CN111464620B - Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof - Google Patents

Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof Download PDF

Info

Publication number
CN111464620B
CN111464620B CN202010236632.3A CN202010236632A CN111464620B CN 111464620 B CN111464620 B CN 111464620B CN 202010236632 A CN202010236632 A CN 202010236632A CN 111464620 B CN111464620 B CN 111464620B
Authority
CN
China
Prior art keywords
user
edge
edge cloud
cloud
true value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010236632.3A
Other languages
Chinese (zh)
Other versions
CN111464620A (en
Inventor
徐佳
杨尚书
丁玉青
周远航
钱一航
徐力杰
鲁蔚锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202010236632.3A priority Critical patent/CN111464620B/en
Publication of CN111464620A publication Critical patent/CN111464620A/en
Application granted granted Critical
Publication of CN111464620B publication Critical patent/CN111464620B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention discloses an edge-assisted mobile crowd sensing true value discovery system and an excitation method thereof, and particularly comprises a true value discovery stage and a budget feasible reverse auction stage. In the true value discovery stage, the invention performs true value discovery on two layers of deep cloud and edge cloud. In the budget feasible reverse auction phase, the present invention proposes a greedy approach to select users under budget constraints to maximize the quality function. The method provided by the invention is effective in calculation, real, feasible in budget and superior to the similar method in terms of true value discovery precision and system expandability, and ensures constant approximation.

