Spatial crowdsourcing scheduling system and method based on edge computing

Abstract

The present invention relates to a spatial crowdsourcing scheduling system based on edge computing, including a central control layer, an edge computing layer and a user device layer; when the central control layer receives the request information of the user device, it will respectively calculate the edge node closest to the user device and send the request task to the corresponding edge node for processing; the edge computing layer includes a number of road network nodes; each road network node is an edge node, responsible for processing the computing tasks issued by the cloud server; the user device layer includes mobile devices and wireless sensors of task initiators and crowdsourcing workers, which are used to upload their respective task request information. The present invention effectively downgrades and reduces the service response time, while avoiding the leakage of location privacy data.

Description

Technical Field

[0001] The present invention relates to the field of spatial crowdsourcing scheduling, and in particular to a spatial crowdsourcing scheduling system and method based on edge computing. Background Art

[0002] The development of 5G technology has promoted the widespread application of intelligent services, and the "crowdsourcing" service platform based on location trajectory has developed particularly rapidly. According to Marketsand Markets, the global market revenue of intelligent services based on location trajectory will grow from US$17.8 billion in 2020 to US$39.2 billion in 2025, with a compound annual growth rate of 17.1% during the forecast period. "Crowdsourcing" is defined as "a business model that outsources tasks traditionally performed by company employees to a large group of free and voluntary non-specific general public in an open form." "Crowdsourcing" has been widely used in various fields, including environmental monitoring, healthcare, Meituan, Didi Taxi, Amazon's Turkish Robot, etc. At the same time, everyone with a mobile terminal smart device can register on the Internet to become a worker in the crowdsourcing platform. In this context, an emerging crowdsourcing model that outsources tasks with spatial characteristics to workers on the Internet has emerged, called spatial crowdsourcing (also known as mobile crowdsourcing). In the spatial crowdsourcing platform, people can query location-related service information at any time and enjoy convenient and fast services.

[0003] In the "spatial crowdsourcing" platform, how to select appropriate users for task assignment and ensure the quality and efficiency of task matching is the main challenge currently faced. First of all, most of the existing spatial crowdsourcing scheduling schemes use Euclidean distance as the basis for task assignment. Such schemes are simple to calculate and highly efficient. However, since they ignore the underlying road network information, the service quality needs to be improved. In addition, most of the current spatial crowdsourcing scheduling schemes are based on central servers and rarely take distributed ideas into consideration. As a result, each task initiator and worker needs to upload their own location information to the cloud server for calculation, which greatly increases the server's computing overhead, thereby affecting the service efficiency of the spatial crowdsourcing platform. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a spatial crowdsourcing scheduling system and method based on edge computing, which effectively downgrades and reduces service response time while avoiding the leakage of location privacy data.

[0005] To achieve the above object, the present invention adopts the following technical solution:

[0006] A spatial crowdsourcing scheduling system based on edge computing includes a central control layer, an edge computing layer and a user device layer; when the central control layer receives request information from a user device, it will respectively calculate the edge nodes closest to the user device and send the request task to the corresponding edge node for processing; the edge computing layer includes a plurality of network nodes; each network node is an edge node responsible for processing the computing tasks sent by a cloud server; the user device layer includes mobile devices and wireless sensors of task initiators and crowdsourcing workers, which are used to upload their respective task request information.

[0007] A scheduling method for a spatial crowdsourcing scheduling system based on edge computing, comprising the following steps:

[0008] Construct a road network structure. Based on the road network structure, calculate the road network distance between the task initiator and the crowdsourcing worker;

[0009] Use a bidirectional dynamic K nearest neighbor mechanism to find the nearest object;

[0010] When new query points and data objects appear in the road network, they are inserted into the query point set Q and the latest data object set O. Correspondingly, when the query points and data objects are successfully matched, they are deleted from the query point set Q and the data object set O.

[0011] For the query point Q i Perform a forward dynamic K nearest neighbor query for the data object O j Perform a reverse dynamic K-nearest neighbor query.

[0012] Furthermore, the road network structure is a directed weighted graph N = (V, E, W), where V represents the set of intersections between road segments in the road network, E represents the set of road segments in the road network, and W represents the weights corresponding to the road segments, wherein the intersection V i ∈V refers to a node formed by the confluence of two or more road sections.

[0013] Furthermore, the road network distance between the task initiator and the crowdsourcing worker is calculated by formula (1):

[0014]

[0015] When the query point Q i With data object O j If they are adjacent, the road network distance between them is the Euclidean distance; otherwise, the road network node v closest to the query point should be found. x , v y ; Then calculate the query point to node v respectively x The distance between the internal data object and the query point to v yThe distance between the two data objects, and then the data object corresponding to the minimum value of the two is the optimal worker near the task initiator;

[0016] where v x and v y They are query points Q i Nearby left and right nodes, d(Q i ,v x ), d(v p ,O m ), d(Q i ,v y ) and d(v q ,Q n ) are calculated by Euclidean distance, NDM(v x ,v p ), NDM(v y ,v q )Through A * The algorithm performs extended precomputation.

