Credible low-cost reward method giving consideration to both data quality and coverage rate improvement

Abstract

The invention discloses a credible low-cost reward method giving consideration to improvement of data quality and coverage rate, and aims to improve data quality, task coverage rate and cost control of crowd-sourcing network data collection. According to the method, a trust mechanism and an incentive strategy based on the task coverage rate and the data quality are introduced, and recovery of compensation reward is considered at the same time, so that the task coverage rate can be improved while workers are incentive to provide high-quality data, and the operation cost of a platform is remarkably reduced. Compared with the prior art, the method effectively solves the problems that the data quality is difficult to guarantee, the task coverage rate is insufficient and the cost is too high in the conventional method, so that the balance optimization of the data quality, the coverage rate and the cost control is realized while the requirements of different application scenes are met.

Description

Technical Field

[0001] The present invention belongs to the field of data collection and remuneration payment of group intelligence networks, and particularly relates to a reliable, low-cost remuneration method that takes into account both data quality and coverage improvement, and a remuneration payment method that takes into account the trust of workers to achieve data quality and coverage improvement at a low cost. Background Art

[0002] Mobile Crowdsensing (MCS), as an emerging data collection technology, uses mobile devices carrying sensors and a large number of user groups to complete perception tasks, and has shown great application potential in many fields such as intelligent transportation, environmental monitoring, and industrial automation. However, the widespread application of MCS technology faces three key challenges: data quality, task coverage, and cost control (i.e., QCC problem).

[0003] Data quality is one of the core issues of MCS. High-quality data needs to be consistent with ground truth data (GTD), but due to the inevitable existence of malicious workers, they may submit false data for their own interests, further exacerbating the uncertainty of data quality and hindering data collection by data requesters. For this reason, the present invention screens workers based on their trustworthiness, so as to achieve the real data quality expected by data requesters as much as possible.

[0004] Task coverage refers to the proportion of tasks released by the platform that have been completed. High coverage is crucial for data-driven services. However, workers often perform tasks based on opportunism and tend to choose low-cost tasks, resulting in remote or difficult tasks being ignored. Although workers can be motivated to perform non-focus tasks by increasing task rewards, this approach will significantly increase platform costs.

[0005] How to control costs is another key issue in MCS. The platform needs to pay workers to compensate for the cost of their sensing tasks, but the platform hopes to obtain high-quality data and high coverage at the lowest cost, while workers expect to get higher remuneration, and there is a conflict of interest between the two. Although existing incentive mechanisms such as auction mechanisms can achieve market equilibrium, they are still insufficient in terms of data quality and coverage. In addition, although incentive mechanisms based on behavioral economics can reduce costs, they ignore the possible dishonesty of workers and the uncertainty of data quality.

[0006] Existing technologies have obvious shortcomings in simultaneously meeting the requirements of high data quality, high task coverage and low cost. Therefore, there is an urgent need for an innovative mechanism that can comprehensively solve the above problems to promote the widespread application of MCS technology. Summary of the invention

[0007] The present invention proposes a reliable, low-cost remuneration method that takes into account both data quality and coverage improvement. By innovatively introducing a trust mechanism and optimizing a remuneration strategy, this method can improve the coverage of tasks while motivating workers to provide high-quality data, and significantly reduce the operating costs of the platform. Compared with the prior art, the present invention effectively solves the problems of difficult to ensure data quality, insufficient task coverage, and excessively high costs in previous methods, thereby achieving a balanced optimization of data quality, coverage, and cost control while meeting the needs of different application scenarios. Overall, the present invention first roughly classifies workers by the platform, including trusted workers and workers with uncertain trust, and assigns their trust values ​​of 1.0 and 0.5; then, each worker quotes based on their own mobility costs and promised data quality, so that the platform calculates the contribution efficiency of each worker to each task; secondly, the platform selects the workers with the highest contribution efficiency to execute the task, and notifies the selected set Each worker u s , inform that they have been selected by the platform; then, the worker provides the data to the platform, and the platform updates the trust and the compensation coefficient based on the difference in data quality before and after; finally, the platform pays direct compensation and additional compensation, and accumulates compensation until the threshold is reached, and then recovers the compensation to offset the compensation utility caused by excessive movement. To achieve the above purpose, the present invention provides a reliable, low-cost compensation method that takes into account both data quality and coverage improvement, which is characterized by comprising the following steps:

[0008] (1) The platform publishes the tasks to be executed. The platform first sets the set of workers as U and the initial set of trusted workers as U c , whose trust is 1, and the set of workers with undetermined trust is U u , whose initial trust is 0.5, and the set of untrustworthy workers is U m , and then the elected worker u s Execute the task, return the data to the platform after completion, and update the trust level in rounds; announce the worker's initial mobile remuneration coefficient q is the data collection of the qth round; the task execution of each round is carried out according to the following steps;

[0009] (2) Workers submit to the platform a set of tasks they are willing to perform Worker j Submit your own information to the platform, including the worker's location information (x i ,y i ), the maximum distance of initial intention to move

[0010] (3) The platform receives a set of workers willing to perform tasks: The set of workers selected by the platform is The platform selects workers in the following way:

[0011] According to the worker's contribution utility rate D s,n , the platform selects k workers with the highest contribution efficiency for each task, where D s,n Calculate according to the following formula: Contribution utility rate is the ratio of the data quality, trust, and willingness to move value committed by the worker to perform the task to his bid; s,n Indicates the data quality committed by the worker, g s,n The larger it is, the higher the quality of the data submitted by the worker is, and thus the greater the contribution rate is, which promotes the improvement of data quality; Indicates workers' trust, The higher it is, the higher the probability that the worker submits real data, and thus the greater the contribution rate. By comparing the difference between the quality of the data submitted by the worker and the quality of the data he promised, the trust degree is changed. The specific calculation is as shown in the calculation of trust degree in (5.1); thus promoting the improvement of data quality; I s Indicates the distance a worker is willing to move to perform a task and the distance l between the worker and the task s,n The ratio of I s When the value is greater than 1, it indicates that the distance that the worker is willing to move is greater than the distance between the worker and the task. Therefore, the larger the value, the greater the probability that the task is executed by the worker, thereby promoting the improvement of task coverage. s,n is the worker's bid. The smaller the worker's bid, the greater the worker's contribution rate.

[0012] (4) The platform notifies the selected collection Each worker u s , telling them that they have been selected by the platform;

[0013] (5) Selected worker u s Execute the task and return the data to the platform after completion. The platform performs the following actions:

[0014] (5.1) The platform determines the data quality g promised by the workers s,n The quality of data submitted by workers G s,n , the trust is calculated according to the following formula: in represents the trust level of the sth worker in the qth round, It is the maximum value of the worker's trust, usually a positive number, such as 1.0.

[0015] (5.2) The platform updates the worker's mobility compensation coefficient according to the following formula based on the worker's completion status. η is the adjustment coefficient and η∈(0,1), The upper limit of the platform movement reward coefficient And f i To distinguish the nature of the task, the calculation is performed according to the following formula: Where σ>1, At the same time, set a threshold value ζ h , for example, h =0.7×ζ i , used to distinguish whether it is considered a hot task; when ζ i ≥ζ h When task t i It is a non-hotspot task, and its hotspot task flag is f i =1, otherwise, f i =1;

[0016] (5.3) Since workers perform non-hotspot tasks, their moving distance may exceed the maximum distance they are willing to move. In order to motivate workers to perform such tasks, when workers move beyond their maximum distance to perform tasks, they will generate compensation utility. The platform is the cumulative excess moving distance greater than the threshold A. h The more the accumulated excess, the greater the increase in the worker's compensation benefit. The platform calculates the worker's compensation benefit A(l o ): and the accumulated excess movement distance l o Related; among them is the worker compensation loss coefficient, which is related to the accumulated compensation remuneration of workers entering the mechanism. The greater the compensation remuneration, the The smaller; at the same time the following formula is evaluated: and A s The independent variables are workers' compensation benefit A and its threshold A h The ratio p s , when p s When it is 1, it means that the worker's threshold A has been reached. h , at this time, workers should recover part of the compensation to offset the additional costs of going to remote areas to perform sensing tasks. τ represents the fitting coefficient of the compensation benefit, for example, 0.2-0.3;