Description

Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof
Technical Field
The invention relates to an edge-assisted mobile crowd sensing true value discovery system and an excitation method thereof.
Background
Mobile crowd sensing is a human driven activity that exploits the popularity of wireless connections to enable a variety of mobile devices with built-in sensing capabilities, as well as inherent user mobility, to create dense and dynamic datasets that can effectively characterize our environmental information. Mobile crowd sensing has become an effective method of data acquisition in sensing applications such as photo selection, public bike travel selection, and indoor positioning systems, among others.
Most existing mobile crowd sensing systems rely on cloud services to accomplish the tasks of collecting/aggregating sensory data, distributing tasks, estimating truth values, and motivating mobile users. Cloud-based mobile crowd-sourced awareness systems have some significant drawbacks, such as poor scalability of crowd-sourced awareness due to computation and communication congestion on cloud servers, difficulty in identifying false locations of awareness data, and the attendant risk of higher data security and exposure to user privacy.
The mobile crowd sensing system based on the edge computing architecture has the following main advantages: the computational complexity is reduced: the edge computing-based mobile crowd-aware architecture may parallelize computing by offloading computing from the cloud to multiple edge servers; reducing delay: little or no communication between the cloud and the mobile user; location awareness: most mobile crowd-sourced tasks depend on location. Edge computing resources (e.g., base stations, access points) are typically located at specific locations: since the edge server collects the sensed data only in its deployment area, it is easy to verify the location attribute of the sensed data; flexible data processing: mobile crowd sensing based on edge computation gives the edge server flexibility in local data processing (e.g., aggregation, truth discovery and temperature, noise level, transportation, air condition inference for specific areas); privacy threat is reduced: the sensor data is distributed among a plurality of edge servers. The distributed storage of the sensed data in the plurality of edge servers not only enhances the security of the data, but also reduces the privacy threat of the user.
Some related studies on mobile crowd sensing based on edge computation have been conducted, which use edge computation to achieve or solve certain target problems, such as edge computation based data processing, privacy protection, reputation management, task offloading of vehicle crowd-sourced applications, and extraction of environmental information, etc. However, there is no truth-finding and motivating method design in the prior art.
Disclosure of Invention
The invention aims to: the invention aims to provide an edge-assisted mobile crowd sensing true value finding system, wherein a mobile crowd sensing task with budget is distributed to each edge cloud; users interested in executing tasks submit their bids along with the sensed data to the respective edge clouds for reference, each edge cloud performing true value discovery for each task.
It is another object of the present invention to provide a method of incentive for an edge-assisted mobile crowd-aware truth-value discovery system, each edge cloud selecting a subset of users as winners to maximize the quality of the winners under budget constraints and to determine the rewards paid to the winners.
The technical scheme is as follows: an edge-assisted mobile crowd sensing true value discovery system is applicable to a mobile crowd sensing system, and is characterized in that: the truth value discovery system is positioned on a platform in the deep cloud, and the platform comprises a set E= { E of r edge clouds 1 ,e 2 ,...,e r And a set of n smartphone users, set u= {1,2,..n }, set U k Is edge cloud e k User set within coverage area, where u=uu k K=1, 2, r, any user belongs to multiple edge clouds;
deep cloud will m tasks t= { T 1 ,t 2 ,...,t m Assigning to all edge clouds, performing mobile crowd sensing, and determining budget for each edge cloud according to importance of sensing data in edge cloud coverage area, so that G= (G) 1 ,G 2 ,...,G r ) For budget overview of all edge clouds, each task t j E T with a task type mu j Associated, the type represents task t j Is of importance of (2); let μ= (μ) 1 ,μ 2 ,...,μ m ) Is the type of all tasks;
each edge cloud e k E recording tasks and assigning them to user subset U k Each user i epsilon U submits a triplet B to the edge cloud to which the user i epsilon U belongs i ＝(T i ,b i ,X i )，T i Is the set of tasks that user i is willing to perform, b i Is the bid of user i, X i Is submitted by user i and is associated with task set T i Corresponding perception data, each task set T i And cost c i Associated costs refer to the cost of submitting data, such as the computational power spent, storage space, time, traffic, etc., for user i to perceive the data; c i Known only to user i, user submits task T i Let x= (X) 1 ,X 2 ,...,X n ) The sensory data submitted for all users is presented,