[0017] Furthermore, the bidirectional dynamic K nearest neighbor query mechanism includes a forward dynamic K nearest neighbor query algorithm and a reverse dynamic K nearest neighbor query algorithm.

[0018] Furthermore, the forward dynamic K nearest neighbor query algorithm is specifically as follows:

[0019] 1) In the pre-computation phase, the road network node matrix and the candidate sets of internal data objects in the road network nodes are constructed respectively;

[0020] 2) In the query phase, the road network nodes near the query point are searched within the initial radius of r miles, and the distances between the query point and the online-hailing vehicles in the road network nodes are compared. The online-hailing vehicle with the smallest distance is the one closest to the passenger, that is, the driver and passenger are successfully matched, and the algorithm ends. Otherwise, k = r + α, and the search continues, where α is the expansion radius parameter;

[0021] 3) In the dynamic update phase, newly appeared query points and data objects are inserted into sets Q and O, and successfully matched query points and data objects are deleted from Q and O.

[0022] Furthermore, the reverse dynamic K nearest neighbor query algorithm is specifically as follows:

[0023] 1) In the pre-computation phase, the road network node matrix and the candidate sets of internal data objects in the road network nodes are constructed respectively;

[0024] 2) In the query phase, search for road network nodes near the query point within the initial radius of r miles, and compare the distances between the data object and the query point in the road network node. The query point with the smallest distance is the passenger closest to the driver, that is, the driver and passenger are matched successfully, and the algorithm ends. Otherwise, k = r + α, and continue searching, where α is the expansion radius parameter;

[0025] 3) In the dynamic update phase, newly appeared query points and data objects are inserted into sets Q and O, and successfully matched query points and data objects are deleted from Q and O.

[0026] Furthermore, the bidirectional dynamic K nearest neighbor mechanism includes forward dynamic K nearest neighbor query (PDKNN) and reverse dynamic K nearest neighbor query (RDKNN). i Execute PDKNN query, for the data object O that appears j Execute RDKNN query. In the early stage of these two query processes, the corresponding target object is searched within the range of r miles. If the search is successful, the algorithm ends. Otherwise, k=r+α, and the query range is continuously expanded until the search is successful.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. The central server of the central control layer of the present invention distributes the query request uploaded by the user device layer to the edge node closest to the device, and the edge node of the edge computing layer processes the corresponding data, thereby sharing the computing task of the central server and reducing the service response time;

[0029] 2. The bidirectional dynamic K-nearest neighbor scheduling algorithm based on edge computing of the present invention can ensure that user data is only processed at the edge of the network, so that the data will only be stored, analyzed and processed locally, avoiding the leakage of location privacy data. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the system model of the present invention;

[0031] Figure 2 This is a road network modeling process in one embodiment of the present invention;

[0032] Figure 3 This is an example of road network distance calculation in one embodiment of the present invention;

[0033] Figure 4 It is a bidirectional dynamic K nearest neighbor mechanism in one embodiment of the present invention;

[0034] Figure 5 is a flow chart of a forward dynamic k-nearest neighbor query algorithm in one embodiment of the present invention;

[0035] Figure 6is a flow chart of a reverse dynamic k-nearest neighbor query algorithm in one embodiment of the present invention;

[0036] Figure 7 is a flow chart of a dynamic update mechanism in one embodiment of the present invention;

[0037] Figure 8 It is a flow chart of a bidirectional dynamic k-nearest neighbor query mechanism in one embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0039] Please refer to Figure 1 The present invention provides a spatial crowdsourcing scheduling system based on edge computing, including a central control layer, an edge computing layer and a user device layer; when the central control layer receives the request information of the user device, it will respectively calculate the edge nodes closest to the user device and send the request task to the corresponding edge node for processing; the edge computing layer includes a plurality of road network nodes; each road network node is an edge node, which is responsible for processing the computing tasks issued by the cloud server; the user device layer includes mobile devices and wireless sensors of task initiators and crowdsourcing workers, which are used to upload their respective task request information.

[0040] Figure 1 In the figure, the left data flow refers to the forward dynamic K-nearest neighbor query, which starts with the task initiator to find the best worker near the task initiator. Correspondingly, the right side is the reverse dynamic K-nearest neighbor query, which starts with the crowdsourcing worker to find the best task initiator near the crowdsourcing worker.

