Dynamic task matching method oriented to crowd sensing and based on Skyline query

Abstract

The invention discloses a dynamic task matching method oriented to crowd sensing and based on Skyline query, and aims to solve the matching problem under the condition that a large number of tasks and participants arrive dynamically. According to the method, the time sensitivities of the tasks and the participants are fully considered, and meanwhile, the constraint relation between the tasks and the participants and the advantages and disadvantages of the attributes of the participants are considered. A task matching window model is used for processing dynamically arriving tasks and data objects, candidates are screened out based on time perception, expressions of participants in multiple dimensions are comprehensively evaluated by means of algorithms such as k-space composite Skyline query and the like, and finally k participants with excellent expressions in all aspects are selected to execute tasks. According to the method, the success rate and quality of task matching are improved through time constraint and Skyline query, and the matching problem of tasks and participants in a dynamic environment is effectively solved.

Description

Technical Field

[0001] The present invention belongs to the field of application of data query in task allocation, and specifically is a dynamic task matching method based on skyline query for crowd intelligence perception. Background Art

[0002] In the context of the rapid development of the Internet and sensor technology, crowd sensing technology, as an emerging means of data collection and monitoring, is playing an increasingly important role in various fields. This technology relies on the participation of large-scale people to collect data through their mobile devices, thereby realizing real-time monitoring and management of the environment, traffic, public safety, etc. However, how to efficiently manage tasks and select high-quality and suitable participants for task matching among many participants has always been a key challenge. This not only affects the quality of perception data and the efficiency of task completion, but is also directly related to the cost and resource utilization of the entire crowd sensing system.

[0003] Previous studies have mostly focused on static task allocation, with less consideration of the actual situation in which tasks and participants arrive dynamically on the platform, ignoring the time sensitivity of tasks and participants, as well as the impact of the participants' own attributes on the quality and efficiency of task completion.

[0004] As an important means of data query, applying skyline query to participant selection in crowd sensing can give full play to the multi-objective decision-making properties of skyline query and assist the platform in finding participants with better performance in multiple dimensions. Skyline query can flexibly adapt to dynamic changes in crowd sensing systems, such as changes in participant positions or changes in task requirements, so that the system can adjust in real time and allocate resources in the best way. Summary of the invention

[0005] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a dynamic task matching method based on skyline query for crowd intelligence perception, which aims to use the task matching window model to flexibly handle dynamically arriving tasks and participants, solve the problem of time sensitivity of tasks and participants through time perception, and further screen k participants with better performance in all aspects for the task through the k-space composite skyline query algorithm and return them to the task requester, so as to realize dynamic matching of participants and tasks.

[0006] The technical solution adopted by the present invention to solve the technical problem is: a dynamic task matching method based on skyline query for crowd intelligence perception, which mainly includes the following steps:

[0007] Step 1: Use a task matching window model that includes a double sliding window and a waiting queue mechanism to process dynamically arriving tasks and participants. This model includes a task sliding window SW p , a participant sliding window SW w and a waiting queue Q. Among them, SW p , SW w , and Q all adopt the data structure of a priority queue. Select the task p i to be processed according to the time priority scheduling algorithm, and sort the objects in SW p , Q, and SW w in the order of time priority;

[0008] Step 2: Start matching participants for the task p i to be processed. First, use the k-space composite Skyline query algorithm to screen out the participants who meet the time constraints in the current participant sliding window, form a participant candidate set C, and judge the relationship between the number of participants |C| in C and the number of participants k required by the task p i . When |C| is greater than k, go to Step 3;

[0009] Step 3: When the number of people in the participant candidate set C is greater than k, call the k-space composite Skyline query algorithm to further process C, screen out the participants who perform relatively well in each given attribute dimension for the current task p i , and form a query set CS; compare the number of participants |CS| in CS with the number of participants k required by the task. If |CS| = k, directly return CS as the matching result to the task requester; if |CS| > k, call the evaluation function based on the dominated area to calculate the evaluation function values of the participants, and select the k participants with the largest evaluation function values to form a screening set CS a as the matching result; if |CS| < k, call the alternate algorithm to calculate k - |CS| alternate participants to make up for the lack of the number of participants, and jointly form the matching result with the participants in CS;

