness index book；SXB refers to based on two-dimensional space (2-D space) The method of description；PRINCE refers to the method based on three-dimensional space (3-D space) description.

In Fig. 4:

PB of one user with task dependency graph: the task of user individual competitive bidding, which relies on, closes System's figure；

Decompose: task dependence graph decomposes；

Greedily selecting: it is selected based on greedy method；

Virtual users with independent SMBs: the Virtual User comprising separate single idea competitive bidding；

Re-combine: task dependence graph recombination；

An adaptive critical-payment computation scheme: based on the adaptive of crucial price Recompense calculation method.

In Fig. 5, Latitude refers to dimension；Longitude refers to longitude.

In Fig. 6, APT refers to the payment reward of averagely each execution task；ANU refers to the competitive bidding task of averagely each user Number；ADL refers to that the competitive bidding of averagely each user describes length；Ours refers to our proposed methods；SMB, SOB, SXB respectively indicate existing Other methods in technology.

Specific embodiment

With reference to the accompanying drawings and detailed description to the task crowdsourcing distribution motivated the present invention is based on personalized competitive bidding at Reason method is further detailed.

In the scene of actual intelligent perception, user is usually because having different sensing capabilities, hobby and field Scape (such as time, space and current event etc.), to have different competitive bidding preferences.Therefore, they are desirable to according to him Oneself personal preference carry out personalized competitive bidding (Personalized bidding).With traffic route shown in Fig. 2 For task, crowdsourcing service platform issues the task of 5 different locations, uses τ_j, j={ 1 ..., 5 }.There are 4 users to want to join With with complete this 5 intelligent perception tasks.But they have different preferences due to different interest and background.For example, As shown in the Dark grey lines in Fig. 2, Bob will pass through task τ₁And τ₂The position at place.Therefore, he thinks that task is completed in competitive bidding τ₁And τ₂, respectively with 20 yuan and 30 yuan of price.Bob wants to do the two tasks or do not do both.Such as the light gray of Fig. 2 Shown in lines, Alice and Lucy will pass through τ₁, τ₂And τ₄.Since Alice has the sufficient time, she wishes to do any wherein several These three a or whole tasks, therefore, any τ of her competitive bidding₁, τ₂And τ₄A subset, respectively with 50,10,10 yuan of valence Lattice.But Lucy but only has the limited time, she it is expected to be only τ₁, τ₂And τ₄One of task, 10,30,10 yuan of difference Price.Solid black lines and dotted line Jack as shown in Figure 2 there are two optional path.Therefore, he wants to be τ₃And τ₄Respectively with valence Lattice 15 and 10 yuan or it is τ₃And τ₅Respectively with price 15 and 20 yuan.

Personalized competitive bidding can not only be met the needs of users, and can also be conducive to intelligent perception platform.Flexible individual character Changing competitive bidding can increase by the task number of competitive bidding, to expand the set for finding the candidate solution of optimal task assignment.Therefore, Personalized competitive bidding can reduce total consuming (social cost) and the total payoff (total that platform completes all tasks payment).It include below two task (τ with one₁And τ₂) simple case, to further illustrate personalized competitive bidding to platform Benefit.As shown in Fig. 2, Bob is with 50 yuan of price together competitive bidding τ₁And τ₂.Alice is with price 50 and 10 yuan difference competitive bidding τ₁And τ₂ Subset.Lucy is but with price 10 and 30 yuan difference competitive bidding τ₁And τ₂In one of them.

If not considering the personalized competitive bidding of user, mechanism will be only with simplest competitive bidding mode, such as single idea Competitive bidding (Single single-minded bidding method, SSB) (reference can be made to document " X.Chen, X.Wu, X.-Y.Li, Y.He,and Y.Liu.Privacy-preserving high-quality map generation with participatory sensing.In Proc.IEEE INFOCOM,2014."；"H.Jin,L.Su,H.Xiao,and K.Nahrstedt.INCEPTION:incentivizing privacy-preserving data aggregation for mobile crowd sensing systems.In Proc.ACM MobiHoc,pages 341–350,2016."； “M.Karaliopoulos,I.Koutsopoulos,and M.Titsias.First learn then earn: optimizing mobile crowdsensing campaigns through data-driven user pro ling.In Proc.ACM MobiHoc,pages 271–280,2016").Assuming that Alice and Lucy think to earn money as much as possible, they will The highest single idea competitive bidding scheme of price is selected from their candidate competitive bidding scheme.Therefore, the bidding documents of Bob, Alice and Lucy Respectively { (τ₁, τ₂, 50) }, { (τ₁, τ₂, 60) } and (τ₂, 30) }.Bid based on these three users, in the competitive bidding machine of single idea Under system, platform formulates following optimal distributing scheme so that totle drilling cost is minimum: Bob completes τ₁And τ₂, totle drilling cost is 50 yuan.So And if it is considered that the personalized competitive bidding of user, the bidding documents of Bob, Alice and Lucy are respectively { (τ₁, τ₂, 50) }, { (τ₁, 50), (τ₂, 10), (τ₁, τ₂, 60) } and (τ₁, 10), (τ₂, 30) }.So, under personalized bidding mechanism, the OPTIMAL TASK point of platform Be with scheme: Lucy and Alice completes τ respectively with 10 yuan of price₁And τ₂, that is, the totle drilling cost of platform is 20 yuan.Therefore, if Consider the personalized competitive bidding of user, platform can will complete task τ₁And τ₂Totle drilling cost be reduced to 20 yuan from 50 yuan.It can be seen that personalization Bidding mechanism not only improves platform and is conducive to users again.Also, personalized competitive bidding is mentioned by meeting the personalized hobby of user Internal motivation in their high psychology, so as to preferably motivate these intelligent perceptions user to participate in the place of crowdsourcing task In reason.

Considered based on above-mentioned factor, the present invention proposes a kind of task crowdsourcing allocation processing side based on personalized competitive bidding excitation Method, as shown in Figure 1, including the following steps:

Step 1, task publication:

Intelligent perception platform is to the waiting task in perception task interested user's release tasks library.Assuming that using T Indicate the task library (or being set of tasks) of platform, i.e. T={ τ_j| j ∈ { 1 ..., M } }, wherein τ_jIt is respectively indicated j-th with M Task and task total number.Assuming that a total of N number of user is interested in these perception tasks here, then user's set can be used U={ u_i| i ∈ { 1 ..., N } } it indicates, u_iIndicate i-th of user therein.

Step 2, task competitive bidding:

User sets task competitive bidding book, and each each task competitive bidding school bag set by user includes one or more competitive bidding lists The competitive bidding relations of distribution between member and competitive bidding unit；Each competitive bidding unit corresponds to set by user one for recording it Or multiple expectation competitive bidding tasks and corresponding expectation are recompensed.

