Sparse crowd sensing task allocation method based on two-stage heuristic search

Abstract

This invention discloses a sparse crowd-sensing task allocation method based on a two-stage heuristic search. The method includes: obtaining the sensing task and its total cost; dividing the sensing task into m sensing cycles in time and n initial sensing sub-regions in the sensing area; and grouping similar initial sensing sub-regions within the n initial sensing sub-regions into a set based on historical data and spatiotemporal correlations, thereby obtaining a set of L sensing sub-regions. This invention addresses the current technical problem of poor data inference accuracy, or poor quality of sensing data, in sparse crowd-sensing task allocation.

Description

Technical Field

[0001] The present invention relates to the technical field of task allocation, and in particular to a sparse crowd-sensing task allocation method based on two-stage heuristic search. Background Art

[0002] As smart mobile devices are equipped with an increasing number of sensors, such as accelerometers, gyroscopes, GPS, cameras, and microphones, mobile crowd sensing (MCS) has become a promising data collection method for smart city projects. This technology leverages the collaborative nature of groups and the sensing, computing, and communication capabilities of mobile devices to effectively address the challenges of large-scale data collection. To ensure high data quality, MCS applications typically require mobilizing a large number of participants to cover a wide geographic area, but this approach often comes with high perception costs. Consequently, much research has focused on finding ways to reduce perception costs without sacrificing data quality. Several researchers have proposed a new framework, Sparse MCS. This framework exploits temporal and spatial correlations in data to implement a data collection strategy focused on a few key sub-areas. This strategy enables inference of data across the entire uncovered area based on limited data input. Through this strategy, Sparse MCS significantly reduces the number of tasks and perception costs while still maintaining high data quality. This not only improves overall data collection efficiency but also provides strong technical support for the sustainable development of smart cities.

[0003] Existing sparse crowd sensing task allocation methods often oversimplify the problem of subregion selection. These approaches typically assume that there are numerous participants within the target sensing area, that all sensing subregions are reachable, and that participants submit accurate sensing data. However, the reality is far more complex. The quality of the sensing data submitted by participants can vary due to differences in individual device performance, improper operation, or malicious behavior. In sparse crowd sensing, in particular, a limited number of subregions are selected, and data inference for the remaining unperceived areas relies on data from these selected subregions. If low-quality data submitted by participants with low credibility is used for inference, it will directly impact the sensing performance and reliability of the entire system. Furthermore, early sensing subregion selection strategies, such as query by committee (QBC), primarily select subregions based on uncertainty. While these approaches can identify areas with insufficient information, they can sometimes overlook other areas that may contain important information due to their excessive focus on uncertainty. This bias can lead to an incomplete assessment of the value of data points, overlooking potentially high-value data points. Summary of the Invention

[0004] The embodiment of the present invention provides a sparse crowd sensing task allocation method based on a two-stage heuristic search, which at least solves the technical problem of poor accuracy of data inference in the current sparse crowd sensing task allocation, that is, poor quality of perception data.

[0005] According to one aspect of an embodiment of the present invention, a sparse crowd sensing task allocation method based on a two-stage heuristic search is provided. The method may include: obtaining a sensing task and a total cost of the sensing task; dividing the sensing task into m sensing cycles in time, and dividing the sensing task into n initial sensing sub-regions in the sensing area; based on the historical data and spatiotemporal correlation of the n initial sensing sub-regions, dividing similar initial sensing sub-regions in the n initial sensing sub-regions into a set to obtain L sensing sub-region sets after division; when any one or more participants are in any one of the m sensing cycles, matching any one or more participants with any initial sensing sub-region in a subset of a partial sensing sub-region set in the L sensing sub-region set based on a two-stage heuristic search algorithm to obtain multiple initial sensing matching pairs, wherein each initial sensing matching pair includes any one or more participants to any initial sensing sub-region in a subset of a partial sensing sub-region set. The geographical distance of the area, the target perception value and the movement cost corresponding to the geographical distance, the target perception value is determined based on the credibility and initial perception value of any participant or multiple participants; based on the spatiotemporal correlation of the set of L perception sub-areas, the initial perception matrix in multiple initial perception matching pairs is inferred using the data inference algorithm F to obtain an inferred data matrix, wherein the initial perception matrix is ​​composed of multiple target perception values; based on the real data matrix and the inferred data matrix, the error between the real data matrix and the inferred data matrix is ​​determined; based on the movement cost and geographical distance corresponding to the geographical distance of each initial perception matching pair in the multiple initial perception matching pairs, the target values ​​of the multiple initial perception matching pairs are determined; when the target values ​​of the multiple initial perception matching pairs are less than the total cost of the perception task, the set of perception matching pairs corresponding to the minimum error between the real data matrix and the inferred data matrix is ​​determined.