[0010] Step 4: Call the task matching window model maintenance strategy to maintain the two sliding windows during the matching process;

[0011] Furthermore, the specific steps of the time priority scheduling algorithm mentioned in Step 1 include:

[0012] (1) Calculate the time priority index TPI of the newly arrived tasks in SW p , Q and the newly arrived participants in SW w ;

[0013] (2) Sort SW p , Q, and SW wSort the objects in the queue, put the object with the highest TPI at the head of each queue, and return the task with the highest TPI as the pending task p i , output the sorted SW p , S.W. w ;

[0014] Furthermore, in the priority scheduling algorithm (1), the time priority index where t s is the time when the current task starts to execute or the time when the participant starts working, t c refers to the current time, where l is a small positive number used to avoid the situation where the denominator is zero, that is, when t s =t c The closer the object is to the current time, the higher its time priority.

[0015] Furthermore, the participant screening method based on time perception in step 2 specifically includes the following steps:

[0016] (1) Initialize index index and traverse the current participant sliding window SW w , get the current participant's working time w i .(st,et);

[0017] (2) If participant w i Working hours i .(st,et) contains task p i The request execution time p i .(t start ,t end ),Right now Then w i Add to the participant candidate set C, where w i .st indicates the participant’s work start time, w i .et indicates the end time of the participant's work, p i .t start Indicates the time when the task needs to start executing, p i .t end Indicates the task execution end time;

[0018] (3) Compare the number of participants |C| in the candidate set C with the number of participants k required for the task:

[0019] a. If |C|=k, the number of participants in the selection C is exactly equal to the number of participants k required for the current task, and C is directly returned as the matching result;

[0020] b. If |C| < k and the number of participants meeting the time constraint fails to reach the number of participants required for the current task, the task matching fails;

[0021] c. If |C| > k, call the k - space composite Skyline query algorithm to further screen high - quality participants, and calculate the time fitness η(w i ,p i ), where w i .t = w i .(st, et), p i .t = p i .(t start ,t end ), η(w i ,p i ) evaluates the time fitness between participant w i and task p i . The larger this value is, the higher the matching degree between participant w i and task p i . Meanwhile, update the time fitness dimension in the participant attribute table;

[0022] Furthermore, the k - space composite Skyline query algorithm mentioned in step three specifically includes the following steps:

[0023] (1) Calculate the distance d(w i , p i ) between participant w i and task p i , and update the participant attribute table;

[0024] (2) Initialize the index variable index to traverse the participants in the candidate set C, and obtain the attribute values of each dimension of the current participant w i ;

[0025] (3) Calculate the dominance of w i in all two - dimensional sub - spaces U xy . U xy is used to represent the distribution and comparison of participants in two specific dimensions, where x and y represent two different attributes, such as price and distance. If w i is a Skyline point in U xy , then U xy is the dominance space A i of w xy . Add the dimension pair (x, y) of A xy to the dimension pair set R i of participant w i ;

[0026] (4) Initialize the index variable index again to traverse the participants in C and obtain the participant w i 's dimension pair set R i ;

[0027] (5) If the union of the dominance spaces A i of w xy (w i ) covers the data space D, that is add w i to the query set CS;

[0028] (6) Compare the number of participants |CS| in CS with the number of participants k required by the task. If |CS| = k, directly return CS as the matching result to the task requester; if |CS| > k, call the evaluation function based on the dominance area to calculate the evaluation function values of the participants, and select the k participants with the largest evaluation function values to form the screening set CS a as the matching result; if |CS| < k, call the alternative algorithm to calculate k - |CS| alternative participants to make up the missing number of participants, and jointly form the matching result with the participants in CS;

[0029] Further, the evaluation function algorithm based on the dominance area mentioned in the k - space composite Skyline query algorithm specifically includes the following steps:

[0030] (1) Obtain the dimension pair set R i of each participant w i , and query its dominance space A xy (w i );

[0031] (2) Calculate the dominance area S i on each dominance space A xy of the participant w xy (w i ), where the dominance area S xy (w i ) is divided into the following three cases:

[0032] When the participant w i is in the two - dimensional subspace U xy , if the values of both attribute dimensions are越小越好 (the smaller the better), then the dominance area size of w i is recorded as S xy (w i )=(1 - p i [r])(1 - p i [m])(r,m∈[1,d]);

[0033] When the participant w i is in the two - dimensional subspace Uxy If the values ​​of both attribute dimensions are as large as possible, then w i The size of the dominant area is denoted as S xy (w i )=p i [r]p i [m](r,m∈[1,d]);

[0034] When participants i In the two-dimensional subspace U xy If the value of one attribute dimension is as small as possible and the value of another attribute dimension is as large as possible, then w i The size of the dominant area is denoted as S xy (w i )=(1-p i [r])(p i [m])(r,m∈[1,d]);

[0035] (3) Calculate the sum of the areas of each participant's controlled space

[0036] (4) Rearrange the participants in descending order of evaluation function values, and the top k participants form the CS a ;

[0037] (5) Return to CS a Give k-space composite skyline query algorithm;

[0038] Furthermore, the candidate participant query algorithm mentioned in the k-space composite skyline query algorithm has the following specific steps:

[0039] (1) Get the participant w i The dimension of the set R i , calculate w i Dominate Space A xy (w i )

[0040] (2) Arrange the participants in C according to their control space A xy (w i ) Sort the numbers in descending order;

[0041] (3) When w i With w j When the number of dominated spaces is the same, the evaluation function algorithm based on the dominated area is called to calculate w i With w j The size of the dominant area is w i With w j Sort again;

[0042] (4) Add the top k-|CS| participants to CS b middle;

[0043] (5) Return to CS b Give k-space composite skyline query algorithm;

[0044] Furthermore, the specific steps of the task sliding window matching model mentioned in step 4 include:

[0045] The specific steps of the task matching window model maintenance strategy are as follows:

[0046] Task sliding window maintenance strategy:

[0047] (1) If task p i With k participants w i Matching success or task p i Expired, the p i Remove from the task sliding window;

[0048] (2) If the number of tasks in the current task sliding window is less than the size of the task sliding window N p , add the tasks that have newly arrived in the task window to the task sliding window according to their time priority;

[0049] Participant sliding window window maintenance strategy:

[0050] (1) If participant w i With p i Match successful or w i If the working time limit is exceeded, i Remove from the task sliding window;

[0051] (2) If the number of tasks in the current participant sliding window is less than the size of the participant sliding window N w , add the participants who newly arrive at the participant window to the participant sliding window according to time priority.

[0052] The beneficial effects of the present invention are as follows: the present invention solves the following problems by adopting a dynamic task matching method based on skyline query for crowd intelligence perception:

[0053] (1) Through the time-aware participant screening method, the problem of traditional static task allocation ignoring the actual available time of participants and the actual needs of tasks, resulting in resource waste and task execution delays, is effectively solved. At the same time, the concept of time fit is introduced to improve the resource utilization efficiency of participants.

[0054] (2) The task matching window model uses a double sliding window and waiting queue mechanism to simultaneously process dynamically arriving tasks and participants, improving the platform's real-time response capability and task management capabilities. On the other hand, by setting the time priority index to sort tasks and participants, it can flexibly handle urgent tasks and participants and achieve efficient matching of tasks and participants.

[0055] (3) Through the k-space composite skyline query algorithm, multi-objective comprehensive decision-making is carried out to screen out the better participants. At the same time, it is avoided that some participants are selected because they are too outstanding in one dimension, while other participants are screened out because they perform slightly worse in this dimension but better in other dimensions. By improving the quality of perception personnel, the quality of perception data can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a flow chart of a dynamic task matching method based on skyline query for crowd intelligence perception of the present invention.

[0057] Figure 2 It is a schematic diagram of the task matching window model provided by the present invention.

[0058] Figure 3 This is an example diagram of task objects in the task sliding window at time t provided by the present invention.

[0059] Figure 4 It is an example diagram of participant objects in the participant sliding window at time t given by the present invention.

[0060] Figure 5 This is an example diagram of the participant's dominated space and dominated area in the result set CS at time t given by the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described examples are only some implementation examples of the present invention, rather than all the embodiments. In addition, it should be understood that after reading the present invention, those skilled in the art can make various modifications or changes to the present invention, and these equivalent forms also fall within the scope defined by the claims attached to the application.