Description method based on bidding documents, user set competitive bidding book according to themselves preference and current scene, including They want the task and desired price completed.In the methods of the invention, user can with the task competitive bidding book of setting individual, Such as arbitrary i-th of user u_iThe task competitive bidding book of setting can be expressed as b_i ^*, each task competitive bidding school bag includes one or more A competitive bidding unit, upper right footmark * indicate competitive bidding relations of distribution label, and using the different competitive bidding relations of distribution, then competitive bidding distribution is closed System's label is also corresponding different, and the competitive bidding distribution principle between each competitive bidding unit in task competitive bidding book passes through the competitive bidding relations of distribution It uses restraint.Here, needing to propose the definition about competitive bidding unit.

Define 1: competitive bidding unit (or being atom competitive bidding, Atomic bid), is in task competitive bidding book set by user One basic unit can be seen as a single idea competitive bidding (Single-minded bid, SMB).Include in task competitive bidding book Any k-th of competitive bidding unitIt may be expressed as:

Wherein, N is total number of users,For the competitive bidding unit number for including in task competitive bidding book, T_i,k、P_i,kIt respectively indicates Competitive bidding unitTotal value is recompensed in the expectation competitive bidding task vector of middle record and expectation, and has T_i,k={ τ_i,k,h|h∈{1,..., H_i,k,τ_i,k,h、p_i,k,hRespectively indicate competitive bidding unitIn h-th expectation competitive bidding task and corresponding Expectation reward, h ∈ { 1 ..., H_i,k, H_i,kIndicate competitive bidding unitIn include expectation competitive bidding task total number, and it is expected Competitive bidding task vector T_i,kIt is the subset of task library T, that is, hasτ_i,k,h∈T.Finally, these users can be by setting The competitive bidding book of all personalizations is submitted on platform.

For platform, a competitive bidding unitIndicate that this user may recompense price with expectation P_i,kCompetitive bidding task vector T it is expected in completion_i,kIn all expectation competitive bidding tasks, if do not complete expectation competitive bidding task if Less than corresponding reward.

And for user u_iFor, in addition to its desired reward price P_i,kOutside, task T is executed there are one it_i,kIt is true It is real to expend C_i,k.In practice, this true consuming is the private information of user, therefore only user's himself is known.It is worth saying It is bright, complete the true consuming C of all tasks_i,kThe sum of be exactly overall society cost (Social cost).Due to the mankind from Private and rationality person's character, they are invariably prone to be offered a high price in competitive bidding in the price actually expended it is expected to obtain more receipts Benefit, therefore, C_i,k≤P_i,k, user u_iIncome beThis is also that platform needs design objective allocation strategy to take precautions against One major consideration of property.

Step 3, task distribution:

Based on task competitive bidding book set by user, platform needs according to the task distribution principle of design to be each user distribution TaskAnd rewardEach user u_iThe distributing to him of the task is executed by perceivingAnd sensing results are uploaded to flat Platform.

Last platform pays reward to the user

In the methods of the invention, the task distribution principle of design is as follows:

It can be seen that in the task crowdsourcing allocation processing method motivated the present invention is based on personalized competitive bidding, task distribution The design of principle, there are corresponding relationships for the description method design with the task competitive bidding book for setting.For appointing for setting In the description design of competitive bidding book of being engaged in, permission user is come using the competitive bidding relations of distribution includes in personalized constraint task competitive bidding book Competitive bidding distribution principle between each competitive bidding unit；For the allocation processing of waiting task, then by establishing waiting task With the competitive bidding corresponding relationship between task competitive bidding book set by user, judge that can platform meet the competitive bidding of task competitive bidding book requirement The relations of distribution, and from can satisfy the competitive bidding relations of distribution require task competitive bidding book in select competitive bidding unit cost performance it is best One as acceptance of the bid competitive bidding book, extract waiting task set up meet it is described acceptance of the bid competitive bidding book in the competitive bidding relations of distribution want It asks and the task packet to match with the expectation competitive bidding task in the optimal competitive bidding unit of the cost performance distributes to user's processing, So that the distribution of waiting task is dependent on the task competitive bidding book for setting, so as to better meet user for The personalization of task competitive bidding distribution, flexibility demand help that intelligent perception user is motivated to participate in the processing of crowdsourcing task； Meanwhile task distribution principle is that have cost performance optimal competing in the task competitive bidding book for will meet the requirement of the competitive bidding relations of distribution An acceptance of the bid competitive bidding book of unit is marked, the distribution of Lai Zhihang task introduces cost performance contrast mechanism, to avoid excessively, blindly The task competitive bidding book acceptance of the bid for raising its expectation reward and influence the interests of platform and task supplier, preferably taken into account task The reasonability of distribution.

However, to provide the flexibility of personalized competitive bidding, such as in the allocation processing of the crowdsourcing task based on bidding mechanism The foundation of fruit bidding mechanism is not abundant enough, then is difficult to meet the needs of user individual competitive bidding, but bidding mechanism abundantization, past Toward but it is easy so that intelligent perception user is more complicated to the setting of competitive bidding function in crowdsourcing service system, cause user setting competing Mark the friendly that function is inconvenient for use, and reduction user uses；Therefore, in the demand and task competitive bidding for how taking into account personalized competitive bidding In the friendly that function uses, existing bidding mechanism is difficult to meet needs well.On the other hand, with competitive bidding personalization and The complication of the promotion of flexibility and the setting of competitive bidding function, becomes the crowdsourcing allocation processing to waiting task It complicates, the candidate collection of task distribution may be significantly increased, so that substantially increasing crowdsourcing service system executes task point Timing calculates the time complexity and processing difficulty of optimal task assignment mode according to allocation rule, in some instances it may even be possible to so that solving most The excellent task method of salary distribution becomes NP-hard problem, and the distribution of execution task is caused to need to expend the complexity of exponential time, thus The task crowdsourcing distribution for realizing exciting torque is set to calculate validity and feasibility severe exacerbation.Therefore, how design objective Crowdsourcing allocation processing mechanism can have personalized competitive bidding drive characteristic, preferably take into account personalized competitive bidding demand and The friendly that task competitive bidding function uses further preferably is taken into account under bidding mechanism for the effective of task crowdsourcing allocation processing Property and feasibility, and become another new issue of crowdsourcing task allocation technique area research.

Another emphasis of the invention is to consider under the premise of the personalized competitive bidding of user is to exciting torque, How to design to better meet two new features of the expression adequacy of personalized competitive bidding incentive mechanism and user friendly, and How the calculating validity (computation efficiency) of in exciting torque task allocation processing is preferably realized With tactful precaution (strategy-proof).But unfortunately, the personalized competitive bidding of high flexible allows user being capable of table Up to a variety of different competitive bidding preferences, thus while bringing abundant expressivity for them, but be also while bringing following three side The technological challenge in face:

1, function ease for use: it is longer and more complicated that the flexibility of personalized competitive bidding describes the competitive bidding of user, to hold User is easily led to using very inconvenient.Therefore, the function ease for use and user friendly that personalized competitive bidding designs bidding mechanism The feature of property is more difficult.