Figure BDA0002431213250000021
let T be k And X is k Respectively, to edge cloud e k Task and perception data submitted;
upon receiving the perception data, each edge cloud e k E calculates each i E U k Weights w of users of (2) i Estimating true values for all perceived tasks through true value discovery
Figure BDA0002431213250000022
Let w be k E is k All users U in (1) k Weight of (1) is set to X * The true value estimated for all final edge clouds.
Further, the cloud e at any edge k The steps of the above truth-value discovery phase are as follows:
step 201: the user collects the sensing data and submits the sensing data to an edge cloud server where the user is located;
step 202: random initialization task true value of edge cloud server
Figure BDA0002431213250000023
Upper iteration limit +_>
Figure BDA0002431213250000024
Step 203: edge Yun Diedai counter initialization k=0;
step 204: will be on edge cloud
Figure BDA0002431213250000031
The value of +.>
Figure BDA0002431213250000032
Step 205: calculating all user i E U k Weights of (2)
Figure BDA0002431213250000033
Wherein the method comprises the steps of
Figure BDA0002431213250000034
std j Is task t j Standard deviation of all perceived data of (2);
step 206: calculate all tasks t j ∈T k True value of (2)
Figure BDA0002431213250000035
Step 207: updating the current iteration number K=K+1;
step 208: if it is
Figure BDA0002431213250000036
And->
Figure BDA0002431213250000037
Step 204 is performed, otherwise step 209 is performed;
step 209: outputting true values of all tasks estimated
Figure BDA0002431213250000038
And weight w of all users k 。
Further, the edge cloud performs budget reverse auction, and a task set T is given k User set U k Budget G k Task type μ and bid strategy
Figure BDA0002431213250000039
Each edge cloud e k Computing winner set +.>
Figure BDA00024312132500000310
And each winner i.epsilon.S k Payment p of (2) i Let->
Figure BDA00024312132500000311
Let p k And p is S respectively k And a payment policy of S,
the utility of any user i is defined as the difference between the payment and its actual cost:
u i ＝p i -c i (1)
for any task t j ∈T i From edge cloud e k Winner set S in (1) k The obtained quality function is defined as:
Figure BDA00024312132500000312
wherein log reflects the diminishing returns of the platform in terms of quality brought by the participating users;
maximizing the quality function so that the total payment does not exceed the budget, the present invention refers to this problem as a budget feasible quality optimization problem, expressed as:
target maximaze V (S) k ) (3)
Constraint:
Figure BDA0002431213250000041
further, the budget feasible reverse auction stage comprises the following steps:
step 301: initializing a winner set
Figure BDA0002431213250000042
Initializing a payment policy p k ＝0；
Step 302: initializing a set of users whose bids do not exceed a budget
Figure BDA0002431213250000043
Step 303: let i * Is that
Figure BDA0002431213250000044
The most valuable users in the collection: />
Figure BDA0002431213250000045
Step 304: step 305 is performed with 2/5 probability and step 306 is performed with 3/5 probability;
step 305: will i * Joining to a winner set S k In (2) and set user i * Is paid as
Figure BDA0002431213250000046
Step 317 is performed;
step 306: at the position of
Figure BDA0002431213250000047
Find->
Figure BDA0002431213250000048
User i with the greatest value, wherein>
Figure BDA0002431213250000049
V i' (S k )＝V(S k ∪{i'})＝V(S k )；
Step 307: if it meets
Figure BDA00024312132500000410
Step 308 is performed, otherwise step 310 is performed;
step 308: adding user i to the winner set S k In (a) and (b);
step 309: at the position of
Figure BDA00024312132500000411
Find->
Figure BDA00024312132500000412
User i with the greatest value, wherein>
Figure BDA00024312132500000413
For the collection->
Figure BDA00024312132500000414
And set S k Is performed in step 307;
step 310: judge pair set S k Whether or not each user i has performed steps 311 to 316; if so, step 317 is performed;
step 311: order the
Figure BDA00024312132500000415
Let->
Figure BDA00024312132500000416
Step 312: finding collections
Figure BDA00024312132500000417
Middle->
Figure BDA00024312132500000418
User i' with the greatest value, wherein>
Figure BDA00024312132500000419
Step 313: if user i' satisfies
Figure BDA00024312132500000420
Step 314 is performed, otherwise step 310 is performed;
step 314: finding collections
Figure BDA00024312132500000421
Middle->
Figure BDA00024312132500000422
User i' with the greatest value, wherein>
Figure BDA00024312132500000423
Step 315: calculation of
Figure BDA00024312132500000424
/>
Step 316: and adding i 'to S' k In (a): s'. k ＝S' k Step 313 is performed for U { i' };
step 317: output winner set S k Payment policy p k 。
Further, the edge cloud submits the estimated true values to the deep cloud, and the deep cloud calculates each edge cloud e k Weight g of E k When the edge isCloud e k The formula of the weight update of (2) is:
Figure BDA0002431213250000051
wherein the method comprises the steps of
Figure BDA0002431213250000052