[0006] Optionally, based on historical data and spatiotemporal correlation of the n initial perception sub-regions, similar initial perception sub-regions among the n initial perception sub-regions are divided into a set, and the expression of the divided set of L perception sub-regions is:

[0007] U sim ={U1,U2,...}={{u1,u2,u3},{u4,u5},...,{u m-2 ,u m-1 ,u m}}

[0008] Among them, {u1,u2,u3}, {u4,u5} and {u m-2 ,u m-1 ,u m} respectively represent that similar initial perception sub-regions in n initial perception sub-regions are divided into a set, U1 is a set of {u1,u2,u3}, U sim is the set of L perception sub-areas after division.

[0009] Optionally, based on a two-stage heuristic search algorithm, any one or multiple participants are matched with any initial perception sub-region in a subset of a part of the perception sub-region set in the L perception sub-region set to obtain multiple initial perception matching pairs, including: a first stage: selecting any participant with the closest geographical distance to any initial perception sub-region for matching to obtain multiple first perception matching pairs, and excluding multiple participants and multiple initial perception sub-regions in the multiple first perception matches in the second stage; a second stage: matching multiple participants remaining unmatched in the first stage with multiple initial perception sub-regions to obtain multiple second initial perception matching pairs; an iterative stage: adjusting the multiple first initial perception matching pairs and the multiple second initial perception matching pairs of the current cycle through inference errors to obtain multiple adjusted first perception matching pairs and multiple second perception matching pairs; and obtaining multiple initial perception matching pairs based on the adjusted multiple first perception matching pairs and multiple second perception matching pairs.

[0010] Optionally, based on the adjusted multiple first perceptual matching pairs and the multiple second perceptual matching pairs, multiple initial perceptual matching pairs are obtained, including: merging the adjusted multiple first perceptual matching pairs and the multiple second perceptual matching pairs to obtain multiple initial perceptual matching pairs.

[0011] Optionally, based on the spatiotemporal correlation of the L sets of perception sub-regions, the data inference algorithm F is used to infer the initial perception matrices in the multiple initial perception matching pairs, and the expression of the inferred data matrix is ​​obtained as follows:

[0012] D infer =F(D sense )

[0013] Among them, D infer is the inference data matrix, F is the data inference algorithm, D sense is the initial perception matrix.

[0014] Optionally, based on the movement cost and geographical distance corresponding to the geographical distance of each initial perceptual matching pair in the multiple initial perceptual matching pairs, the target values ​​of the multiple initial perceptual matching pairs are determined, including: multiplying the movement cost corresponding to the geographical distance of each initial perceptual matching pair in the multiple initial perceptual matching pairs by the geographical distance to obtain the target value of each initial perceptual matching; adding the target values ​​of each initial perceptual matching pair to obtain the target values ​​of the multiple initial perceptual matching pairs.

[0015] Beneficial effects of the present invention:

[0016] The present invention proposes a sparse crowd perception task allocation method based on a two-stage heuristic search. The method uses the historical task execution data of participants stored on the platform to comprehensively evaluate the trust of participants from four aspects: task completion rate, data accuracy, timeliness, and activity. At the same time, the temporal and spatial features of the perception sub-areas are extracted for cluster analysis, and the perception sub-areas with similar temporal and spatial features are divided into the same subset. Afterwards, a sparse crowd perception task allocation problem is constructed based on a bipartite graph, that is, the matching problem of participants and perception sub-areas. A two-stage heuristic search algorithm is used to solve the problem and generate a series of <participant, perception sub-area> pairs, which solves the current technical problem of poor accuracy of data inference in sparse crowd perception task allocation, that is, poor quality of perception data, and achieves the task allocation under the constraint of total cost in the sparse crowd perception task allocation process, thereby improving the accuracy of data inference, that is, improving the quality of perception data. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 is a flowchart of a sparse crowd sensing task allocation method based on two-stage heuristic search according to an embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of a cluster analysis result in a similarity analysis according to an embodiment of the present invention;

[0020] Figure 3 2 is a schematic diagram of participant credibility evaluation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and to describe a specific order or precedence. It should be understood that the terms used in this way are interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products or devices.