[0062] The present invention discloses a dynamic task matching method based on skyline query for crowd intelligence perception, and the specific operation process includes:

[0063] Step 1: Use a task matching window model that includes a double sliding window and a waiting queue mechanism to handle dynamically arriving tasks and participants. The model includes a task sliding window SW p , Participant sliding window SW w and a waiting queue Q, where SWp , SW w , Q both use the data structure of priority queue. The specific working process is as follows: Figure 2 As shown, according to the time priority scheduling algorithm, the pending task p is selected i , and output SW sorted in time priority order p , Q and SW w ;

[0064] Specifically, the main steps of the time priority scheduling algorithm mentioned are as follows:

[0065] (1) Calculate SW p , new tasks and SW in Q w Time priority index TPI of new participants;

[0066] (2) According to the TPI value, SW p , Q and SW w Sort the objects in the queue, put the object with the highest TPI at the head of each queue, and return the task with the highest TPI as the pending task p i , output the sorted SW p , S.W. w ;

[0067] Specifically, the time priority index mentioned in the priority scheduling algorithm (1) where t s is the time when the current task starts to execute or the time when the participant starts working, t c refers to the current time, where l is a small positive number used to avoid the situation where the denominator is zero, that is, when t s =t c When the object is closer to the current time, its time priority is higher;

[0068] Step 2: The current pending task p i To start matching participants, we first use a time-aware participant screening method to screen out participants that meet the time constraints in the current participant sliding window to form a candidate participant set C. Then we determine the number of participants in C |C| and p i The relationship between the required number of participants k, and perform corresponding operations according to different relationships;

[0069] Specifically, the main steps of the time-aware participant screening method mentioned are as follows:

[0070] (1) Initialize index index and traverse the current participant sliding window SW w , get the current participant's working time w i .(st,et);

[0071] (2) If participant w i 's working hours w i .(st, et) contains the request execution time p i of task p i .(t start , t end ), that is then add w i to the participant candidate set C, where w i .st represents the start time of the participant's work, w i .et represents the end time of the participant's work, p i .t start represents the time when the task needs to start execution, p i .t end represents the end time of task execution;

[0072] (3) Compare the number of participants |C| in the candidate set C with the number of participants k required for the task:

[0073] a. If |C| = k, the number of participants in the selected set C is exactly equal to the number of participants k required for the current task, and directly return C as the matching result;

[0074] b. If |C| < k, the number of participants that meet the time constraints is less than the number of participants required for the current task, then the task matching fails;

[0075] c. If |C| > k, call the k - space composite Skyline query algorithm to further screen high - quality participants, and calculate the time fitness η(w i , p i ), where w i .t = w i .(st, et), p i .t = p i .(t start , t end ), η(w i , p i ) evaluates the time fitness between the participant object w i and the task p i . The larger this value is, the higher the matching degree between the participant object w i and the task p i , and at the same time update the time fitness dimension in the participant attribute table;

[0076] Step 3: When the number of candidates calculated in step 2 |C|>k, use the proposed k-space composite skyline query algorithm to select k participants who perform well in all dimensions for the current task;

[0077] Specifically, the main steps of the k-space composite skyline query algorithm are as follows:

[0078] (1) Calculate the participant w i With task p i The distance d(w i ,p i ), update the participant attribute table;

[0079] (2) Initialize the index variable index to traverse the participants in the candidate set C and obtain the current participant w i The attribute values ​​of each dimension;

[0080] (3) Calculate w i In all two-dimensional subspaces U xy The dominance of U xy It is used to represent the distribution and comparison of participants on two specific dimensions, where x and y represent two different attributes, such as quote and distance. If w i in U xy is a skyline point, then U xy w i The dominance space A xy , A xy Add the dimension pair (x,y) of the participant w i The dimension of the set R i middle;

[0081] (4) After the above traversal calculation is completed, the index variable index is initialized again to traverse the participants in C and obtain the participant w i The dimension of the set R i ;

[0082] (5) If w i The dominance space A xy (w i ) covers the data space D, that is, w i Add to query set CS;