Alpha and beta are two constants, alpha is a reliability coefficient, beta is an importance coefficient, alpha is an influence of the number of people of the edge cloud on the final true value discovery, the edge cloud with more people has larger weight when alpha is larger, the edge cloud with more budgets has larger weight when the true value discovery of the deep cloud is larger, and the edge cloud with more budgets has larger weight when the true value discovery of the deep cloud is larger, wherein alpha epsilon [0,1],β∈[0,1]α+β=1; finally according to X * Estimating final true values X for all perceived tasks by means of a truth-value discovery method ** 。
Furthermore, the edge-assisted mobile crowd sensing true value discovery excitation method is budget-feasible.
And (3) proving: is provided with
Figure BDA0002431213250000053
And->
Figure BDA0002431213250000054
For all i.epsilon.S k' \S k Assume that there is
Figure BDA0002431213250000055
Then add all inequalities together, have +.>
Figure BDA0002431213250000056
Equivalent to->
Figure BDA0002431213250000057
So that the assumption is not true at the beginning, there is +.>
Figure BDA0002431213250000058
Now set S 0 Is an empty set, S 1 For only one user, and so on. Assuming the presence of a user
Figure BDA0002431213250000059
Can bid
Figure BDA00024312132500000510
Still becomes the winner (user j originally bid for b j ) At this point j increases the bid to b' j Others remain unchanged.
It can be found that
Figure BDA00024312132500000511
Thus j remains in winner set S j-1 Is a kind of medium.
The combination selected before j is included in the winning combination is denoted by S. Thus, there are
Figure BDA0002431213250000061
Figure BDA0002431213250000062
It can be assumed that
Figure BDA0002431213250000063
This is indeed the case, otherwise S ∈ { j } =s k U is U-shaped
Figure BDA0002431213250000064
/>
Thus there is
Figure BDA0002431213250000065
Let r=s k S, applying equation (5) and the above equation can be obtained:
for user r 0 ∈R\{ j }, have
Figure BDA0002431213250000066
Is known to be
Figure BDA0002431213250000067
And then get->
Figure BDA0002431213250000068
Combining the inequality to obtain
Figure BDA0002431213250000069
Can be obtained
b'(S k ∪S)-b'(S∪{j})＝b'(R\{j})＝b(R\{j})≤b(S k )。
Because of
Figure BDA00024312132500000610
For->
Figure BDA00024312132500000611
All are true, thus there is->
Figure BDA00024312132500000612
And->
Figure BDA00024312132500000613
Can obtain
Figure BDA00024312132500000614
Thus V (S) k ) < 2V (S.u.j.), combining inequality (6), finally obtaining
To:
Figure BDA00024312132500000615
i.e. < ->
Figure BDA00024312132500000616
So that the quotations of all users are added up to be less than or equal to G k 。
Furthermore, the edge-assisted mobile crowd sensing true value discovery excitation method is computationally efficient. And (3) proving: first, the truth value on the edge cloud is analyzed to find the phase time complexity. Updating the weights of all users in each edge cloud (step 205) takes time O (nm). Updating the true phase of all tasks in each edge cloud (step 206) requires time O (nm). The number of iterations (steps 203-208) is at most
Figure BDA00024312132500000617
Thus, the run time of the truth-finding phase is O (nm). The time complexity of true value discovery in deep clouds is the same as the time complexity of true value discovery on edge clouds. Since the entire true phase discovery requires time O (nm).
Next, the time complexity of the budget viable reverse auction is analyzed. Only the time complexity of the random method 3/5 probability branch (steps 306-317) needs to be analyzed as it dominates the run time of the budget feasible reverse auction. Finding the user with the greatest marginal benefit takes time O (nm) to calculate V i (S k ) The time spent takes O (m). Since there are m tasks, each winner should contribute at least one new task, the number of winners is at most m. Thus, the cycle of steps 307-309 requires time O (nm 2 ). In each iteration of the loop of steps 311-316, a process similar to steps 307-309 is performed. Thus, the payment determination takes time O (nm 3 ). The run time of the budget viable reverse auction phase is controlled by the payment determination phase, so the time complexity of the budget viable reverse auction phase is O (nm 3 )。
Furthermore, the edge-assisted mobile crowd sensing true value discovery excitation method is real.
And (3) proving: to prove that the excitation method of the present invention is realistic, it is to be proved that the method complies with the melsen theorem. First, monotonicity is demonstrated, at 2/5 probability, the user who directly selects the largest contribution as the winner, whatever his bid is, and at 3/5 probability, we useThe idea of greedy algorithm is to select winners, who submit lower offers, which will cause them to
Figure BDA0002431213250000076
Larger, top ranking, still will be the winner, so the method satisfies the monotonicity of melsen's theorem.
Next, the threshold payment is demonstrated, and under 2/5 probability, the user with the greatest contribution is directly selected as the winner, and his entire budget G is directly paid k As a compensation payment, a winning user cannot bring more benefits to himself even if changing own quotation, and under the 3/5 probability, winners are selected by adopting the thought of greedy algorithm, and in the payment stage
