CN115713213A - Crowd-sourcing task allocation method based on man-machine cooperation - Google Patents
Crowd-sourcing task allocation method based on man-machine cooperation Download PDFInfo
- Publication number
- CN115713213A CN115713213A CN202211480715.2A CN202211480715A CN115713213A CN 115713213 A CN115713213 A CN 115713213A CN 202211480715 A CN202211480715 A CN 202211480715A CN 115713213 A CN115713213 A CN 115713213A
- Authority
- CN
- China
- Prior art keywords
- task
- opportunistic
- cluster
- tasks
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention provides a crowd sourcing task allocation method based on man-machine cooperation, and belongs to the technical field of crowd sourcing perception. The method introduces a large number of mobile users, vehicles and mobile unmanned vehicles in a city to jointly complete tasks submitted by the users, predicts future mobile routes by analyzing historical routes of opportunistic users and opportunistic vehicles, and designs a bidirectional expected value matching algorithm for task distribution; the task bidding documents of the participatory users are auction by a reverse auction mechanism to carry out task allocation; and clustering the remaining unallocated tasks by a clustering method, allocating the unmanned vehicles to the clusters closest to the unmanned vehicles, and sequentially allocating the tasks closest to the unmanned vehicles. The invention fully utilizes various movable resources in the city, effectively overcomes the limitation of single body perception, improves the perception task completion rate, saves the perception task completion cost and can better adapt to various perception task requests in the city.
Description
Technical Field
The invention belongs to the technical field of crowd sensing, and relates to a crowd sensing task allocation method based on man-machine cooperation.
Background
Mobile crowd-sourcing awareness has become a new paradigm for smart city awareness. Different from the prior art that the sensor is fixed in a city for participating in environment perception, the mobile crowd sensing emphasizes mobility, and the wide and flexible environment perception is realized by taking a large number of mobile crowds in the city as carriers and a large number of mobile devices carried by the carriers. In the crowd sensing based on mobile users, the task completion modes of the users are flexible, and the users are mainly divided into two categories, namely opportunistic users for completing some tasks by forward movement and participatory users for completing some tasks by active movement. However, because the user applies and receives a perception task based on subjective willingness, although the crowd-sourcing platform implements some incentive mechanisms to encourage the user, according to practical experience, some popular, thick and low-cost perception tasks are always considered to be executed by the user in priority, and the rest of a large number of tasks cannot be completed on time, so the crowd-sourcing perception based on the mobile user has certain instability and uncontrollable property, and how to improve the task completion rate of the crowd-sourcing perception is an urgent problem to be solved.
With the development of technologies such as cloud computing, automatic driving, 5G communication and urban internet of vehicles, crowd sensing based on mobile vehicles gradually becomes a new platform for urban sensing. A large number of traditional vehicles and unmanned vehicles running in a city can realize the sensing along the running track through various sensors carried by the vehicles, and sensing data are transmitted to a crowd sensing platform by utilizing communication technologies such as 5G and cellular network after being preprocessed to realize city sensing. Compared with a mobile user, the traditional vehicle can execute some perception tasks along the road, the mobile unmanned vehicle can execute some cold perception tasks through instructions of the crowd perception platform, and can also replace the user to finish some relatively dangerous tasks, and the unmanned vehicle can move in the city for a long time in a long distance, so that lower-cost environmental perception is realized, and a large number of controllable mobile unmanned vehicles create favorable conditions for realizing wider perception in the city.
Inspired by the problems, the invention provides a crowd-sourcing task allocation method based on man-machine cooperation. Unlike previous mobile crowd sensing models, the participants in the sensing process are not limited to a single user or vehicle, but consist of a large number of mobile users, vehicles, and mobile unmanned vehicles distributed throughout a city. By simulating the real situation, the mobile user can finish some perception tasks along the way and also can apply for the perception tasks according to the subjective intention of the mobile user; vehicles moving in cities can also finish some perception tasks along the road; the rest sensing tasks which are not accepted by people or are not suitable for people can be finished by sending instructions to the movable unmanned vehicle by the crowd sensing platform, and the completion of the sensing tasks is guaranteed to the maximum extent through human-computer cooperation.