[0023] Example 1

[0024] According to an embodiment of the present invention, a sparse crowd sensing task allocation method based on a two-stage heuristic search is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system comprising at least one set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0025] Figure 1 is a flow chart of a sparse crowd sensing task allocation method based on two-stage heuristic search according to an embodiment of the present invention, such as Figure 1 As shown, the method may include the following steps:

[0026] Step S101: Acquire the perception task and the total cost of the perception task.

[0027] In the technical solution provided in the above step S101 of the present invention, a perception task is obtained, wherein the perception task can be an image classification task or an image segmentation task, etc., and the total cost of the perception task is limited to C.

[0028] Step S102: Divide the sensing task into m sensing cycles in time, and divide the sensing task into n initial sensing sub-areas in the sensing area.

[0029] In the technical solution provided in step S102 of the present invention, the sensing task is divided into m sensing cycles in time: T = {t1, t2, ... t m}, the perception area is divided into n perception sub-areas: U = {u1,u2,...,u n}, the sensing activity can actually be regarded as a combination of m×n sensing tasks, each of which is performed at a specific time and place, {u1,u2,...,u n} represent the first perception sub-region, the second perception sub-region and the nth perception sub-region respectively.

[0030] Step S103 : Based on the historical data and spatiotemporal correlation of the n initial perceptual sub-regions, similar initial perceptual sub-regions among the n initial perceptual sub-regions are divided into a set to obtain L sets of divided perceptual sub-regions.

[0031] In the technical solution provided in step S103 of the present invention, the method of evenly dividing the perceptual sub-regions may result in some perceptual sub-regions having similar perceptual data, resulting in perceptual redundancy during selection. Therefore, the historical data and spatiotemporal correlation of the n initial perceptual sub-regions are combined to group similar initial perceptual sub-regions among the n initial perceptual sub-regions into a set, thereby obtaining a set of L perceptual sub-regions after division. According to an embodiment of the present invention, a schematic diagram of cluster analysis results in similarity analysis is provided. Figure 2 FIG. 1 is a schematic diagram of a cluster analysis result in a similarity analysis according to an embodiment of the present invention. Figure 2 As shown in FIG, similar initial perception sub-regions are determined by the perception sub-region similarity analysis. The perception sub-region similarity analysis is based on historical data, and the spatiotemporal feature extraction and cluster analysis of the perception data in each sub-region are performed. The perception sub-regions with similar spatiotemporal features of the data belong to the same cluster and are finally divided into the same set, obtaining the cluster analysis result.

[0032] Step S104: When any one or more participants are in any one of the m perception cycles, the any one or more participants are matched with any initial perception sub-area in the subset of the partial perception sub-area set in the L perception sub-area set based on a two-stage heuristic search algorithm to obtain multiple initial perception matching pairs, wherein each initial perception matching pair includes the geographical distance from any one or more participants to any initial perception sub-area in the subset of the partial perception sub-area set, the target perception value, and the movement cost corresponding to the geographical distance, and the target perception value is determined based on the credibility and initial perception value of the one or more participants.

[0033] In the technical solution provided in step S104 of the present invention, when any one or more participants are in any one of the m perception cycles, a two-stage heuristic search algorithm is used to match any one or more participants with any initial perception sub-region in the subset of the partial perception sub-region set in the L perception sub-region set, thereby obtaining a plurality of initial perception matching pairs, wherein each initial perception matching pair includes a geographical distance L (p) from any one or more participants to any initial perception sub-region in the subset of the partial perception sub-region set. k ,u j ), target perception value d ij The cost of movement C corresponding to the geographical distance pk , target perception value d ij It is determined based on the credibility and initial perception value of any one or more participants. L(p k ,u j ) k represents the Kth participant, u j represents any initial perception sub-area in the subset of the partial perception sub-area set, L represents the geographical distance, C pk It represents the mobile version corresponding to the K-th participant, where the initial perception value is the perception value of any participant or multiple participants to any initial perception sub-area of ​​the subset of the partial perception sub-area set. The calculation process of the target perception value of a participant is as follows:

[0034] The reputation R of a participant can be expressed as the weighted average of C, A, T, and L as shown in the following formula:

[0035] R=w1·C+w2·A+w3·T+w4·L

[0036] Where: C is the task completion rate, A is the data accuracy, T is the timeliness, and L is the activity. w1, w2, w3, and w4 are the weights of these indicators, and they satisfy w1+w2+w3+w4=1. According to an embodiment of the present invention, a participant credibility evaluation is provided. Figure 3 is a schematic diagram of participant reputation evaluation according to an embodiment of the present invention, such as Figure 3 As shown, in Figure 3 The R value distribution is not much different from the fitted normal distribution curve. Therefore, for participants from channels without historical task execution data, the median can be used as the initial credibility setting. Based on the R value results, the proportions of high, medium, and low credibility participants are set to 0.25, 0.5, and 0.25.