Figure BDA0002431213250000072
To determine the payment paid to user i, when winning bid b i ≤p i At this time, a lower bid may still make i the winner, when b i ＞p i There are two cases when:
(1) When (when)
Figure BDA0002431213250000073
The winner will not be reached due to the violation of the conditions of step 307.
(2) When (when)
Figure BDA0002431213250000074
Due to presence->
Figure BDA0002431213250000075
I ' will replace i ' in position so that i's return is reduced and even cannot be the winner, so the user cannot increase his own return by changing his own price.
The threshold payment in compliance with melsen theorem.
In summary, the excitation method of the present invention is realistic.
Furthermore, the approximate ratio of the edge-assisted mobile crowd sensing true value discovery excitation method is 1/5.
And (3) proving: set S * For optimal solution, there are, due to the sub-model and monotonicity of the cost function V ():
Figure BDA0002431213250000081
at this time take
Figure BDA0002431213250000082
There may be->
Figure BDA0002431213250000083
Since k is not in the winning set, there is +.>
Figure BDA0002431213250000084
As can be derived from the above formula,
Figure BDA0002431213250000085
so can obtain 2V (S k )≤2(V(S k-1 )+V({k}))≤2(V(S k-1 )+V({i * })). Let S be the final result of the method in order to prove the approximation ratio. The method adopts greedy algorithm to obtain S with 3/5 probability k With a probability of 2/5, i is directly obtained by the optimal single-user scheme * This result. Thus, there are:
Figure BDA0002431213250000086
wherein E (V (S)) is V (S), can be obtained
Figure BDA0002431213250000087
The beneficial effects are that:
compared with the prior art, the edge-assisted mobile crowd sensing true value discovery excitation method has the following advantages:
1. the invention distributes the mobile crowd sensing task with budget to each edge cloud; the truth value estimation can be carried out on regional crowd sensing data, and the pertinence of the truth value estimation is improved.
2. The present invention selects a subset of users as winners for each edge cloud with increased accuracy in true value estimation accuracy to maximize the quality of the winners under budget constraints and to determine the compensation paid to the winners. The calculation cost is low, and the value obtained by the platform is at least not lower than 1/5 of the optimal solution.
3. The truth value estimation is used as an evaluation method of the crowd sensing data quality, and the crowd sensing data quality is improved.
Drawings
FIG. 1 is a schematic diagram of an edge-aided mobile crowd sensing truth value discovery system;
FIG. 2 is a truth-finding flow chart;
FIG. 3 is a budget viable reverse auction flow diagram.
Detailed Description
The invention relates to an edge-assisted mobile crowd sensing true value discovery system and an excitation method thereof, which comprise an edge-assisted mobile crowd sensing temperature detection system, and fig. 1 is a schematic diagram of the edge-assisted mobile crowd sensing true value discovery system.
In a mobile crowd-aware temperature detection system, the system is located on a platform in a deep cloud, the platform comprises a set E= { E of r edge clouds 1 ,e 2 ,...,e r And a set of n smartphone users interested in performing tasks, u= {1,2,., n }; set U k Is edge cloud e k User set within coverage area, where u=uu k ,k＝1,2,...,r。
The platform first detects m temperature tasks t= { T 1 ,t 2 ,...,t m Assigned to all edge clouds for mobile crowd sensing. Each edge cloud has a budget that is determined by the importance of the sensed data in the edge cloud coverage area. Let g= (G 1 ,G 2 ,...,G r ) For all sidesBudget profile of the edge cloud. Each task t j E T are all associated with a task type mu j And (5) associating. Let μ= (μ) 1 ,μ 2 ,...,μ m ) Is the type of all tasks.
Each edge cloud e k E records the task and assigns it to user subset U k . Each user i epsilon U submits a triplet B to the edge cloud to which it belongs i ＝(T i ,b i ,X i )，X i Is the perceived temperature of user i. Let x= (X) 1 ,X 2 ,...,X n ) Perceptual data submitted for all users, wherein
Figure BDA0002431213250000091
Let T be k And X is k Respectively, to edge cloud e k Submitted tasks and awareness data.
Upon receiving the perception data, each edge cloud e k E calculates each i E U k Weights w of users of (2) i And estimating true values for all perceived tasks through true value discovery
Figure BDA0002431213250000092
Let w be k E is k All users U in (1) k Is a weight of (2). Set X * The true value estimated for all final edge clouds.
Alternatively, the edge cloud may submit the estimated true values to the deep cloud. Deep cloud computing each edge cloud e k Weight g of E k And according to X * Estimating final true values X for all perceived tasks by means of a truth-value discovery method ** . Finally, each edge cloud performs a budget-feasible reverse auction. Each edge cloud e k Computing winner set
Figure BDA0002431213250000093
And each winner i.epsilon.S k Payment p of (2) i . Let->
Figure BDA0002431213250000094
Let p k And p are respectivelyIs S k And a payment policy of S.