[0017] (6) After the worker completes the sensing task assigned by the platform, the platform pays the worker a reward; the reward is divided into two parts. One part is paid immediately after the participant completes the task, which is called direct reward D s,n Calculate according to the following formula: in is the weight of fixed compensation, is the weight of the worker's additional reward, and For example, take 0.4-0.6; φ is the data quality factor, for example, take about 1.2, r s,n A fixed payment related to the distance the worker has to perform the task; s,n It is the additional compensation for workers, which is related to the data quality and is calculated according to the following formula Among them, G s,n is the actual data quality of the worker; g s,n is the data quality that the worker commits to before performing the task; Δ g It represents the difference between the actual data quality and the promised data quality; ω>1 is the penalty coefficient, and 0<ν,η<1, which is expressed as the reward adjustment factor and the penalty adjustment factor; additional rewards or penalties are given based on the difference in data quality before and after;

[0018] (7) The other part is compensation E m,n , and continues to accumulate until the workers' compensation benefits exceed the maintenance threshold A h , will gradually make the workers recover, to offset the extra costs of performing non-hotspot tasks; the compensation E′ recovered by the workers m,n Calculated as follows: in Indicates that the workers' compensation benefit reaches the maintenance threshold A h The round set, E′ m,n The accumulated compensation for workers to perform sensing tasks, μ s is its weight, that is, each payment will be directly drawn from the total compensation according to a certain proportion; when workers perform non-hotspot tasks, they accumulate compensation benefits and thus get the opportunity to recover compensation. In order to motivate workers to go to remote areas to perform tasks, a weight μ is given to each recovered compensation. s , is the attenuation coefficient, the value range is (0,1), and it is set by the platform; The total number of rounds from when a worker starts to perform a task to when he finally exits the mechanism is r, and s is the round in which the worker receives a bonus. After receiving the bonus, the worker needs to perform the perception task as many times as possible in succession so that he can recover more compensation when he exits. If the worker exits directly, the compensation E m,n Cannot be recycled.

[0019] (8) If the task rounds are not completed, repeat steps (2) to (7) until the task is completed.

[0020] Beneficial Effects

[0021] The present invention proposes a method for comprehensively solving the three key problems of data quality, task coverage and cost control (QCC problem) in mobile crowdsourcing sensing (MCS) technology. By introducing a trust scoring mechanism and a dynamic reward adjustment strategy, the present invention can screen out credible workers and motivate them to submit high-quality data, while improving task coverage and controlling costs. Experimental results show that the present invention performs well in data quality, and the data accuracy increases rapidly with the increase of task rounds, and tends to stabilize more quickly when the ratio of the initial number of workers increases. In terms of task coverage, the task coverage of the method of the present invention is as high as 83% or more, and even reaches 96% in some cases, which is much higher than the 25%-70% of the traditional method. In addition, the present invention balances the relationship between data quality, task coverage and cost by dynamically screening workers with high commitment quality and low quotations, so that the platform utility steadily increases with the increase of task rounds and eventually tends to stabilize. When the number of workers is large, the return on investment of the present invention is similar to that of the traditional method, but it performs better in task coverage and data quality. The present invention also has the characteristics of strong dynamic adaptability, and can provide real-time feedback and adjustments based on the task execution and worker performance, and flexibly respond to different task requirements and worker behavior changes. In summary, the present invention achieves optimization in multiple key dimensions, providing an efficient, reliable and innovative solution for the widespread application of MCS technology, which has important practical application value and promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1 A schematic diagram of the process of the method is shown;

[0024] Figure 2 The worker trust score is updated as the number of task execution rounds increases;

[0025] Figure 3 The task coverage of this method;

[0026] Figure 4 The data quality of this method;

[0027] Figure 5 The return on investment of the method platform; DETAILED DESCRIPTION

[0028] In order to facilitate the understanding of the present invention, the present invention will be described more comprehensively and meticulously below in conjunction with the accompanying drawings and preferred embodiments of the specification, but the protection scope of the present invention is not limited to the following specific embodiments.