2, allocation processing validity: the time of task distribution has been significantly greatly increased due to increasing competitive bidding quantity in personalized competitive bidding Selected works close, to substantially increase the time complexity for calculating optimal task assignment.Chapter 3, it will demonstrate that this problem is NP- Hard problem.More seriously, even if the approximate solution for obtaining invariant is also required to expend exponential time complexity, to make Realize that this feature of the calculating validity of exciting torque is extremely difficult.

3, tactful precaution: different competitive biddings selects the distribution to task to increase complicated constraint, asks task distribution Inscribe it is more difficult, in traffic route task example as shown in Figure 2, Bob need task 1 and task 2 while distributing To him.In addition, the constraint dependence between task distribution can be had strategically by selfish user using more to obtain Income, so that incentive mechanism is more difficult tactful prevention.

So, the technological challenge based on aforementioned three aspects, the design problem of personalized competitive bidding excitation specifically include that (i) How design personalized competitive bidding describes method, the preference based on themselves is allowed the user to describe bidding documents, to meet The good expressivity of user (expressiveness) and user friendly (user-friendliness)；(ii) how to design Task and reward allocation rule, so that all task T are distributed to user U based on the personalized competitive bidding of user by platform, thus most Smallization completes the overall society cost (social cost) of all tasks, while meeting and calculating validity (computation ) and tactful precaution (strategy-proof) efficiency.

It is directed to how to solve the technological challenge in terms of above three, deployment analysis and explanation below.

A, the personalized competitive bidding based on three-dimensional expression space describes method design.

1, the competitive bidding relations of distribution description based on two-dimentional expression of space.

Define 2: subset selects the relations of distribution (the alternatively referred to as OR competitive bidding relations of distribution, abbreviation OR competitive bidding) description；Based on AND With OR logical operator, it then follows following two step:

(1), AND atom competitive bidding: each user u is constructed_iAn atom competitive bidding can be submitted, is usedIt indicates, including any Quantity (i.e. H_i,k) AND task to (τ_i,k,h,p_i,k,h), h ∈ (1 ..., H_i,k).It indicates that user it is expected to complete all tasks T_i,k={ τ_i,k,h|h∈{1,...,H_i,kTo receive remunerationOtherwise any task is not done and obtains task Remuneration.

So atom competitive bidding can be expressed as

(2), OR competitive bidding: each user u is constructed_iThe a subset selection relations of distribution can be submitted, and (i.e. OR competitive bidding distribution is closed System) task competitive bidding book, the subset selection relations of distribution indicate that expectation is assigned in task competitive bidding book comprising whole competitive bidding lists One among member or mutually multiple (that is, being the subset for the competitive bidding unit set for including in task competitive bidding book) without intersection, And obtain corresponding expectation reward；That is, the task competitive bidding book of the subset selection relations of distribution may be expressed as:

Wherein,Indicate i-th of user u_iThe task competitive bidding book of the set subset selection relations of distribution, upper right footmark O Indicating that the competitive bidding relations of distribution are that subset selects the relations of distribution, ∪ indicates that subset selects relations of distribution operator,When, haveThat is, in the case where subset selects relations of distribution arithmetic logic It is expected that be assigned in task competitive bidding book multiple competitive bidding units when, multiple competitive bidding unit is mutually without intersection；Expression task competitive bidding K-th of competitive bidding unit in book, For user u_iThe competitive bidding unit for including in set task competitive bidding book Number.

Proposition 1: when the operator number of OR competitive bidding is x, length is averagely describedOR competitive bidding descriptive power E^O(x) it can be indicated by formula (1).But it cannot indicate user that all personalized competitive biddings describe demand.

Proof: firstly, including exclusion principle, provable E using in Dynamic Programming and Combinational Mathematics^O(x) by formula (1) Indicate (reference can be made to document " R.L.Graham.Handbook of Combinatorics, volume 1.Elsevier, 1995").Then, it can prove that the description of OR competitive bidding cannot indicate that all personalized competitive biddings describe demand by counterevidence method. That is, although the competitive bidding relations of distribution of AND and OR logical combination description have certain personalization and flexibility, but not All PB (Personalized Bidding, personalized competitive bidding) description demand can completely be met.Therefore, OR competitive bidding is retouched All expression of space cannot be accommodated by stating.

In order to solve this deficiency, we supplement the expression competitive bidding relations of distribution in conjunction with AND and xor logic.

It defines 3: selecting a selection relations of distribution (the also referred to as XOR competitive bidding relations of distribution, abbreviation XOR competitive bidding) description；Based on AND It is constructed as follows with xor operation symbol:

(1), the competitive bidding of AND atom is constructed: as defined shown in AND atom competitive bidding in 2.

(2), XOR competitive bidding: each user u is constructed_iOne can be submitted to select a selection relations of distribution, and (i.e. XOR competitive bidding distributes Relationship) task competitive bidding book, it is described select a selection relations of distribution indicate only expectation be assigned in task competitive bidding book comprising all it is competing Any one among unit is marked, and obtains corresponding expectation reward；That is, selecting the task competitive bidding book of a selection relations of distribution indicates Are as follows:

Proposition 2: when the operator number of XOR is x, then averagely describing length isXOR competitive bidding describes energy Power E^X(x) it can be indicated by formula (2).Meanwhile it can indicate that all personalized competitive biddings of user describe demand.

Based on formula (2), we have further obtained E^X(x) increment is

According to formula (3), we can construct 2-D expression of space as shown in Figure 3, and wherein x-axis indicates XORs competitive bidding Quantity, y-axis indicate the increment Delta E of ability to express^X(x).Therefore, according to formula (2) and (3), 2 dimension space figures are can be used in we It goes to indicate the expression of space that XOR competitive bidding describes.(x=2 when xor operation accords with number maximum^M) two-dimentional expression of space can express All user's exposition needs.Therefore, we, which can be used in 2 dimension spaces of AND and xor operation symbol, indicates that all user's competitive biddings need It asks.

In practice, in order to trade off between user's expressivity and computational complexity, we are by limiting one-dimensional length (i.e. the quantity of XOR) (indicates) the number limitation of xor operation symbol to reduce two-dimentional expression of space with constant R.It is used in the space R The XOR competitive bidding of xor operation symbol describes method and is known as SXB method, according toIts ability to express can be obtained are as follows:

2, the description based on three-dimensional expression space.

Although XOR competitive bidding description has sufficient convincingness, there are still description length length, described the shortcomings that low efficiency. Based on above-mentioned two-dimentional expression of space, it is intended that further expand to three bit spaces by OR operator.For this purpose, based on AND, The competitive bidding relations of distribution description of XOR and OR is defined as follows.

Define 4: the mixing selection relations of distribution (the also referred to as XOR-of-OR competitive bidding relations of distribution, abbreviation XOR-of-OR competitive bidding) Description；Following competitive bidding is constructed based on AND, XOR and OR:

(1), atom competitive bidding and OR competitive bidding are constructed: as defined shown in OR competitive bidding description in 2.