[0083] For example, suppose the data set P contains 3 points {p 1 , p 2 , p 3}, the data space D consists of three dimensions {A, B, C}, p 1 The space dominated by {A AB , ABC}, p 2 's domination space is {A AB , A AC}, p 3 's domination space is {A BC}, because the union of the domination spaces of p 1 covers the entire dimensional space D, and the union of the domination spaces of p covers the entire dimensional space D, so CS = {p 2 , p }; 1 , p 2};

[0084] (6) Compare the number of participants |CS| in CS with the number of participants k required by the task. If |CS| = k, directly return CS as the matching result to the task requester; if |CS| > k, call the evaluation function based on the domination area to calculate the evaluation function values of the participants, and select the k participants with the largest evaluation function values to form the screening set CS a as the matching result; if |CS| < k, call the alternative algorithm to calculate k - |CS| alternative participants to make up for the lack of the number of participants, and jointly form the matching result with the participants in CS;

[0085] Specifically: The evaluation function algorithm based on the domination area mentioned in the k - space composite Skyline query algorithm includes the following specific steps:

[0086] (1) Obtain the set of dimension pairs R i of each participant w i , and query its domination space A xy (w i );

[0087] (2) Calculate the domination area S i on each domination space A xy of the participant w xy (w i ) in size. Specifically, among them, the domination area S xy (w i ) is divided into the following three cases:

[0088] When the participant w i is in the two - dimensional subspace U xy , if the values of both attribute dimensions are better when they are smaller, then the size of the domination area of w i is recorded as S xy (w i ) = (1 - p i [r])(1 - p i [m])(r, m ∈ [1, d]);

[0089] When participants i In the two-dimensional subspace U xy If the values ​​of both attribute dimensions are as large as possible, then w i The size of the dominant area is denoted as S xy (w i )=p i [r]p i [m](r,m∈[1,d]);

[0090] When participants i In the two-dimensional subspace U xy If the value of one attribute dimension is as small as possible and the value of another attribute dimension is as large as possible, then w i The size of the dominant area is denoted as S xy (w i )=(1-p i [r])(p i [m])(r,m∈[1,d]);

[0091] (3) Calculate the sum of the areas of each participant's controlled space

[0092] (4) Participants are sorted according to the evaluation function value, and the top k participants form the CS a ;

[0093] (5) Return to CS a Give k-space composite skyline query algorithm.

[0094] Specifically, the candidate participant query algorithm mentioned in the k-space composite skyline query algorithm has the following specific steps:

[0095] (1) Get the participant w i The dimension of the set R i , calculate w i Dominate Space A xy (w i )

[0096] (2) Arrange the participants in C according to their control space A xy (w i ) Sort the numbers in descending order;

[0097] (3) When w i With w j When the number of dominated spaces is the same, the evaluation function algorithm based on the dominated area is called to calculate w i With w j The size of the dominant area is w i With w j Sort again;

[0098] (4) Add the top k-|CS| participants to CS b middle;

[0099] (5) Return to CS b Give k-space composite skyline query algorithm.

[0100] Step 4: In order to more flexibly handle dynamically arriving tasks and participant objects, the task matching window model maintenance strategy is called during the matching process to maintain the two sliding windows;

[0101] Specifically, the specific steps of the task matching window model maintenance strategy are as follows:

[0102] Task sliding window maintenance strategy:

[0103] (1) If task p i With k participants w i Matching success or task p i Expired, the p i Remove from the task sliding window;

[0104] (2) If the number of tasks in the current task sliding window is less than the size of the task sliding window N p , add the tasks that have newly arrived in the task window to the task sliding window according to their time priority;

[0105] Participant sliding window window maintenance strategy:

[0106] (1) If participant w i With p i Match successful or w i If the working time limit is exceeded, i Remove from the task sliding window;

[0107] (2) If the number of tasks in the current participant sliding window is less than the size of the participant sliding window N w , add the participants who newly arrive at the participant window to the participant sliding window according to time priority.