For any task t j ∈T i The invention will be from edge cloud e k Winner set S in (1) k The obtained quality function is defined as:
Figure BDA0002431213250000095
further, the edge-assisted mobile crowd-sensing truth value discovery excitation method comprises two stages, namely a truth value discovery stage and a budget feasible reverse auction stage, wherein the truth value discovery stage and the budget feasible reverse auction stage are respectively carried out on any edge cloud e k The flow of the above truth value discovery phase is shown in fig. 2, and the steps are as follows:
simply simulate the execution on one edge cloud, assuming one edge cloud e k There are 4 users u i (i=1, 2,3, 4) to complete two tasks t 1 ,t 2 。
Step 201:4 users collect two temperature data through sensors
Figure BDA0002431213250000101
And submitting the sensing data to the edge cloud e where the user is located k Is a server of (1);
step 202: random initialization task true value of edge cloud server
Figure BDA0002431213250000102
Upper iteration limit +_>
Figure BDA0002431213250000103
Step 203: edge Yun Diedai counter initialization k=0;
step 204: will be on edge cloud
Figure BDA0002431213250000104
The value of +.>
Figure BDA0002431213250000105
Step 205: calculating all user i E U k Weights of (2)
Figure BDA0002431213250000106
Wherein the method comprises the steps of
Figure BDA0002431213250000107
std j Is task t j Standard deviation of all perceived data of (a).
From the formula w can be obtained 1 ＝4.3982，w 2 ＝3.3692，w 3 ＝2.5392，w 4 ＝0.5392。
Step 206: calculate all tasks t j ∈T k True value of (2)
Figure BDA0002431213250000108
Step 207: updating the current iteration number K=K+1;
step 208: if it is
Figure BDA0002431213250000109
And->
Figure BDA00024312132500001010
Step 204 is performed, otherwise step 209 is performed;
step 209: outputting true values of all tasks estimated
Figure BDA00024312132500001011
And weight w of all users k ；
Wherein the method comprises the steps of
Figure BDA00024312132500001012
w 1 ＝3.3986，w 2 ＝10.7871，w 3 ＝8.5912，w 4 = 3.1094. And (3) users on other edge clouds are the same, and finally, all the edge clouds are summarized by using the same algorithm to obtain a final temperature true value.
Further, an edge-assisted movement according to the inventionThe crowd-aware truth-value discovery incentive method also comprises a budget-enabled reverse auction stage for simulating a bid price b for each user 1 ＝5,b 2 ＝4,b 3 ＝4,b 4 =6, edge cloud e k The budget obtained is 10, mu 1 ＝1，μ 2 ＝2。
The flow is as shown in fig. 3, and the steps are as follows:
step 301: initializing a winner set
Figure BDA00024312132500001013
Initializing a payment policy p k ＝0；
Step 302: initializing a set of users whose bids do not exceed a budget
Figure BDA00024312132500001014
So that
Figure BDA00024312132500001015
/>
Step 303: let i * Is that
Figure BDA00024312132500001016
The most valuable users in the collection: />
Figure BDA00024312132500001017
V values were calculated by w for 4 users as 4.4326, 8.7678, 7.5219, 5.9214, respectively. So the largest user is u 2 ；
Step 304: step 305 is performed with a 2/5 probability, step 306 is performed with a 3/5 probability, and for the sake of showing the integrity of the algorithm, it is assumed here that step 306 is performed with a 3/5 probability;
step 305: will i * Joining to a winner set S k In (2) and set user i * Is paid as
Figure BDA0002431213250000111
Step 317 is performed;
step 306: at the position of
Figure BDA0002431213250000112
Find->
Figure BDA0002431213250000113
User i with the greatest value, wherein>
Figure BDA0002431213250000114
V i' (S k )＝V(S k ∪{i'})-V(S k ) Of the current 4 users, u 2 Is->
Figure BDA0002431213250000115
Is->
Figure BDA0002431213250000116
Is the largest.
Step 307: if it meets
Figure BDA0002431213250000117
Step 308 is performed, otherwise step 310 is performed, it is apparent that user u 2 Meets this constraint;
step 308: adding user 2 to the winner set S k In (a) and (b);
step 309: at the position of
Figure BDA0002431213250000118
Find->
Figure BDA0002431213250000119
User i with the greatest value, wherein>
Figure BDA00024312132500001110
For the collection->
Figure BDA00024312132500001111
And set S k If step 307 is performed, the final set S is found k ＝{2,3}
Step 310: judge pair set S k Whether or not each user i performs a stepStep 311 to step 316; if so, step 317 is performed;
step 311: first calculate the reward of user 2, let
Figure BDA00024312132500001112
Let->
Figure BDA00024312132500001113
At this time->
Figure BDA00024312132500001114
Step 312: finding collections
Figure BDA00024312132500001115
Middle->
Figure BDA00024312132500001116
User i' with the greatest value, wherein>
Figure BDA00024312132500001117
At this point, user 3.
Step 313: if user i' satisfies
Figure BDA00024312132500001118
Step 314 is performed and otherwise step 310 is performed, where user 3 is in compliance with this constraint.
Step 314: calculation of
Figure BDA00024312132500001119
Step 315: and adds user 3 to S' k In (a): s'. k ＝S' k Step 313 is performed, and after one more round, user 1 is added to S' k Thereby obtaining p 2 ＝4.66。
Step 316: similarly, the consideration p of the user 3 is calculated by the loop 3 ＝4.19
Step 317: output winner set S k = {2,3}, pay policy p k ＝{p 2 ＝4.66，p 3 ＝4.19}。