(2), XOR-of-OR competitive bidding: each user u is constructed_iA mixing selection relations of distribution (i.e. OR competitive bidding can be submitted The relations of distribution) task competitive bidding book, the mixing selection relations of distribution indicate that only expectation is assigned to set multiple competitive biddings and appoints Any one among business group, includes one or more competitive bidding units in set each competitive bidding task groups, and each competing The relations of distribution of subset selection each other between each competitive bidding unit in task groups are marked, i.e., each competitive bidding task groups indicate that expectation is assigned to One among whole competitive bidding unit wherein included or mutually without the multiple of intersection；That is, the task of the mixing selection relations of distribution is competing Bidding documents indicates are as follows:

Proposition 3: XOR and OR operator number, the average description length of XOR-of-OR competitive bidding description are respectively indicated with x and y ForAbility to express E^XO(x, y) can be indicated with formula (5).Meanwhile it can indicate the competitive bidding exposition need of all users.

Wherein x ∈ 1 ..., E^O(y)},y∈{1,...,M}。

According to formula (5), E^XOThe increment of (x, y) is represented by

According to formula (5) and (6), we can construct 3-D expression of space as shown in Figure 3, wherein x-axis and y-axis difference (i.e. x) (i.e. y), also, z-axis is indicated with XOR and OR operator the quantity of expression xor operation symbol with the quantity of OR operator The increment E of several increase abilities to express^XO(x,y).Therefore, we can be used 3 dimension figures and go to indicate XOR-of-OR competitive bidding description side The expression of space of method.When XOR and OR operator number maximum, which can indicate that all user's competitive biddings are retouched State demand.Therefore, we can be with indicating all user individual competitive biddings based on 3 dimensional spaces of AND, OR and XOR.

Similar to 2 dimension spaces, we by limitation one dimension (such as: length XOR or OR), which is constant R, to be reduced 3 dimension expression of space.According to formula (5), the main dimension in entire expression of space about computation complexity is XOR.Therefore, Wo Menyong Constant R constrains xor operation and accords with number, then averagely describes lengthIts ability to express is represented by

3, personalized competitive bidding describes method design.

It can see by above-mentioned analysis, method described using XOR-of-OR competitive bidding and removes description user task competitive bidding book The competitive bidding relations of distribution are optimal selection schemes.For the ease of the operation and use of user, following user personality can be used Change competitive bidding and describe method:

(1) describe a group task: if user it is expected to be assigned a group task, his (she) is that this group task creates One atom competitive bidding, otherwise, his (she) are the one atom competitive bidding of each task creation.

(2) describe the combination of task: if the user desired that distributing any subset of certain tasks, his (she) uses OR operator Plan is created based on these atom competitive biddings.

(3) it generates the exclusive plan of R: being based on two above step, each user can iteratively be created with xor operation symbol Exclusive plan with maximum quantity limitation R.

The competitive bidding relations of distribution that method removes description user task competitive bidding book are described using XOR-of-OR competitive bidding, it can be enough Demand is described in all personalized competitive biddings of expression user, meets personalization, spirit that user distributes task competitive bidding well Active demand, while the problem that the logical description in turn avoiding the competitive bidding relations of distribution is too long, excessively complicated, it is user-friendly, The compromise between friendly is used to successfully realize user's expressivity and user.

By taking traffic route task shown in Fig. 2 as an example, the competitive bidding that each user is described with above description method is needed It asks.There are two exclusive plans by user Jack.He first in the works, Jack want in the first step, τ₁And τ₃In conjunction with rise Come, and creates atom competitive bidding { (τ₁,$15)∧(τ₃,$10)}.Because he does not have the combination of task, then there is { τ₁∧τ₃,$25}.? In the works, Jack is not intended to a τ to his second₁And τ₂It does together, is then the one atom competitive bidding of each task creation, such as (τ₁,$15),(τ₂,$35).In second step, Jack wants any subset of the two tasks, therefore (i.e. by OR operator ∪) obtain { (τ₁,$15)∪(τ₂,$35)}.Exclusive in the works at these, the individual character that XOR-of-OR describes him can be used in Jack Change competitive bidding are as follows:

B, task and reward allocation rule design.

Arbitrary i-th of user u_iThe task competitive bidding book of the mixing selection relations of distribution of setting can be formalized further Ground is expressed asWherein (τ_i,l,k,h,p_i,l,k,h) expression task, ∧, ∪,It respectively indicates The operator (i.e. AND, OR and XOR) of 3 different dependency logic relationships between task distribution.For example, in the competitive bidding of Jack In description, such as formula (8), τ₁And τ₃There is the dependence of AND, it should distribute together；τ₁And τ₂There is the dependence of OR, they Any subset can be assigned.(τ₁∧τ₃) and (τ₁∪τ₂) there is the dependence of XOR, therefore, at most one of them can be with Distribution.Therefore, the personalized competitive bidding description of user is made of many different types of complex task dependences.We use task Dependence graph indicates, as shown in figure 4, wherein circle indicates different task, Dark grey, light gray and black arrow difference Indicate AND, OR and XOR dependence.On the basis of task dependence graph, relied on by decomposing and reconfiguring for task Relational graph proposes a kind of method for allocating tasks based on dependence and adaptive critical payment calculation method.

1, the method for allocating tasks decomposed based on personalized competitive bidding description.

Since the task dependence of personalized user competitive bidding description is usually very complicated, so that directly distributing extremely difficult. In order to solve this problem, the task dependence graph that competitive bidding describes resolved into independent competitive bidding unit, even resolved into solely Vertical single idea competitive bidding (SMB) is then distributed by greedy algorithm these tasks distributing to user.That is, for establishing There is the waiting task of competitive bidding corresponding relationship, in the case where being corresponding with multiple tasks competitive bidding book, then decomposes each task respectively The competitive bidding unit for including in competitive bidding book, and distinguish according to decomposing the expectation competitive bidding task for obtaining each competitive bidding unit and it is expected to recompense The cost performance for determining each competitive bidding unit, the task competitive bidding book where selecting the optimal competitive bidding unit of cost performance are competing as getting the bid Bidding documents extracts corresponding waiting task from task library, sets up the competitive bidding relations of distribution met in the acceptance of the bid competitive bidding book and wants The task packet asked and matched with the expectation competitive bidding task in the optimal competitive bidding unit of the cost performance, distributes to the acceptance of the bid User corresponding to competitive bidding book.

In the methods of the invention, the competitive bidding point that method removes description user task competitive bidding book is described using XOR-of-OR competitive bidding It is optimal selection scheme with relationship.When the competitive bidding unit for including in task resolution competitive bidding book, distribution is selected to close for mixing The task competitive bidding book of system, isolation is:

Firstly, at least one band is added in each competitive bidding unit in the task competitive bidding book of the mixing selection relations of distribution Markd virtual task, it is ensured that the label of the virtual task added in each competitive bidding unit that each competitive bidding task groups include is mutual It is not identical, and every two belong to all include in the competitive bidding units of different competitive bidding task groups same tag virtual task, each Virtual task is idle task, and identical virtual task is marked to be considered as the same idle task, marks different virtual tasks It is considered as different idle tasks；Thus respectively by each competitive bidding cell translation be the competitive bidding selecting unit added with virtual task, The competitive bidding relations of distribution between each competitive bidding selecting unit are determined as the subset selection relations of distribution simultaneously, that is, indicate expectation distribution To one among whole competitive bidding selecting units or mutually without the multiple of intersection.