[0037] To ensure that the evaluation results more accurately reflect current performance, the latest data for each indicator should have a greater impact on the final evaluation than earlier data. By using an exponentially decaying weighting method, the most recent data points are given a higher weight, and this weight gradually decreases as the data point becomes older. The participant reputation assessment is determined based on the participant's historical observation data.

[0038] w(t i )=e -λ·(T-t)

[0039] 1: Task completion rate C

[0040]

[0041] 2: Data Accuracy A

[0042] For a completed task, data accuracy A is the average difference between the submitted data and the true data.

[0043]

[0044] 3: Timeliness T

[0045] For completed tasks, timeliness T is the average difference between the task submission time and the deadline time.

[0046]

[0047] 4: Activity

[0048] Activity L is the frequency with which participants complete tasks within a certain period of time.

[0049]

[0050] in:

[0051] N is the total number of tasks assigned to the participants, and M is the total number of tasks completed by the participants;

[0052] w(t i ) is based on the completion time t of the i-th task i time-weighted factors;

[0053] s i is the completion status of the i-th task, 1 for completed and 0 for uncompleted;

[0054] d i is the submission data of the i-th task, is the real data of the i-th task;

[0055] t i is the completion time of the i-th task, sti is the start time of the i-th task, et i is the end time of the i-th task;

[0056] O is the length of the observation period.

[0057] Step S105 , based on the spatiotemporal correlation of the L sets of perception sub-regions, the data inference algorithm F is used to infer the initial perception matrices in the multiple initial perception matching pairs to obtain an inferred data matrix, wherein the initial perception matrix is ​​composed of multiple target perception values.

[0058] In the technical solution provided in the above step S105 of the present invention, based on the spatiotemporal correlation of the set of L perception sub-areas, the data inference algorithm F is used to infer the initial perception matrix in multiple initial perception matching pairs to obtain an inferred data matrix, wherein the initial perception matrix is ​​composed of multiple target perception values. For example, if the initial perception matrix is ​​30 data, then according to the data inference algorithm F, the inferred data matrix of the n perception sub-areas of the perception task can be inferred, wherein the inferred data matrix is ​​close to the real perception data.

[0059] Step S106 : determining the error between the real data matrix and the inferred data matrix based on the real data matrix and the inferred data matrix.

[0060] In the technical solution provided in step S106 of the present invention, based on the real data matrix and the inferred data matrix, the calculation expression for determining the error between the real data matrix and the inferred data matrix is ​​as follows:

[0061]

[0062] Among them, for tasks that collect continuous initial perception values ​​(such as temperature and humidity), the mean absolute error (MAE) is used to define E. For classification perception tasks (such as air quality index (AQI)), the classification error (CE) is used to define E. When the perception task is a continuous perception task, D infer (i, j) represents the inferred data of the jth initial perception sub-region observed in the i-th perception cycle, D real (i, j) represents the real data of the jth initial perception sub-area observed in the i-th perception cycle, where the value of i ranges from 1 to m and the value of j ranges from 1 to n. When the perception task is a discrete perception task, ψD infer (i, j) represents the inferred data of the jth initial perception sub-region observed in the i-th perception cycle, ψD real(i, j) represents the real data of the jth initial perception sub-area observed in the i-th perception cycle, where the value of i ranges from 1 to m, and the value of j ranges from 1 to n. 1 means that when ψD infer (i,j) and ψD real When (i, j) are not equal, it is recorded as 1.

[0063] Step S107 : determining target values ​​of the multiple initial perceptual matching pairs based on the movement cost and the geographical distance corresponding to the geographical distance of each of the multiple initial perceptual matching pairs.

[0064] In the technical solution provided in the above step S107 of the present invention, the movement cost and the geographical distance corresponding to the geographical distance of each of the multiple initial perception matching pairs are calculated to obtain the target values ​​of the multiple initial perception matching pairs.