[0029] Unless otherwise defined, all professional terms used below have the same meanings as those generally understood by those skilled in the art. The professional terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the scope of protection of the present invention.

[0030] Unless otherwise specified, various raw materials, reagents, instruments and equipment used in the present invention can be purchased from the market or prepared by existing methods.

[0031] Example:

[0032] In the crowd intelligence perception network, it is necessary to obtain the real data of a certain place, and a higher data quality is required, such as temperature and humidity, air quality, traffic conditions and other data for perception application scenarios, and the perceived data is uploaded to the platform for processing. When the platform releases tasks, participants apply to participate in the task, and the trust of different participants and the true value data of different tasks are calculated by the method of the present invention through the different data provided by the workers, and the evaluation indicators of the execution of different tasks are obtained. Then, through the execution of the task rounds, the trust of the workers is updated with the iteration of the rounds, so as to obtain a method for collecting higher quality perception data.

[0033] The experimental results of the inventive method are given below.

[0034] Figure 1 A schematic diagram of the process of the present invention is provided, which mainly includes:

[0035] 1. Initialize the workers and publish the tasks on the platform. The platform publishes the worker selection strategy and remuneration payment plan, such as Figure 1 Steps (1)-(4) in ;

[0036] 2. The worker generates a set of tasks that he is willing to perform based on his own situation. For any task, it includes location information and mark information. The location information is the location coordinates of the worker, such as step (5) in 1.

[0037] 3. Workers submit the set of tasks they are willing to perform to the platform. The set of workers willing to perform tasks is, submitting their own information to the platform. The worker's information includes the worker's location information and the maximum distance they are willing to move initially, such as Figure 1 Step (6) in

[0038] 4. The platform selects workers, such as Figure 1 Step (7) in

[0039] 5. The platform adjusts the worker's mobile remuneration coefficient and calculates the worker's compensation benefits based on the worker's completion status, such as Figure 1 Steps (8) and (9) in

[0040] 6. The platform calculates and updates the trust level based on the worker’s completion status, such as Figure 1 Step (10) in

[0041] 7. The platform calculates the true value of the task data based on the worker’s completion status, such as Figure 1 Step (11) in

[0042] 8. When the worker's compensation benefits exceed the maintenance threshold, the worker can recover part of the compensation to offset the additional costs of performing non-hotspot tasks, such as Figure 1 Step (12) in

[0043] Figure 2 The update of the worker trust score of the method of the present invention with the number of task execution rounds is given. From the experimental results, it can be seen that the method of the present invention allows the worker's trust score to change with the number of rounds, and the screening is carried out through the trust score increase and decrease mechanism to achieve the requirement of cultivating trustworthy workers.

[0044] Figure 3 The experimental results of the task coverage rate of workers performed by the method of the present invention are given when a few trusted workers and most workers with undetermined trust are known at the beginning. As can be seen from the experiment, as the method runs, the task coverage rate of the previous method 1 (which only considers the trust of workers but does not consider the incentive method of the compensation utility brought by workers' excessive movement) is lower than that of the other two methods under various combinations of the number of workers and the number of tasks. The task coverage rate is only 25%-70%, while the task coverage rate of the other methods is as high as 83% or more. This is due to the lack of external incentives, which leads to the lack of motivation for workers to go to more distant areas. Only workers who meet the trust requirements in the initial rounds are continuously selected, resulting in stagnant task coverage. The reward method proposed in the present invention not only takes trust into account, but also introduces a direct reward and compensation reward recovery mechanism. The trust is updated through the evaluation of data quality in each round, and the trusted workers are reselected. This strategy aims to motivate more workers to participate in non-hotspot tasks on the one hand, thereby improving task coverage; on the other hand, it aims to cultivate more trusted workers. As the number of workers increases, the method proposed in this invention shows a higher peak in task coverage, with a maximum coverage of about 96%, and a large gap after about 25 rounds with the previous method 2 (this method only considers the collection of perception data through incentives, but does not consider the bias of workers' trust, and the incentive method that regards all data as true values). In addition, for the four experimental results, in an average of 53.1% of the rounds, the method proposed in this invention performs better than the previous method in task coverage. This is because we not only select individuals with higher trust, but also select individuals with lower trust.