Specifically, by decomposing to XOR dependence, OR is converted by the XOR-of-OR competitive bidding description of user Competitive bidding description.For example, the virtual task of addition is indicated with D.It indicates the virtual task there is no value and cost, indirect earth's surface It is constrained up to XOR.For user u_iCompetitive bidding description inWe are eachAdd a virtual taskFor example, the competitive bidding description of Jack can be broken down into formula (8):

{(τ₁∧τ₃∧D_1,2∧D_1,3,$25)∪(τ₁∧D_1,2,$15)∪(τ₂∧D_1,3,$35)}。

Due to (τ₁∧τ₃∧D_1,2∧D_1,3) and (τ₁∧D_1,2) and (τ₂∧D_1,3) intersection, therefore, { (τ₁∧τ₃, $ 25) } and {(τ₁,$15)∪(τ₂, $ 35) } it cannot be by joint distribution.

Secondly, needing further to describe to convert with independent SMB by the competitive bidding of OR by decomposing OR dependence.For This, can add the Virtual User with independent SMB to indicate that OR is relied on, i.e., select the task competitive bidding book of the relations of distribution for mixing Corresponding user belongs to the Virtual User of the user for each competitive bidding selecting unit correspondence establishment one.

Specifically, for example, being user u_iEach ofCreation one has independent SMBVirtual User v_i,l,k。 Since the competitive bidding unit number for including in task competitive bidding book isTherefore it is directed to user u_i, the quantity of the Virtual User of creation Also it isIt is a.Therefore, user u_iCorresponding Virtual User collection is thenv_ikIt indicates to correspond to user u_i K-th of competitive bidding unit Virtual User.For example, the competitive bidding description of Jack can be broken down into three independent Virtual User, i.e., Virtual User v_1,1, it is { τ that competitive bidding, which describes SMB,₁∧τ₃∧D_1,2∧D_1,3, $ 25 }；Virtual User v_1,2, competitive bidding describes SMB For { τ₁∧D_1,2,$15}；Virtual User v_1,3, it is { τ that competitive bidding, which describes SMB,₂∧D_1,3,$35}.Such conversion is that OR is utilized Like attribute between dependence and task distribution.According to defining 4, the atom competitive bidding collection of an OR competitive bidding has independent property Matter.In other words, just as the distribution of the task of different user, different atom competitive biddings can be independently allocated to this user.

It selects the task competitive bidding book of the relations of distribution to decompose mixing as a result, and is respectively corresponding with different virtual use in order to multiple The competitive bidding selecting unit of family and the relations of distribution of subset selection each other.It will pass through the XOR- for jointly considering task competitive bidding book in this way To there are relationship in the of-OR competitive bidding relations of distribution between three kinds of logical operators the complicated task competitive bidding relations of distribution to rely on The personalized competitive bidding dismantling of (Task-dependency Graph) is schemed into multiple independent, simple single idea competitive biddings, so that Task distribution principle is converted from the complicated XOR-of-OR competitive bidding relations of distribution for relatively simple OR dependence, is reduced The complexity and operand that task distribution calculates, handle task distribution more easily, to realize user's expressivity, user Compromise between friendly, calculating validity three.

Proposition 4: there is the user individual competitive bidding of description length lambda, by most adding λ²A virtual task, can be of equal value It is converted into the λ Virtual User with λ independence SMB competitive bidding description.

Based on above-mentioned conversion, appointing for SMB competitive bidding is converted by the task optimal scheme problem that user individual competitive bidding describes Business assignment problem.This problem is proved to be np hard problem.Therefore, we obtain the task of near-optimization by greedy algorithm Distribution.

In order to determine that task distributes assessed cost performance, the concrete mode of the cost performance of determining competitive bidding unit can be designed Are as follows: for arbitrary i-th of user u_iAny k-th of competitive bidding unit for including in the task competitive bidding book of settingPass through calculating Its payment efficiencyDetermine corresponding competitive bidding unitCost performance, payment efficiency ε_i,kValue it is more big, determine Competitive bidding unitCost performance it is better；Wherein, T_i,kIndicate competitive bidding unitThe expectation competitive bidding task vector of middle record, P_i,kTable Show competitive bidding unitTotal value is recompensed in the expectation of middle record, and has T_i,k={ τ_i,k,h|h∈{1,...,H_i,k,τ_i,k,h、p_i,k,hRespectively indicate competitive bidding unitIn h-th expectation competitive bidding task and corresponding expectation Reward, h ∈ { 1 ..., H_i,k, H_i,kIndicate competitive bidding unitIn include expectation competitive bidding task total number；|T_i,k| indicate meter Calculate expectation competitive bidding task vector T_i,kIn include expectation competitive bidding task number, i.e., | T_i,k|=H_i,k。

Formally, it is directed to SMB competitive bidding (T_i,k,P_i,k) Virtual User v_i,k, payment efficiency is calculated asWherein | T_i,k| it is T_i,kNumber of tasks, and virtual task (i.e. D_i,j) contribution be 0.

And when selection determines acceptance of the bid competitive bidding book, for the task competitive bidding book of the mixing selection relations of distribution, select determination side Formula are as follows: according to the expectation competitive bidding task and expectation of each competitive bidding selecting unit in the task competitive bidding book of each mixing selection relations of distribution Reward, determines the cost performance of each competitive bidding selecting unit respectively, selects times where the optimal competitive bidding selecting unit of cost performance Business competitive bidding book is as acceptance of the bid competitive bidding book, the use that the corresponding Virtual User of the optimal competitive bidding selecting unit of the cost performance is belonged to Family is determined as the user that gets the bid, and corresponding waiting task is extracted from task library, sets up and selects with the optimal competitive bidding of the cost performance The task packet that the expectation competitive bidding task in unit matches is selected, get the bid user described in dispensing.Specifically, can be calculated by greediness Method iteratively selects the Virtual User v with maximum payment efficiency_i,k, until distributing all tasks.

2, combined reward distribution method is described based on personalized competitive bidding.