[0065] Step S108 , when the target values ​​of the multiple initial perceptual matching pairs are less than the total cost of the perceptual task, determining a set of perceptual matching pairs corresponding to the minimum error between the real data matrix and the inferred data matrix.

[0066] In the technical solution provided in step S108 of the present invention, when the target values ​​of the multiple initial perceptual matching pairs are less than the total cost of the perceptual task, a set of perceptual matching pairs corresponding to the minimum error between the real data matrix and the inferred data matrix is ​​determined, as shown in the following expression:

[0067]

[0068] Among them, Φ is a set containing the mapping relationship of <participant, perception sub-area> pairs, representing the participant p k Assigned to the perceptual subregion u i Perform the perception task. P×U is the Cartesian product of all possible <actor, perception subregion> pairs, The set of perceptual matching pairs corresponding to the minimum error between the real data matrix and the inferred data matrix, It indicates that the target value of multiple initial perception matching pairs is less than the total cost of the perception task, and C represents the total cost of the perception task.

[0069] The above method of this embodiment is further introduced below.

[0070] As an optional embodiment, in step S103, similar initial perceptual subregions among the n initial perceptual subregions are divided into a set based on historical data and spatiotemporal correlation of the n initial perceptual subregions. The expression for the divided set of L perceptual subregions is:

[0071] U sim={U1,U2,...}={{u1,u2,u3},{u4,u5},...,{u m-2 ,u m-1 ,u m}}

[0072] Among them, {u1,u2,u3}, {u4,u5} and {u m-2 ,u m-1 ,u m} respectively represent that similar initial perception sub-regions in n initial perception sub-regions are divided into a set, U1 is a set of {u1,u2,u3}, U sim is the set of L perception sub-areas after division.

[0073] As an optional implementation method, step S104 matches any one or multiple participants with any initial perception sub-region in a subset of a part of the perception sub-region set in the L perception sub-region set based on a two-stage heuristic search algorithm to obtain multiple initial perception matching pairs, including: the first stage: selecting any participant with the closest geographical distance to any initial perception sub-region for matching to obtain multiple first perception matching pairs, and excluding multiple participants and multiple initial perception sub-regions in the multiple first perception matches in the second stage; the second stage: matching multiple participants remaining unmatched in the first stage with multiple initial perception sub-regions to obtain multiple second initial perception matching pairs; the iterative stage: adjusting the multiple first initial perception matching pairs and the multiple second initial perception matching pairs of the current cycle through the inference error to obtain multiple adjusted first perception matching pairs and multiple second perception matching pairs; based on the adjusted multiple first perception matching pairs and multiple second perception matching pairs, multiple initial perception matching pairs are obtained.

[0074] In this embodiment, the problem of assigning participants to perception units to perform tasks in each perception cycle can be regarded as a weighted bipartite graph optimal matching problem. All participants are represented as the left nodes of the bipartite graph, and all perception units are represented as the right nodes of the bipartite graph. For each participant p k And each perception unit u j , if p k KU j , then add a line from p to the bipartite graph k to u j The initial weight of the edge is defined as w kj .

[0075]

[0076] In the first stage, for U sim For each subset in , select the closest actor and perception unit pair (pk ,u j ) to match. Add this match to the initial matching set M0, and k and u j Exclude from the subsequent matching process to avoid repeated matching and resource waste. The main purpose of this step is to quickly reduce the number of uncovered perceptual units and ensure that at least one unit in each subset is guaranteed by preliminary data collection.

[0077] In the second stage, a bipartite graph G′=(P′,U′,E′) is constructed between the remaining unmatched participants and the perception units, and the KM algorithm is applied to find the optimal matching M′ on G′, considering the cost constraint C / m, where P′ represents the remaining participants excluding the participants in the first stage, U′ represents the remaining perception sub-areas excluding the perception sub-areas in the first stage, and E′ represents the weight of the edge in the second stage.

[0078] Finally, the algorithm merges the matching results of the two stages to form a matching solution for the current perception cycle.

[0079] M=M0∪M′

[0080] Iteration: After each perception cycle, the data inference and error calculation for the current cycle are performed. Based on the inference error, the weights of the edges in the bipartite graph are adjusted. (Specifically, if the data provided by a participant results in a large inference error, the weight of the edge connecting that participant to their assigned perception unit is increased; otherwise, the weight is decreased. The optimal matching process is repeated using the adjusted weights until the predetermined number of cycles is reached.)