[0045] Figure 4 The experimental results of the data quality of the perception data submitted by the workers with the operation of the network are given when a small number of trustworthy workers and most workers with uncertain trust are initially known. It can be seen from the experiment that the accuracy of the data obtained by the method of the present invention gradually increases with the operation of the method. Compared with the previous methods, it can be seen that the data quality of the invention is more stable, and as the ratio of the initial number of workers increases, the faster the data quality obtained tends to stabilize, the higher the data quality.

[0046] Figure 5 The experimental results of the platform utility with the operation of the network using the method of the present invention are given. It can be seen from the experiment that with the operation of the method, the changing trend of the platform utility with the task rounds under 5 kinds of compensation rewards are considered respectively. The platform utility is defined as the sum of the task values ​​minus the sum of the rewards paid to the workers. Under the method of the invention, the platform utility gradually increases with the increase of the task rounds, grows faster in the first 15 rounds, then grows slowly, and gradually stabilizes after about 30 rounds of tasks. Among them, when the compensation reward is 10, it finally stabilizes at a level of about 2700, while the platform utility under other compensation reward conditions is stable at about 2550. This mechanism can effectively screen out workers with higher commitment quality and lower bids to perform tasks. With the increase of task rounds, the unselected workers will dynamically adjust their own information, resulting in an increase in the volatility of the commitment quality. This dynamic adjustment mechanism enables the platform utility to gradually improve with the increase of rounds.

Claims

1. A reliable and low-cost compensation method that takes into account both data quality and coverage improvement, characterized by The following steps are involved: (1) The set of workers is U. Initially, a small proportion of workers are trusted workers. The initial set of trusted workers is U c , whose trust is 1, and the set of workers with undetermined trust is U u , whose initial trust is 0.5, and the set of untrustworthy workers is U m , which is empty initially. (2) The platform publishes the tasks to be performed and announces the workers’ initial mobility compensation coefficient q is the data collection of the qth round; the task execution of each round is carried out according to the following steps; (3) Workers submit to the platform a set of tasks they are willing to perform Worker i Submit your own information to the platform, including the worker's location information (x i ,y i ), the maximum distance of initial intention to move (4) The platform receives a set of workers willing to perform tasks: The set of workers selected by the platform is The platform selects workers in the following way: According to the worker's contribution utility rate D s,n , the platform selects k workers with the highest contribution efficiency for each task, where D s,n Calculate according to the following formula: Contribution utility rate D s,n is the ratio of the worker's promised data quality, trust, and willingness to move value to perform the task to his bid; g s,n Indicates the data quality committed by the worker, g s,n The larger it is, the higher the quality of the data submitted by the worker is, and thus the greater the contribution rate is, which promotes the improvement of data quality; Indicates workers' trust, The higher it is, the higher the probability that the worker submits real data, and thus the greater the contribution rate. By comparing the difference between the quality of the data submitted by the worker and the quality of the data he promised, the trust level is changed. The specific calculation is as shown in the calculation of trust level in (6.1); thus promoting the improvement of data quality; I s Indicates the distance a worker is willing to move to perform a task and the distance between the worker and the task l s,n The ratio of I s When the value is greater than 1, it indicates that the distance that the worker is willing to move is greater than the distance between the worker and the task. Therefore, the larger the value, the greater the probability that the task is executed by the worker, thereby promoting the improvement of task coverage. s,n is the worker's bid. The smaller the worker's bid, the greater the worker's contribution rate. (5) The platform notifies the selected collection Each worker u s , telling them that they have been selected by the platform; (6) Selected worker u s Execute the task and return the data to the platform after completion. The platform performs the following actions: (6.1) The platform determines the data quality g promised by the workers s,n The quality of data submitted by workers G s,n , the trust is calculated according to the following formula: in represents the trust of the sth worker in the qth round, It is the maximum value of the worker's trust, usually a positive number, such as 1.