[0041] In this embodiment, a scheduling method of a spatial crowdsourcing scheduling system based on edge computing includes the following steps:

[0042] Construct a road network structure. Based on the road network structure, calculate the road network distance between the task initiator and the crowdsourcing worker;

[0043] Use a bidirectional dynamic K nearest neighbor mechanism to find the nearest object;

[0044] When new query points and data objects appear in the road network, they are inserted into the query point set Q and the latest data object set O. Correspondingly, when the query points and data objects are successfully matched, they are deleted from the query point set Q and the data object set O.

[0045] For the query point Q i Perform a forward dynamic K nearest neighbor query for the data object O j Perform a reverse dynamic K-nearest neighbor query.

[0046] refer to Figure 2In this embodiment, the road network structure is a directed weighted graph N = (V, E, W), such as Figure 2 As shown in (b), V represents the set of intersections between road segments in the road network, E represents the set of road segments in the road network, and W represents the weight corresponding to the road segment. i ∈V refers to a node formed by the confluence of two or more road sections.

[0047] In this embodiment, the road network distance between the task initiator and the crowdsourcing worker is calculated by formula (1):

[0048]

[0049] When the query point Q i With data object O j If they are adjacent, the road network distance between them is the Euclidean distance; otherwise, the road network node v closest to the query point should be found. x , v y ; Then calculate the query point to node v respectively x The distance between the internal data object and the query point to v y The distance between the two data objects, and then the data object corresponding to the minimum value of the two is the optimal worker near the task initiator;

[0050] like Figure 3 As shown, where v x and v y They are query points Q i Nearby left and right nodes, d(Q i ,v x ), d(v p ,O m ), d(Q i ,v y ) and d(v q ,O n ) are calculated by Euclidean distance, NDM(v x ,v p ), NDM(v y ,v q )Through A * The algorithm performs extended precomputation.

[0051] In this embodiment, the bidirectional dynamic K nearest neighbor query mechanism includes a forward dynamic K nearest neighbor query algorithm and a reverse dynamic K nearest neighbor query algorithm.

[0052] In this embodiment, the forward dynamic K nearest neighbor query algorithm (PDKNN) is given a set of data objects O and a set of query points Q. This query can find the objects from the O set within K miles of the query point Q. i The most recent data object. Figure 4As shown, the red solid circle area is the PDKNN query, where the query point Q1 can find the nearest object in the area of ​​radius k1, specifically:

[0053] 1) In the pre-computation phase, the road network node matrix and the candidate sets of internal data objects in the road network nodes are constructed respectively;

[0054] 2) In the query phase, the road network nodes near the query point are searched within the initial radius of r miles, and the distances between the query point and the online-hailing vehicles in the road network nodes are compared. The online-hailing vehicle with the smallest distance is the one closest to the passenger, that is, the driver and passenger are successfully matched, and the algorithm ends. Otherwise, k = r + α, and the search continues, where α is the expansion radius parameter;

[0055] 3) In the dynamic update phase, newly appeared query points and data objects are inserted into sets Q and O, and successfully matched query points and data objects are deleted from Q and O.

[0056] The relevant pseudo code is as follows:

[0057]

[0058] In this embodiment, the reverse dynamic K nearest neighbor query algorithm (RDKNN) refers to a given set of data objects O and a set of query points Q. This query can find the nearest neighbor from the data object O within a range of K miles from the set Q. i The most recent query point Q i .like Figure 4 As shown, the blue dotted circle is the RDKNN query, where the query point O7 can find the nearest data object in the area of ​​radius k2. Specifically:

[0059] The algorithm is also divided into pre-calculation phase, query phase and update phase. The pre-calculation phase and update phase are the same as the PDKNN algorithm steps, the difference is that the RDKNN algorithm searches for road network nodes near the query point within the initial radius of r miles in the query phase, and compares the distance between the data object and the query point in the road network node. The query point with the smallest distance is the passenger closest to the driver, that is, the driver and passenger match successfully, and the algorithm ends. Otherwise, k = r + α, and continue to search, where α is the expansion radius parameter.

[0060] The relevant pseudo code is as follows:

[0061]

[0062]

[0063] In this embodiment, the dynamic update mechanism means that when new query points and data objects appear in the road network, they need to be inserted into the query point set Q and the latest data object set O. Correspondingly, when the query points and data objects are successfully matched respectively, they need to be deleted from the query point set Q and the data object set O.

[0064]

[0065]

[0066] In this embodiment, the bidirectional dynamic K nearest neighbor mechanism includes a forward dynamic K nearest neighbor query (PDKNN) and a reverse dynamic K nearest neighbor query (RDKNN). i Execute PDKNN query, for the data object O that appears j Execute RDKNN query. In the early stage of these two query processes, the corresponding target object is searched within the range of r miles. If the search is successful, the algorithm ends. Otherwise, k=r+α, and the query range is continuously expanded until the search is successful.