[0081]

[0082] Among them, E base is the pre-set error limit, is the updated weight, Indicates the weights that need to be updated in the first or second stage, v ijk represents the value of the jth initial perception sub-area observed by participant k in the i-th perception cycle, D real (i, j) represents the true value of the jth initial perception sub-region observed in the i-th perception cycle.

[0083] As an optional implementation method, multiple initial perceptual matching pairs are obtained based on the adjusted multiple first perceptual matching pairs and multiple second perceptual matching pairs, including: merging the adjusted multiple first perceptual matching pairs and multiple second perceptual matching pairs to obtain multiple initial perceptual matching pairs.

[0084] In this embodiment, the adjusted multiple first perceptual matching pairs and the multiple second perceptual matching pairs are merged to obtain multiple initial perceptual matching pairs.

[0085] As an optional embodiment, in step S105, based on the spatiotemporal correlation of the L sets of perceptual sub-regions, the initial perceptual matrices in the multiple initial perceptual matching pairs are inferred using a data inference algorithm F, and the expression of the inferred data matrix is ​​obtained as follows:

[0086] D infer =F(D sense )

[0087] Among them, D infer is the inference data matrix, F is the data inference algorithm, D sense is the initial perception matrix.

[0088] As an optional implementation method, step S107 determines the target values ​​of multiple initial perceptual matching pairs based on the movement cost and geographical distance corresponding to the geographical distance of each initial perceptual matching pair in the multiple initial perceptual matching pairs, including: multiplying the movement cost corresponding to the geographical distance of each initial perceptual matching pair in the multiple initial perceptual matching pairs by the geographical distance to obtain the target value of each initial perceptual matching; adding the target values ​​of each initial perceptual matching pair to obtain the target values ​​of multiple initial perceptual matching pairs.

[0089] In this embodiment, the movement cost corresponding to the geographical distance of each initial perceptual matching pair in the multiple initial perceptual matching pairs is multiplied by the geographical distance to obtain the target value of each initial perceptual matching, and the target values ​​of the multiple initial perceptual matching pairs are added together to obtain the target values ​​of the multiple initial perceptual matching pairs.

[0090] In an embodiment of the present invention, by obtaining a perception task and a total cost of the perception task; dividing the perception task into m perception cycles in time, and dividing the perception task into n initial perception sub-regions in the perception area; based on the historical data and spatiotemporal correlation of the n initial perception sub-regions, dividing similar initial perception sub-regions in the n initial perception sub-regions into a set to obtain L perception sub-region sets after division; when any one or more participants are in any one of the m perception cycles, matching any one or more participants with any initial perception sub-region in a subset of a partial perception sub-region set in the L perception sub-region set based on a two-stage heuristic search algorithm to obtain multiple initial perception matching pairs, wherein each initial perception matching pair includes the geographical distance from any one or more participants to any initial perception sub-region in the subset of the partial perception sub-region set, the target perception value and the movement cost corresponding to the geographical distance, and the target perception value is calculated based on the credibility and The initial perception value is determined; based on the spatiotemporal correlation of the set of L perception sub-areas, the initial perception matrix in the multiple initial perception matching pairs is inferred using the data inference algorithm F to obtain an inferred data matrix, wherein the initial perception matrix is ​​composed of multiple target perception values; based on the real data matrix and the inferred data matrix, the error between the real data matrix and the inferred data matrix is ​​determined; based on the movement cost and geographical distance corresponding to the geographical distance of each initial perception matching pair in the multiple initial perception matching pairs, the target values ​​of the multiple initial perception matching pairs are determined; when the target values ​​of the multiple initial perception matching pairs are less than the total cost of the perception task, the perception matching pair set corresponding to the minimum error between the real data matrix and the inferred data matrix is ​​determined, which solves the current technical problem of poor accuracy of data inference in sparse crowd perception task allocation, that is, poor quality of perception data, and achieves task allocation under the constraint of total cost in the process of sparse crowd perception task allocation, thereby improving the accuracy of data inference, that is, improving the quality of perception data.

[0091] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0092] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0094] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected to achieve the purpose of the present embodiment according to actual needs.