0. (6.2) The platform updates the worker's mobility compensation coefficient according to the following formula based on the worker's completion status. η is the adjustment coefficient and η∈(0,1), The upper limit of the platform movement reward coefficient And f i To distinguish the nature of the task, the calculation is performed according to the following formula: Where σ>1, At the same time, set a threshold value ζ h , for example, h =0.7×ζ i , used to distinguish whether it is considered a hot task; when ζ i ≥ζ h When task t i It is a non-hotspot task, and its hotspot task flag is f i =1, otherwise, f i =1; (6.3) Since workers perform non-hotspot tasks, their moving distance may exceed the maximum distance they are willing to move. In order to motivate workers to perform such tasks, when workers move beyond their maximum distance to perform tasks, they will generate compensation utility. The platform is the cumulative excess moving distance greater than the threshold A. h The more the accumulated excess, the greater the increase in the worker's compensation benefit. The platform calculates the worker's compensation benefit A(l o ): and the accumulated excess movement distance l o Related; among them is the worker compensation loss coefficient, which is related to the accumulated compensation remuneration of workers entering the mechanism. The greater the compensation remuneration, the The smaller; at the same time the following formula is evaluated: and A s The independent variables are workers' compensation benefit A and its threshold A h The ratio p s , when p s When it is 1, it means that the worker's threshold A has been reached. h , at this time, workers should recover part of the compensation to offset the additional costs of going to remote areas to perform sensing tasks. τ represents the fitting coefficient of the compensation benefit, for example, 0.2-0.3; (7) After the worker completes the sensing task assigned by the platform, the platform pays the worker a reward; the reward is divided into two parts. One part is paid immediately after the participant completes the task, which is called direct reward D s,n Calculate according to the following formula: in is the weight of fixed compensation, is the weight of the worker's additional reward, and For example, take 0.4-0.6; φ is the data quality factor, for example, take about 1.2, r s,n A fixed payment related to the distance between the worker and the task to be performed; s,n The additional compensation for workers is related to data quality and is calculated according to the following formula Among them, G s,n is the actual data quality of the worker; g s,n is the data quality that the worker commits to before performing the task; Δ g It represents the difference between the actual data quality and the promised data quality; ω>1 is the penalty coefficient, and 0<ν,η<1, which is expressed as the reward adjustment factor and the penalty adjustment factor; additional rewards or penalties are given based on the difference in data quality before and after; (8) The other part is compensation E m,n , and continues to accumulate until the workers' compensation benefits exceed the maintenance threshold A h , will gradually make the workers recover, to offset the extra costs of performing non-hotspot tasks; the compensation E′ recovered by the workers m,n Calculated as follows: in Indicates that the workers' compensation benefit reaches the maintenance threshold A h The round set, E′ m,n The accumulated compensation for workers to perform sensing tasks, μ s is its weight, that is, each payment will be directly drawn from the total compensation according to a certain proportion; when workers perform non-hotspot tasks, they accumulate compensation benefits and thus get the opportunity to recover compensation. In order to motivate workers to go to remote areas to perform tasks, a weight μ is given to each recovered compensation. s , is the attenuation coefficient, the value range is (0,1), and it is set by the platform; The total number of rounds from when a worker starts to perform a task to when he finally exits the mechanism is r, and s is the round in which the worker receives a bonus. After receiving the bonus, the worker needs to perform the perception task as many times as possible in succession so that he can recover more compensation when he exits. If the worker exits directly, the compensation E m,n Cannot be recycled. (9) If the task rounds are not completed, repeat steps (3) to (8) until the task is completed.

Images