[0108] For example, in the dynamic task matching method based on Skyline query for crowd intelligence perception, for the data in the task matching window model at a certain time t, such as Figure 3 , Figure 4 Shown: Task sliding window SW p Task and participant sliding window SW w The participants in are arranged according to time priority. In this example, N p and N ware set to 5 and 15 respectively. Let each participant object w i There are four attribute dimensions, including its own spatial coordinate w i .loc, available time range w i .(st,et), quote w i .quo, credibility w i .μ. Each task object p i Contains its own spatial coordinates p i .loc, given execution time range p i .(t start ,t end ). The platform first checks the task sliding window SW p The first task in the waiting queue Q is neutralized and the task with the highest time priority is selected as the task to be processed. 1 Three participants are required and scheduled to be executed between 2:00 and 4:00. Through the time-aware participant screening method, a candidate participant set C = {w 1 ,w 2 ,w 3 ,w 4 ,w 5 ,w 6 ,w 7 ,w 8}. Since the number of participants in the participant candidate set C exceeds the required 3 (k=3), further calculation of participants w i With task p 1 The time fit η and distance d of the participant are used to update the attribute table of the participant. Then, a k-space composite skyline query is performed on the participant candidate set C. First, w is calculated. i Dominance in all two-dimensional subspaces, in this case w i There are four dimensions in total, which can be divided into six two-dimensional subspaces (for the convenience of calculation and understanding, the four dimensions of credibility, quotation, distance, and time fit in this example are represented by A, B, C, and D respectively). According to the k-space composite skyline query, the query set CS = {w 3 ,w 4 ,w 5 ,w 7}, because the number of participants in CS |CS| is greater than k = 3, the evaluation function is called to calculate the dominant area of ​​the participants and sort them according to the size of the dominant area, such as Figure 5 As shown. Finally, the return is task p 1 The top 3 participants who successfully matched {w 3 ,w 4 ,w 5}.

[0109] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation methods. They can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the figures shown and described herein.