In order to enable the reward distribution of task crowdsourcing task distribution can have certain anti-tactic, need to consider user Reward distribute factor, formulation targetedly recompense allocation plan, such as can use critical payment scheme for it is each virtually User calculates the reward with anti-tactic.Specifically, any remuneration should be cannot get without the Virtual User of task distribution. On the other hand, in critical payment scheme, in the Virtual User v for calculating the task that is assigned with_i,kObtained reward total value When, need to consider Virtual User v_i,kCorresponding crucial competition userBrought competition, to improve the anti-plan of task distribution Slightly property, it may be assumed that

Wherein, Virtual User v_i,kCorresponding crucial competition userIt is determined by such as under type:

v_i,kExpression belongs to user u_iK-th of Virtual User；T_i,kIndicate Virtual User v_i,kExpectation competitive bidding task to Amount；|T_i,k| indicate expectation competitive bidding task vector T_i,kIn include expectation competitive bidding task number；It is crucial determined by indicating Compete user；Indicate crucial competition userExpectation competitive bidding task vector；Indicate crucial competition userExpectation competitive bidding task vectorIn include expectation competitive bidding task number；Indicate crucial competition user Expectation recompense total value；Indicate Virtual User v_i,kThe reward total value being paid for；ε_i,kIndicate Virtual User v_i,kPayment effect Rate, and Indicate that candidate key competes user；Indicate the payment effect of candidate key competition user Rate, and Indicate that candidate key competes userExpectation competitive bidding task vector； For i-th of user u_iThe competitive bidding unit number for including in set task competitive bidding book； It isA userThe competitive bidding unit number for including in set task competitive bidding book；N is total number of users；Indicate empty set.

Although the reward distribution based on crucial competition user ensure that anti-tactic of the user in SMB competitive bidding, it is right It cannot meet in the personalized competitive bidding distribution of user.Itself the reason is as follows that: as previously mentioned, having the user of personalized competitive bidding can be with Resolve into multiple Virtual User.One individual Virtual User cannot directly be improved by spurious report oneself income (such as: Money), but he can strategically help other Virtual User to improve income by spurious report, to improve the user's Practical total revenue.By taking Jack as an example, it will be assumed that v_1,1It is selected, and v_1,2It is v_1,1Crucial competition user.According to formula (9), The remuneration of Jack isTherefore, Jack can be from strategically spurious report P_1,2To improve total receipts of himself Benefit.

In order to solve the problems, such as this joint spurious report, the task distribution of reconfigurable user individual competitive bidding, which relies on, is closed System's figure, and design adaptive critical remuneration calculation method.Specifically, for each selected user, we are from being different from Other users of the actual user select crucial competition user, thus should be obtained according to formula (9) to calculate the user Remuneration.Formally, for user u_iK-th of Virtual User v_i,k, crucial competition user can be obtainedAre as follows:

Wherein, v_i,kExpression belongs to user u_iK-th of Virtual User；T_i,kIndicate Virtual User v_i,kExpectation competitive bidding Task vector；|T_i,k| indicate expectation competitive bidding task vector T_i,kIn include expectation competitive bidding task number；Expression determines Crucial competition user；Indicate crucial competition userExpectation competitive bidding task vector；Indicate crucial competition UserExpectation competitive bidding task vectorIn include expectation competitive bidding task number；Indicate crucial competition userExpectation recompense total value；Indicate Virtual User v_i,kThe reward total value being paid for；ε_i,kIndicate Virtual User v_i,kBranch Efficiency is paid, and Indicate that candidate key competes user；Indicate the payment of candidate key competition user Efficiency, and Indicate that candidate key competes userExpectation competitive bidding task vector； For i-th of user u_iThe competitive bidding unit number for including in set task competitive bidding book； It isA userThe competitive bidding unit number for including in set task competitive bidding book；N is total number of users；Indicate empty set.

For example, being based on formula (10), the v of Jack will not be selected after the recombination of task dependence graph_1,2It is used as key competition Family.Therefore, Jack cannot strategically improve his effectiveness.The user using above-mentioned based on critical payment recompenses dispenser as a result, System further limits the constraint dependence between task distribution, can preferably avoid selfish user appointing by tactic Business competitive bidding book specific mode carrys out the wrongful interests for obtaining more incomes and damaging task publisher and platform, so that crowdsourcing is appointed Business distribution has better anti-tactic.

3, the task distribution of special circumstances.

In addition, as a kind of special circumstances, for the waiting task for having competitive bidding corresponding relationship is established, if be only corresponding with One task competitive bidding book extracts corresponding waiting task then using the task competitive bidding book as acceptance of the bid competitive bidding book from task library, User corresponding to the acceptance of the bid competitive bidding book is distributed to, and pays corresponding reward.However, only one this user is competing The special circumstances of mark waiting task are practical to be equivalent to no competitive bidding relationship, and for intelligent perception platform, usually include A large amount of user, and these users have diversified technical ability and are distributed widely, therefore only one user's competitive bidding waits for This special circumstances of processing task are the events of very small probability.

Assessment based on actual experiment data.

In order to assess the present invention is based on the performance of the task crowdsourcing allocation processing method of personalized competitive bidding excitation, we are based on The user trajectory data of actual acquisition are (reference can be made to " https: //archive.ics.uci.edu/ml/datasets/GPS+ Trajectories, Trajectory traces.2015. "), to simulate Gigwalk application.

Actual user's track data and participating user: as shown in figure 5, we are simulated with the real trace of 200 automobiles The movement of user vehicle, and 20 positions (being indicated by light grey triangle) is randomly selected in these tracks.Each position There is the task u (1,5) with random amount.For example, there are many different tasks in shop or community.We simulate 500 A intelligent perception participant, they it is expected to complete " Gigwalk " task on their driving path.Each participant selects at random Select a starting point, such as family or the job site of his (she).The true driving that we are 3 kilometers using the starting point apart from participant Potential path of the path as his (she), and he (she) can only select a wherein paths only driving traveling.

Time and cost: each participant requires certain time to complete task, because parking lot is looked in an individual demand Parking is (reference can be made to " https: //blog.spothero.com/park-smarter-parking-search-time/.Stop Wasting time searching for parking. "), it then could complete task.According to " Stop wasting time searching for parking.https://blog.spothero.com/park-smarter-parking-search- Time/. report ", we are arranged residence time average out to 10 minutes of each place.In addition, different users have it is different The limitation of free time time, i.e. u (10,120) (min).Each user randomly chooses a limited number of positions on the path of his (she), And any one subset is selected in the set of tasks of these positions to complete.In each place, selected after user or parking Any number of task is completed or not parking any task is not done.The bidding price of each task is according to user, traveling Path and task location carry out being uniformly distributed setting, i.e. u ($ 1, $ 100).In practice, price additionally depend on task type, Platform and region.For example, tender price is from 0.10 dollar of random variation if the ceiling price that sets of platform is 10 dollar To 10 dollars.