[0067]

[0068]

[0069] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A scheduling method for a spatial crowdsourcing scheduling system based on edge computing, characterized in that: The following steps are involved: Construct a road network structure. Based on the road network structure, calculate the road network distance between the task initiator and the crowdsourcing worker; Use a bidirectional dynamic K nearest neighbor mechanism to find the nearest object; When new query points and data objects appear in the road network, they are inserted into the query point set Q and the latest data object set O. Correspondingly, when the query points and data objects are successfully matched, they are deleted from the query point set Q and the data object set O. For the query point Q i Perform a forward dynamic K nearest neighbor query for the data object O j Perform reverse dynamic K nearest neighbor query; The bidirectional dynamic K nearest neighbor query mechanism includes a forward dynamic K nearest neighbor query algorithm and a reverse dynamic K nearest neighbor query algorithm; The forward dynamic K nearest neighbor query algorithm is specifically as follows: 1) In the pre-computation phase, the road network node matrix and the candidate sets of internal data objects in the road network nodes are constructed respectively; 2) In the query phase, search for road network nodes near the query point within the initial radius of r miles, and compare the distances between the query point and the online-hailing vehicles in the road network nodes. The online-hailing vehicle with the smallest distance is the one closest to the passenger, that is, the driver and passenger are successfully matched, and the algorithm ends. Otherwise, k = r + α, and continue searching, where α is the expansion radius parameter, k represents the query radius of the forward dynamic K nearest neighbor query algorithm and the reverse dynamic K nearest neighbor query algorithm, and r is the initial radius; 3) In the dynamic update phase, newly appeared query points and data objects are inserted into sets Q and O, and successfully matched query points and data objects are deleted from Q and O; The reverse dynamic K nearest neighbor query algorithm is specifically as follows: 1) In the pre-computation phase, the road network node matrix and the candidate sets of internal data objects in the road network nodes are constructed respectively; 2) In the query phase, search for road network nodes near the query point within the initial radius of r miles, and compare the distances between the data object and the query point in the road network node. The query point with the smallest distance is the passenger closest to the driver, that is, the driver and passenger are matched successfully, and the algorithm ends. Otherwise, k = r + α, and continue searching, where α is the expansion radius parameter; 3) In the dynamic update phase, newly appeared query points and data objects are inserted into sets Q and O, and successfully matched query points and data objects are deleted from Q and O.

2. The scheduling method of the spatial crowdsourcing scheduling system based on edge computing according to claim 1 is characterized in that: The road network structure is a directed weighted graph N = (V, E, W), where V represents the set of intersections between road segments in the road network, E represents the set of road segments in the road network, and W represents the weights corresponding to the road segments. i ∈V refers to a node formed by the confluence of two or more road sections.

3. The scheduling method of the spatial crowdsourcing scheduling system based on edge computing according to claim 1 is characterized in that: The road network distance between the task initiator and the crowdsourcing worker is calculated by formula (1): In the formula, v p 、v q is a road network node in the road network, O m , O n Data objects in the route network, L a1 , L a2 Q i and O j Latitude; L o1 , L o2 Q i and O j longitude; When the query point Q i With data object O j If they are adjacent, the road network distance between them is the Euclidean distance; otherwise, the road network node v closest to the query point should be found. x , v y ; Then calculate the query point to node v respectively x The distance between the internal data object and the query point to v y The distance between the two data objects, and then the data object corresponding to the minimum value of the two is the optimal worker near the task initiator; where v x and v y They are query points Q i Nearby left and right nodes, d(Q i ,v x ), d(v p ,O m ), d(Q i ,v y ) and d(v q ,O n ) are calculated by Euclidean distance, NDM(v x ,v p ), NDM(v y ,v q )Through A * The algorithm performs extended precomputation.

4. The scheduling method of the spatial crowdsourcing scheduling system based on edge computing according to claim 1 is characterized in that: The bidirectional dynamic K nearest neighbor mechanism includes forward dynamic K nearest neighbor query PDKNN and reverse dynamic K nearest neighbor query RDKNN. i Execute PDKNN query, for the data object O that appears j Execute RDKNN query. In the early stage of these two query processes, the corresponding target object is searched within the range of r miles. If the search is successful, the algorithm ends. Otherwise, k=r+α, and the query range is continuously expanded until the search is successful.

5. A spatial crowdsourcing scheduling system based on edge computing, characterized in that: The system is implemented by the scheduling method described in any one of claims 1 to 4, including a central control layer, an edge computing layer and a user device layer; when the central control layer receives the request information of the user device, it will respectively calculate the edge nodes closest to the user device and send the request task to the corresponding edge node for processing; the edge computing layer includes a number of road network nodes; each road network node is an edge node, responsible for processing the computing tasks issued by the cloud server; the user device layer includes mobile devices and wireless sensors of task initiators and crowdsourcing workers, which are used to upload their respective task request information.

Images