[0095] In addition, the functional units in various embodiments of the present invention may be integrated into a first processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0096] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A sparse crowd-sensing task allocation method based on two-stage heuristic search, characterized by: include: Get the perception task and the total cost of the perception task; The perception task is divided into m perception cycles in time, and the perception task is divided into n initial perception sub-areas in the perception area; Based on the historical data and spatiotemporal correlation of the n initial perception sub-regions, similar initial perception sub-regions among the n initial perception sub-regions are divided into a set to obtain L perception sub-region sets after division; When any one or more participants are in any one of the m perception cycles, the any one or more participants are matched with any initial perception sub-region in the subset of the partial perception sub-region set in the L perception sub-region set based on a two-stage heuristic search algorithm to obtain multiple initial perception matching pairs, wherein each initial perception matching pair includes the geographical distance from any one or more participants to any initial perception sub-region in the subset of the partial perception sub-region set, the target perception value, and the movement cost corresponding to the geographical distance, and the target perception value is determined according to the reputation and initial perception value of any one or more participants; Based on the spatiotemporal correlation of the L sets of perception sub-regions, the data inference algorithm F is used to infer the initial perception matrix of multiple initial perception matching pairs to obtain an inferred data matrix, where the initial perception matrix is ​​composed of multiple target perception values; determining an error between the true data matrix and the inferred data matrix based on the true data matrix and the inferred data matrix; determining target values ​​for the plurality of initial perceptual matching pairs based on a movement cost and a geographic distance corresponding to a geographic distance of each of the plurality of initial perceptual matching pairs; When the target values ​​of the plurality of initial perceptual matching pairs are less than the total cost of the perceptual task, a set of perceptual matching pairs corresponding to the minimum error between the true data matrix and the inferred data matrix is ​​determined.

2. The method according to claim 1, characterized in that Based on the historical data and spatiotemporal correlation of the n initial perception sub-regions, similar initial perception sub-regions in the n initial perception sub-regions are divided into a set. The expression of the divided set of L perception sub-regions is: U sim ={U1,U2,...}={{u1,u2,u3},{u4,u5},...,{u m-2 ,u m-1 ,u m }} Among them, {u1,u2,u3}, {u4,u5} and {u m-2 ,u m-1 ,u m } respectively represent that similar initial perception sub-regions in n initial perception sub-regions are divided into a set, U1 is a set of {u1,u2,u3}, U sim is the set of L perception sub-areas after division.

3. The method according to claim 1, characterized in that Based on a two-stage heuristic search algorithm, any one or more participants are matched with any initial perception sub-region in a subset of the perception sub-region sets in the L perception sub-region sets to obtain multiple initial perception matching pairs, including: Phase 1: any participant with the closest geographical distance to any initial perception sub-region is selected for matching, to obtain multiple first initial perception matching pairs, and multiple participants and multiple initial perception sub-regions in the multiple first initial perception matching pairs are excluded in the second phase; The second stage: matching the remaining unmatched participants in the first stage with the multiple initial perception sub-areas to obtain multiple second initial perception matching pairs; Iteration stage: multiple first initial perception matching pairs and multiple second initial perception matching pairs of the current cycle are adjusted according to the inference error to obtain multiple adjusted first perception matching pairs and multiple second perception matching pairs; Based on the adjusted multiple first perceptual matching pairs and multiple second perceptual matching pairs, multiple initial perceptual matching pairs are obtained.

4. The method according to claim 3, characterized in that Based on the adjusted plurality of first perceptual matching pairs and the plurality of second perceptual matching pairs, a plurality of initial perceptual matching pairs are obtained, including: The adjusted multiple first perceptual matching pairs and the multiple second perceptual matching pairs are merged to obtain multiple initial perceptual matching pairs.

5. The method according to claim 1, wherein Based on the spatiotemporal correlation of the L sets of perception sub-regions, the data inference algorithm F is used to infer the initial perception matrix of multiple initial perception matching pairs, and the expression of the inferred data matrix is ​​obtained as follows: D infer =F(D sense ) Among them, D infer is the inference data matrix, F is the data inference algorithm, D sense is the initial perception matrix.

6. The method according to claim 1, characterized in that Determining target values ​​of the plurality of initial perceptual matching pairs based on the movement cost and the geographical distance corresponding to the geographical distance of each of the plurality of initial perceptual matching pairs includes: Multiplying the movement cost corresponding to the geographical distance of each initial perception matching pair in the plurality of initial perception matching pairs by the geographical distance to obtain a target value for each initial perception matching pair; The target values ​​of each of the multiple initial perceptual matches are added together to obtain the target values ​​of the multiple initial perceptual matching pairs.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

Images