Evaluation index and result: we compare Picasso and other method SMB, SOB and SXB in the prior art Compared with.The competitive bidding description of SMB method is comprising only including a monatomic competitive bidding, and also referred to as single idea competitive bidding is (reference can be made to document “H.Jin,L.Su,H.Xiao,and K.Nahrstedt.INCEPTION:incentivizing privacy-preserving data aggregation for mobile crowd sensing systems.In Proc.ACM MobiHoc,pages 341-350,2016. " and " M.Karaliopoulos, I.Koutsopoulos, and M.Titsias.First learn then earn:optimizing mobile crowdsensing campaigns through data-driven user pro ling.In Proc.ACM MobiHoc,pages 271–280,2016").The competitive bidding description of SOB method is comprising multiple mutual Disjoint atom competitive bidding is (reference can be made to document " Y.Chen, B.Li, and Q.Zhang.Incentivizing crowdsourcing systems with network effects.In Proc.IEEE INFOCOM,pages 1–9, 2016 " and " L.Duan, L.Huang, C.Langbort, A.Pozdnukhov, J.Walrand, and L.Zhang.Human- in-the-loop mobile networks:A survey of recent advancements.IEEE Journal on Selected Areas in Communications,35(4):813–831,2017").SXB method refers to be used in the space R The XOR competitive bidding of xor operation symbol describes method.Other than comparing user and describing efficiency ADL, we also compare these methods and complete to appoint The platform level payment (APT) of business and each user are averaged competitive bidding number of tasks (ANU) to respectively indicate the optimal of platform income Property and user's competitive bidding description ability to express.Each user can be more with the task of competitive bidding, user's ability to express of this mechanism Better.We repeat to have done 20 experiments, take its average result.As shown in fig. 6, compared with SMB and SOB, the platform of Picasso Level payment APT has reduced by more than 61%, and each user competitive bidding number of tasks ANU that is averaged at least increases 9.7 times, therefore, Picasso can greatly improve the ability to express of user and reduce platform payt.On the other hand, with SXB method phase Than although the platform level payment APT of the two and each user are averaged, competitive bidding number of tasks ANU is similar, describes efficiency in user On ADL, Picasso improves 74%, and therefore, Picasso greatly improves use under the conditions of not influencing user's ability to express Family describes efficiency.In conclusion platform payment cost can be not only greatly lowered in Picasso, but also user can be improved Internal motivation improve their ability to express, so that platform and user are benefited, the mutual benefit for realizing platform and user is double The balance of win.

In conclusion the present invention is based on the task crowdsourcing allocation processing method of personalized competitive bidding excitation, for for setting In the description design of fixed task competitive bidding book, user is allowed to come in personalized constraint task competitive bidding book using the competitive bidding relations of distribution The competitive bidding distribution principle between each competitive bidding unit for including；For the allocation processing of waiting task, then by establishing wait locate Competitive bidding corresponding relationship between reason task and task competitive bidding book set by user, from times that can satisfy the requirement of the competitive bidding relations of distribution Optimal one of the cost performance of competitive bidding unit is selected in business competitive bidding book as acceptance of the bid competitive bidding book, is set up full task packet and is distributed to pair It is handled using family, so that the distribution of waiting task is dependent on the task competitive bidding book for setting, so as to better Meet personalization, the flexibility demand that user distributes task competitive bidding, helps that intelligent perception user is motivated to participate in crowdsourcing In the processing of task, and avoids excessive, blindness the task competitive bidding book acceptance of the bid for raising its expectation reward and influence platform and appoint The interests of business supplier, have preferably taken into account the reasonability of task distribution.In addition, task crowdsourcing allocation processing method of the invention It is designed by the description design to task competitive bidding book and to the Task-decomposing of task allocation processing, is ensuring the distribution of crowdsourcing task While having personalized competitive bidding drive characteristic, too long, the excessively complicated problem of the logical description of the competitive bidding relations of distribution is avoided, It is user-friendly, and complexity and operand that task distribution calculates are reduced, handle task distribution more easily, from And realize user's expressivity, user friendly, calculate validity three between compromise, preferably taken into account function ease for use And the validity and feasibility of task allocation processing.

Finally, it is stated that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although referring to reality Example is applied to describe the invention in detail, those skilled in the art should understand that, it can be to technical side of the invention Case is modified or replaced equivalently, and without departing from the objective and range of technical solution of the present invention, should all be covered in the present invention Scope of the claims in.

Claims

1. the task crowdsourcing allocation processing method based on personalized competitive bidding excitation, which comprises the steps of:

Waiting task in release tasks library；

User sets task competitive bidding book, and each each task competitive bidding school bag set by user includes one or more competitive bidding units, with And the competitive bidding relations of distribution between competitive bidding unit；Each competitive bidding unit corresponds to one or more set by user for recording it It is expected that competitive bidding task and corresponding expectation reward；

Statistics establishes the competitive bidding corresponding relationship between the waiting task in task library and task competitive bidding book set by user, and needle There is each waiting task of competitive bidding corresponding relationship to foundation, from the task competitive bidding book that can satisfy the requirement of the competitive bidding relations of distribution Cost performance optimal one for selecting competitive bidding unit extracts waiting task as acceptance of the bid competitive bidding book from task library, sets up Meet it is described acceptance of the bid competitive bidding book in the competitive bidding relations of distribution require and it is competing with the expectation in the optimal competitive bidding unit of the cost performance The task packet that mark task matches distributes to user corresponding to the acceptance of the bid competitive bidding book, to realize there is competitive bidding pair to foundation The allocation processing for each waiting task that should be related to.

2. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that right In arbitrary i-th of user u_iThe task competitive bidding book of settingUpper right footmark * indicates competitive bidding relations of distribution label, wherein including Any k-th of competitive bidding unitIt indicates are as follows:

Wherein, N is total number of users,For the competitive bidding unit number for including in task competitive bidding book, T_i,k、P_i,kRespectively indicate competitive bidding UnitTotal value is recompensed in the expectation competitive bidding task vector of middle record and expectation, and has T_i,k={ τ_i,k,h|h∈{1,...,H_i,k,τ_i,k,h、p_i,k,hRespectively indicate competitive bidding unitIn h-th expectation competitive bidding task and corresponding expectation Reward, h ∈ { 1 ..., H_i,k, H_i,kIndicate competitive bidding unitIn include expectation competitive bidding task total number, and it is expected competitive bidding appoint Business vector T_i,kIt is the subset of task library T, that is, hasτ_i,k,h∈T。

3. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that institute Stating the competitive bidding relations of distribution includes the subset selection relations of distribution, and the subset selection relations of distribution indicate that expectation is assigned to task competitive bidding Comprising one among whole competitive bidding units or mutually without the multiple of intersection in book；That is, the task competitive bidding of the subset selection relations of distribution Book indicates are as follows:

Wherein,Indicate i-th of user u_iThe task competitive bidding book of the set subset selection relations of distribution, upper right footmark O are indicated The competitive bidding relations of distribution are that subset selects the relations of distribution, and ∪ indicates that subset selects relations of distribution operator；Expression task competitive bidding book In k-th of competitive bidding unit, For user u_iThe competitive bidding unit for including in set task competitive bidding book Number.

4. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that institute Stating the competitive bidding relations of distribution includes selecting a selection relations of distribution, and a selection relations of distribution of selecting indicate that only it is competing to be assigned to task for expectation Include any one among whole competitive bidding units in bidding documents；That is, selecting the task competitive bidding book of a selection relations of distribution indicates are as follows:

Wherein,Indicate i-th of user u_iThe set task competitive bidding book for selecting a selection relations of distribution, upper right footmark o are indicated The competitive bidding relations of distribution are to select a selection relations of distribution,To select a selection relations of distribution operator, expression only distributes its connection pair Any one as among；K-th of competitive bidding unit in expression task competitive bidding book, For user u_iInstitute The competitive bidding unit number for including in the task competitive bidding book of setting.

5. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that institute Stating the competitive bidding relations of distribution includes the mixing selection relations of distribution, and it is set that the mixing selection relations of distribution indicate that only expectation is assigned to Multiple competitive bidding task groups among any one, include one or more competitive bidding lists in set each competitive bidding task groups Member, and subset selects the relations of distribution, i.e., each competitive bidding task groups table each other between each competitive bidding unit in each competitive bidding task groups Show that expectation is assigned among whole competitive bidding unit wherein included one or mutually without the multiple of intersection；That is, mixing selection distribution The task competitive bidding book of relationship indicates are as follows:

Wherein,Indicate i-th of user u_iThe task competitive bidding book of the set mixing selection relations of distribution, upper right footmark XO are indicated The competitive bidding relations of distribution are the mixing selection relations of distribution,To select a selection relations of distribution operator, expression only distributes its connection pair Any one as among, ∪ indicate that subset selects relations of distribution operator；Indicate that the task of the mixing selection relations of distribution is competing Bidding documentsIn set first of competitive bidding task groups, andL_iTask competitive bidding bookMiddle packet The competitive bidding task groups sum contained,Expression task competitive bidding bookIn n-th in set first of competitive bidding task groups it is competing Mark unit, n ∈ { 1 ..., K_i,l, and Expression task competitive bidding book In k-th of competitive bidding unit,For user u_iThe competitive bidding unit number for including in set task competitive bidding book.

6. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that needle There is the waiting task of competitive bidding corresponding relationship that it is competing to decompose each task respectively if being corresponding with multiple tasks competitive bidding book foundation The competitive bidding unit for including in bidding documents, and the expectation competitive bidding task of each competitive bidding unit is obtained according to decomposition and it is expected to recompense true respectively The cost performance of fixed each competitive bidding unit, the task competitive bidding book where selecting the optimal competitive bidding unit of cost performance are used as acceptance of the bid competitive bidding Book, extracts corresponding waiting task from task library, set up the competitive bidding relations of distribution met in the acceptance of the bid competitive bidding book require, And the task packet to match with the expectation competitive bidding task in the optimal competitive bidding unit of the cost performance, distribute to the acceptance of the bid competitive bidding User corresponding to book.

7. the task crowdsourcing allocation processing method according to claim 6 based on personalized competitive bidding excitation, which is characterized in that institute Stating the competitive bidding relations of distribution includes the mixing selection relations of distribution, and it is set that the mixing selection relations of distribution indicate that only expectation is assigned to Multiple competitive bidding task groups among any one, include one or more competitive bidding lists in set each competitive bidding task groups Member, and subset selects the relations of distribution, i.e., each competitive bidding task groups table each other between each competitive bidding unit in each competitive bidding task groups Show expectation be assigned to whole competitive bidding unit wherein included one or more of；

When the competitive bidding unit for including in task resolution competitive bidding book, for the task competitive bidding book of the mixing selection relations of distribution, decompose Mode specifically:

It selects to add at least one with markd in each competitive bidding unit in the task competitive bidding book of the relations of distribution for mixing Virtual task, it is ensured that the label of the virtual task added in each competitive bidding unit that each competitive bidding task groups include is different, And every two belong to all include in the competitive bidding units of different competitive bidding task groups same tag virtual task, each virtual task It is idle task, and identical virtual task is marked to be considered as the same idle task, different virtual tasks is marked to be considered as not Same idle task；It thus by each competitive bidding cell translation is respectively the competitive bidding selecting unit added with virtual task, while it will be each The competitive bidding relations of distribution between a competitive bidding selecting unit are determined as the subset selection relations of distribution, that is, it is all competing to indicate that expectation is assigned to Mark selecting unit among one or mutually without the multiple of intersection；

User corresponding to task competitive bidding book for the mixing selection relations of distribution is each competitive bidding selecting unit correspondence establishment one A Virtual User for belonging to the user；

As a result, will mixing selection the relations of distribution task competitive bidding book decompose in order to it is multiple be respectively corresponding with different Virtual User and The competitive bidding selecting unit of the relations of distribution of subset selection each other.

8. the task crowdsourcing allocation processing method according to claim 7 based on personalized competitive bidding excitation, which is characterized in that When selection determines acceptance of the bid competitive bidding book, for the task competitive bidding book of the mixing selection relations of distribution, method of determination is selected are as follows: according to each mixed The expectation competitive bidding task of each competitive bidding selecting unit and expectation reward in the task competitive bidding book of the selection relations of distribution are closed, is determined respectively The cost performance of each competitive bidding selecting unit is selected during the task competitive bidding book where the optimal competitive bidding selecting unit of cost performance is used as Competitive bidding book is marked, the user that the corresponding Virtual User of the optimal competitive bidding selecting unit of the cost performance is belonged to is determined as acceptance of the bid and is used Corresponding waiting task is extracted at family from task library, sets up and the expectation in the optimal competitive bidding selecting unit of the cost performance The task packet that competitive bidding task matches, get the bid described in dispensing user.

9. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that really Determine the concrete mode of the cost performance of competitive bidding unit are as follows: for arbitrary i-th of user u_iInclude in the task competitive bidding book of setting Any k-th of competitive bidding unitBy calculating its payment efficiencyDetermine corresponding competitive bidding unitSexual valence Than payment efficiency ε_i,kValue it is more big, determine competitive bidding unitCost performance it is better；Wherein, T_i,kIndicate competitive bidding unitIn The expectation competitive bidding task vector of record, P_i,kIndicate competitive bidding unitTotal value is recompensed in the expectation of middle record, and has T_i,_k={ τ_i,k,h |h∈{1,...,H_i,k,τ_i,k,h、p_i,k,hRespectively indicate competitive bidding unitIn h-th of expectation competitive bidding Task and corresponding expectation reward, h ∈ { 1 ..., H_i,k, H_i,kIndicate competitive bidding unitIn include expectation competitive bidding task it is total Number；|T_i,k| it indicates to calculate expectation competitive bidding task vector T_i,kIn include expectation competitive bidding task number, i.e., | T_i,_k|=H_i,k。

10. the task crowdsourcing allocation processing method according to claim 1 based on personalized competitive bidding excitation, which is characterized in that For the waiting task for having competitive bidding corresponding relationship is established, if being only corresponding with a task competitive bidding book, by the task competitive bidding book As acceptance of the bid competitive bidding book, corresponding waiting task is extracted from task library, distributes to use corresponding to the acceptance of the bid competitive bidding book Family.