Claims

1. A dynamic task matching method based on skyline query for crowd intelligence perception, characterized in that: It includes the following steps: Step 1: Use the task matching window model to handle dynamically arriving tasks and participants. The model uses a double sliding window and waiting queue mechanism, including the task sliding window SW p , Participant sliding window SW w and a waiting queue Q, where SW p , SW w , Q both use the data structure of priority queues and select the pending tasks p according to the priority scheduling algorithm. i And according to the time priority for SW p , Q and SW w Sort objects in ; The specific steps of the priority scheduling algorithm are as follows: (1) Calculate SW p , new tasks and SW in Q w Time Priority Index (TPI) of new participants; (2) According to the TPI value, SW p , Q and SW w Sort the objects in the queue, put the object with the highest TPI at the head of each queue, and return the task with the highest TPI as the pending task p i , output the sorted SW p , S.W. w ; Step 2: For the pending task p i Start matching participants. First, use the time-aware participant screening method to screen out participants that meet the time constraints in the current participant sliding window to form a candidate set C. The specific steps of the participant screening method based on time awareness are as follows: (1) Initialize index index and traverse the current participant sliding window SW w , get the current participant's working time w i .(st,et); (2) If participant w i Working hours i .(st,et) contains task p i The request execution time p i .(t start ,t end ),Right now Then w i Add to the participant candidate set C, where w i .st indicates the participant’s work start time, w i .et indicates the end time of the participant's work, p i .t start Indicates the time when the task needs to start executing, p i .t end Indicates the task execution end time; (3) Compare the number of participants |C| in the candidate set C with the number of participants k required for the task: a. If |C| = k, the number of participants in the selected set C is exactly equal to the number of participants k required for the current task, and directly return C as the matching result; b. If |C| < k, the number of participants that meet the time constraint cannot reach the number of participants required for the current task, and the task matching fails; c. If |C|>k, call the k-space composite skyline query algorithm to further screen high-quality participants and calculate the time fit between the participant and the task η(w i ,p i ), where w i .t=w i .(st,et),p i .t=p i .(t start ,t end ),η(w i ,p i ) evaluated the participants w i With task p i The greater the value of the temporal fit, the more likely the participant w i With task p i The higher the degree of match, the time fit dimension in the participant attribute table is updated simultaneously; Step 3: When the number of participants in the candidate set C is greater than k, call the k-space composite skyline query algorithm to further process C and filter out the participants that are relevant to the current task p. i Participants who perform well in each given attribute dimension; The specific steps of the k-space composite Skyline query algorithm are as follows: (1) Calculate the participant w i With task p i The distance d(w i ,p i ), update the participant attribute table; (2) Initialize the index variable index to traverse the participants in the candidate set C and obtain the current participant w i The attribute values ​​of all dimensions; (3) Calculate w i In each two-dimensional subspace U xy The dominance of U xy Used to represent the distribution and comparison of participants on two specific dimensions, where x and y represent two different attributes, such as quote and distance. If w i in U xy is a skyline point, then U xy w i The dominance space A xy , A xy Add the dimension pair (x,y) of the participant w i The dimension of the set R i middle; (4) Initialize the index variable index again, traverse the participants in C, and obtain participant w i The dimension of the set R i ; (5) If w i The dominance space A xy (w i ) covers the data space D, that is, w i Add to query set CS; (6) Compare the number of participants in CS |CS| with the number of participants k required for the task. If |CS|=k, CS is directly returned to the task requester as a matching result. If |CS|>k, the evaluation function based on the dominant area is called to calculate the evaluation function value of the participant, and the k participants with the largest evaluation function value are selected to form the screening set CS. a As a result of the match; If |CS| < k, call the alternative algorithm to calculate k - |CS| alternative participants to make up for the lack of participants, and jointly form the matching result with the participants in CS; The specific steps of the evaluation function algorithm based on the dominated area are as follows: (1) Get each participant w i The dimension of the set R i , query its dominance space A xy (w i ); (2) Calculate the participant w i Each control space A xy The dominant area S xy (w i ) size, where the dominant area S xy (w i ), which can be divided into the following three cases: When participants i In the two-dimensional subspace U xy If the values ​​of both attribute dimensions are as small as possible, then w i The size of the dominant area is denoted as S xy (w i )=(1-p i [r])(1-p i [m])(r,m∈[1,d]); When participants i In the two-dimensional subspace U xy If the values ​​of both attribute dimensions are as large as possible, then w i The size of the dominant area is denoted as S xy (w i )=p i [r]p i [m](r,m∈[1,d]); When participants i In the two-dimensional subspace U xy If the value of one attribute dimension is as small as possible and the value of another attribute dimension is as large as possible, then w i The size of the dominant area is denoted as S xy (w i )=(1-p i [r])(p i [m])(r,m∈[1,d]); (4) Calculate the sum of the areas of each participant's controlled space (5) Rearrange the participants in descending order of evaluation function values, and the top k participants form the CS a ; (6) Return to CS a Give k-space composite skyline query algorithm; The alternative participant query algorithm mentioned in the k-space composite Skyline query algorithm is as follows: (1) Get the participant w i The dimension of the set R i , calculate w i Dominate Space A xy (w i ) (2) Arrange the participants in C according to their control space A xy (w i ) Sort the quantities from most to least; (3) When w i With w j When the number of dominated spaces is the same, the evaluation function algorithm based on the dominated area is called to calculate w i With w j The size of the dominant area is w i With w j Sort by the dominant area value in descending order; (4) Add the top k-|CS| participants to CS b middle; (5) Return to CS b Give k-space composite skyline query algorithm; Step 4: During the matching process, call the task matching window model maintenance strategy to maintain the two sliding windows; The specific steps of the task matching window model maintenance strategy are as follows: Task sliding window maintenance strategy: (1) If task p i With k participants w i Matching success or task p i Expired, the p i Remove from the task sliding window; (2) If the number of tasks in the current task sliding window is less than the size of the task sliding window N p , add the tasks that have newly arrived in the task window to the task sliding window according to their time priority; Participant sliding window maintenance strategy: (1) If participant w i With p i Match successful or w i If the working time limit is exceeded, i Remove from the task sliding window; (2) If the number of tasks in the current participant sliding window is less than the size of the participant sliding window N w , add the participants who newly arrive at the participant window to the participant sliding window according to time priority.

2. According to the method for dynamic task matching based on skyline query for crowd intelligence perception according to claim 1, the time priority index TPI in step 1 (1) is characterized by: The time priority index where t s is the time when the current task starts to execute or the time when the participant starts working, t c refers to the current time, where is a small positive number used to avoid s =t c When the time denominator is zero, the closer the object is to the current time, the higher its time priority.

Images