Claims (2)

1. An edge-assisted mobile crowd sensing true value discovery system is applicable to a mobile crowd sensing system, and is characterized in that: the edge-assisted mobile crowd sensing true value finding system is positioned on a platform in a deep cloud, and the platform comprises a group of r edge clouds which form a set E= { E 1 ,e 2 ,...,e r And a set of n smartphone users, u= {1,2,..n }, set U k Is edge cloud e k User set in coverage area, where u= U k K=1, 2, r, any user belongs to multiple edge clouds;
deep cloud will m tasks t= { T 1 ,t 2 ,...,t m Assigning to all edge clouds, performing mobile crowd sensing, and determining budget for each edge cloud according to importance of sensing data in edge cloud coverage area, so that G= (G) 1 ,G 2 ,...,G r ) For budget overview of all edge clouds, each task t j E T with a task type mu j Associated, the type represents task t j Is of importance of (2); let μ= (μ) 1 ,μ 2 ,...,μ m ) Is the type of all tasks;
each edge cloud e k E recording tasks and assigning them to user subset U k Each user i epsilon U submits a triplet B to the edge cloud to which the user i epsilon U belongs i ＝(T i ,b i ,X i )，T i Is the set of tasks that user i is willing to perform, b i Is the bid of user i, X i Is submitted by user i and is associated with task set T i Corresponding perception data, each task set T i And cost c i Associated, c i Known only to user i, user submits task T i Let x= (X) 1 ,X 2 ,...,X n ) Perceptual data submitted for all users, wherein
Figure QLYQS_1
Let T be k And X is k Respectively, to edge cloud e k A submitted task set and awareness data;
upon receiving the perception data, each edge cloud e k E calculates each i E U k Weights w of users of (2) i Estimating edge cloud e by truth-value discovery algorithm k True value of all tasks of (a)
Figure QLYQS_2
Let w be k E is k All users U in (1) k Weight of (1) is set to X * True values of all final edge cloud estimates;
for arbitrary edge cloud e k The steps of the truth value discovery algorithm are as follows:
step 201: the user collects the perception data and submits the perception data to the edge cloud where the user is located;
step 202: edge cloud randomly initializing edge cloud e k True value of all tasks of (a)
Figure QLYQS_3
Upper iteration limit +_>
Figure QLYQS_4
Step 203: edge Yun Diedai counter initialization
Figure QLYQS_5
Step 204: edge cloud e is set on edge cloud k True value of all tasks of (a)
Figure QLYQS_6
The value assigned to edge cloud e k Temporary truth value for all tasks of (1)>
Figure QLYQS_7
Step 205: calculating all user i E U k Weights of (2)
Figure QLYQS_8
Wherein the method comprises the steps of
Figure QLYQS_9
std j Is task t j Standard deviation of all perceived data of +.>
Figure QLYQS_10
Is task t j Is a temporary true value of (a);
step 206: calculate each task t j ∈T k True value of (2)
Figure QLYQS_11
Step 207: updating the current iteration number
Figure QLYQS_12
Step 208: if it is
Figure QLYQS_13
And->
Figure QLYQS_14
Step 204 is performed, otherwise step 209 is performed;
step 209: outputting an estimated edge cloud e k True value of all tasks of (a)
Figure QLYQS_15
And weight w of all users k ；
The edge cloud submits the estimated true value to the deep cloud, and the deep cloud calculates each edge cloud e k Weight g of E k Edge cloud e k The formula of the weight update of (2) is:
Figure QLYQS_16
wherein the method comprises the steps of
Figure QLYQS_17
Alpha and beta are two constants, alpha isThe reliability coefficient, beta is an importance coefficient, alpha is the influence of the number of users of edge clouds on the final true value discovery, the greater alpha is, the edge clouds with more users have larger weight, the larger beta is, the edge clouds with more budget have larger weight, the larger alpha epsilon [0,1 ] are in the true value discovery process],β∈[0,1],α+β＝1；/>
Figure QLYQS_18
Is edge cloud e k Task t j Is (are) perception data>
Figure QLYQS_19
Is task t j According to the true value of X * Estimating final true value X of all tasks through true value discovery method ** 。
2. A method of excitation of an edge-assisted mobile crowd-aware truth finding system according to claim 1, characterized by: given edge cloud e k Task set T of (1) k Edge cloud e k User set U of (2) k Edge cloud e k Budget G of (2) k Type μ and bid strategy for all tasks
Figure QLYQS_20
Each edge cloud e k Computing winner set +.>
Figure QLYQS_21
And each winner i.epsilon.S k Payment p of (2) i Let->
Figure QLYQS_22
Where r is the edge cloud number; let p k And p is S respectively k And a payment policy of S,
the utility of any user i is defined as the difference between the payment and its actual cost:
u i ＝p i -c i (1)
wherein u is i Is the utility of user i, c i Is a task set T i Cost of (2);
for any task t j ∈T i From edge cloud e k Winner set S in (1) k The obtained quality function is defined as:
Figure QLYQS_23
wherein T is i Is the set of tasks that user i is willing to perform, v i ∈w k ，v i Is the weight of user i; mu (mu) j Is task t j Log reflects the diminishing returns of the platform in terms of quality brought by the participating users;
the quality function is maximized so that the total payment does not exceed the budget, expressed as:
target maximaze V (S) k ) (3)
Constraint:
Figure QLYQS_24
said each edge cloud e k Computing winner set
Figure QLYQS_25
And each winner i.epsilon.S k Payment p of (2) i Implemented by budget feasible reverse auction, the steps are as follows:
step 301: initializing a winner set
Figure QLYQS_26
Initializing a payment policy p k ＝0；
Step 302: initializing a set of users whose bids do not exceed a budget
Figure QLYQS_27
Wherein b i Is the bid of user i;
step 303: let i * Is that
Figure QLYQS_28
The most valuable users in the collection: />
Figure QLYQS_29
Step 304: step 305 is performed with 2/5 probability and step 306 is performed with 3/5 probability;
step 305: will i * Joining to a winner set S k In (2) and set user i * Is paid as
Figure QLYQS_30
Step 317 is performed;
step 306: at the position of
Figure QLYQS_31
Find->
Figure QLYQS_32
User i with the greatest value, wherein>
Figure QLYQS_33
V i' (S k )＝V(S k ∪{i'})-V(S k )；
Step 307: if it meets
Figure QLYQS_34
Step 308 is performed, otherwise step 310 is performed;
step 308: adding user i to the winner set S k In (a) and (b);
step 309: at the position of
Figure QLYQS_35
Find->
Figure QLYQS_36
User i with the greatest value, wherein>
Figure QLYQS_37
For the collection->
Figure QLYQS_38
And set S k Is performed in step 307;
step 310: judge pair set S k Whether or not each user i has performed steps 311 to 316; if so, step 317 is performed; if not, go to step 311;
step 311: order the
Figure QLYQS_39
Let->
Figure QLYQS_40
Step 312: finding collections
Figure QLYQS_41
Middle->
Figure QLYQS_42
User i' with the greatest value, wherein>
Figure QLYQS_43
Step 313: if user i' satisfies
Figure QLYQS_44
Step 314 is performed, otherwise step 310 is performed;
step 314: finding collections
Figure QLYQS_45
Middle->
Figure QLYQS_46
User i' with the greatest value, wherein>
Figure QLYQS_47
Step 315: calculation of
Figure QLYQS_48
Step 316: and adding i 'to S' k In (a):
Figure QLYQS_49
step 313 is performed;
step 317: output winner set S k Payment policy p k 。
CN202010236632.3A 2020-03-30 2020-03-30 Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof Active CN111464620B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010236632.3A CN111464620B (en) 2020-03-30 2020-03-30 Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010236632.3A CN111464620B (en) 2020-03-30 2020-03-30 Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof

Publications (2)

Publication Number Publication Date
CN111464620A CN111464620A (en) 2020-07-28
CN111464620B true CN111464620B (en) 2023-05-23

Family

ID=71679263

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010236632.3A Active CN111464620B (en) 2020-03-30 2020-03-30 Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof

Country Status (1)

Country Link
CN (1) CN111464620B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112787809B (en) * 2020-12-31 2022-08-12 华南理工大学 Efficient crowd sensing data stream privacy protection truth value discovery method
CN112950251B (en) * 2021-05-11 2024-01-26 北京航空航天大学 Reputation-based vehicle crowd sensing node reverse combination auction excitation optimization method

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108876567A (en) * 2018-05-25 2018-11-23 华南理工大学 A kind of intelligent perception motivational techniques based on perception maximization of utility

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110189174A (en) * 2019-05-29 2019-08-30 南京工业大学 A kind of mobile intelligent perception motivational techniques based on quality of data perception

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108876567A (en) * 2018-05-25 2018-11-23 华南理工大学 A kind of intelligent perception motivational techniques based on perception maximization of utility

Also Published As

Publication number Publication date
CN111464620A (en) 2020-07-28

Similar Documents

Publication Publication Date Title
CN111754000B (en) Quality-aware edge intelligent federal learning method and system
US10715962B2 (en) Systems and methods for predicting lookalike mobile devices
Yang et al. On designing data quality-aware truth estimation and surplus sharing method for mobile crowdsensing
CN104850935B (en) It is a kind of with minimize payment be target mobile intelligent perception motivational techniques
US9002774B2 (en) Systems and methods for generating a forecasting model and forecasting future values
CN110189174A (en) A kind of mobile intelligent perception motivational techniques based on quality of data perception
CN111464620B (en) Edge-assisted mobile crowd sensing true value discovery system and excitation method thereof
Li et al. Multi-criteria group decision making based on trapezoidal intuitionistic fuzzy information
CN111932318B (en) Region division method and device, electronic equipment and computer readable storage medium
CN111461812A (en) Object recommendation method and device, electronic equipment and readable storage medium
CN111626563B (en) Dual-target robust mobile crowd sensing system and excitation method thereof
EP3695349A1 (en) Systems and methods for using geo-blocks and geo-fences to discover lookalike mobile devices
Anderson et al. Bayesian cluster detection via adjacency modelling
CN114817946A (en) Credible execution environment-based federated learning gradient boosting decision tree training method
Yu et al. Participant incentive mechanism toward quality-oriented sensing: understanding and application
CN116669111A (en) Mobile edge computing task unloading method based on blockchain
Yang et al. A decentralized trust inference approach with intelligence to improve data collection quality for mobile crowd sensing
US20220374484A1 (en) Graph learning and automated behavior coordination platform
Wang et al. TVD-RA: A truthful data value discovery based reverse auction incentive system for mobile crowd sensing
Li et al. Federated learning based on Stackelberg game in unmanned-aerial-vehicle-enabled mobile edge computing
Tang et al. A Semi-supervised Sensing Rate Learning based CMAB scheme to combat COVID-19 by trustful data collection in the crowd
CN117217779A (en) Training method and device of prediction model and information prediction method and device
Zhang et al. A reputation-based and privacy-preserving incentive scheme for mobile crowd sensing: a deep reinforcement learning approach
CN114862065B (en) Social work task planning method and device, electronic equipment and storage medium
Tham et al. A spatio-temporal incentive scheme with consumer demand awareness for participatory sensing

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
CB02 Change of applicant information
CB02 Change of applicant information

Address after: 210003 Gulou District, Jiangsu, Nanjing new model road, No. 66

Applicant after: NANJING University OF POSTS AND TELECOMMUNICATIONS

Address before: Yuen Road Qixia District of Nanjing City, Jiangsu Province, No. 9 210000

Applicant before: NANJING University OF POSTS AND TELECOMMUNICATIONS

GR01 Patent grant
GR01 Patent grant