Disclosure of Invention
Aiming at the limitation of single body perception in the existing crowd sensing method, the invention provides a crowd task allocation method based on man-machine cooperation, which introduces a large number of mobile users, vehicles and movable unmanned vehicles in a city to jointly complete tasks submitted by the users, predicts future mobile routes by analyzing historical routes of opportunistic users and opportunistic vehicles, and designs a bidirectional expected value matching algorithm to allocate the tasks; the task bidding documents of the participatory users are auctioned through a reverse auction mechanism to carry out task allocation; and clustering the remaining unallocated tasks by a clustering method, allocating the unmanned vehicles to the clusters closest to the unmanned vehicles, and sequentially allocating the tasks closest to the unmanned vehicles.
The technical scheme of the invention is as follows:
a crowd-sourcing task allocation method based on human-computer cooperation is characterized in that task allocation work totally comprises three stages: in the first stage, the crowd sensing platform predicts the possibility of future moving destinations by analyzing historical routes of the opportunistic users and the opportunistic vehicles for task matching; in the second stage, the crowd sensing platform issues a task notification, the participating users actively apply for tasks through task bidding documents, and the crowd sensing platform carries out auction on the bidding documents to complete task allocation of the participating users; and in the third stage, the crowd sensing platform clusters the remaining unallocated tasks, allocates the unmanned vehicles to the closest clusters, and sequentially allocates the tasks to the unmanned vehicles closest to the unmanned vehicles. The method comprises the following specific steps:
step 1: the crowd sensing platform analyzes the task request of the customer and extracts specific information such as a task set T, a task place, task time, a task type and task budget.
Step 2: the crowd-sourcing aware platform performs task allocation to a set of opportunistic users OPP and a set of opportunistic vehicles OPV in a first phase. And (3) assuming that the historical routes of the opportunistic users meet Poisson distribution, the opportunistic vehicles have GPS historical records, analyzing past running places of the opportunistic users and the opportunistic vehicles to calculate the possibility of the future occurrence of the task positions, and performing task distribution through a designed bidirectional expected value matching algorithm. The specific process is as follows:
step 2.1: suppose an opportunistic user passes a certain task t in one day i Position loc of i And (3) according with Poisson distribution:
where s is the number of passes, λ i For task location loc i The coefficient of (a).
Assessing opportunistic user transit task location loc within a day i Has a probability of
Counting the number of each opportunistic user opp in the task t i Past week before start at task location loc i Average probability of occurrence to predict the future occurrence of an opportunistic user at task location loc i Possibility of (2) is denoted as E opp (loc i )。
Step 2.2: through the history GPS data of the unmanned vehicles, the history driving records of each opportunistic vehicle opv in one week are counted, and the opportunistic vehicles are predicted to appear at a certain task position loc i The probability of (c) is calculated as follows:
where N represents the number of tasks, count (loc) i ) Indicating the presence of an opportunistic vehicle at the mission location loc i The number of times.
Step 2.3: each task t i All have a desired value
Will appear at the task location loc i Possibility of (2)
By an opportunistic user or
The opportunistic vehicle is considered to be capable of successfully completing the task, if the maximum number of tasks matched by the opportunistic user and the opportunistic vehicle is q, the opportunistic user set OPP and the opportunistic vehicle set OPV are sequentially traversed, and the task is respectively completed according to E opp (loc i ) And E opv (loc i ) From big to small to task t i And marking the tasks after sorting.
Step 2.4: during traversal, if an already flagged task occurs, a backtrack is made to see if an unmarked task exists for the previously traversed opportunistic user or the opportunistic vehicle. If the unmarked task is successfully found in the backtracking, the previously traversed opportunistic user or the opportunistic vehicle marks the unmarked task, and cancels the marking of the previously traversed opportunistic user or the opportunistic vehicle to the current task, and simultaneously marks the current task by the currently traversed opportunistic user or the opportunistic vehicle.
Step 2.5: if the backtracking does not find a task that has not been flagged, the currently traversed opportunistic user or the opportunistic vehicle skips the flagging of the current task and selects the remaining tasks that have not been flagged for flagging.
Step 2.6: and repeating the step 2.3 to the step 2.5 until the opportunistic user and the opportunistic vehicle are traversed to the end.
Step 2.7: the crowd sensing platform assigns tagged tasks to the opportunistic users and the opportunistic vehicles and waits for the opportunistic users and the opportunistic vehicles to upload task results.
And step 3: and the crowd sensing platform issues the residual unassigned task set RT in the first stage, and the participating user set PAT in the city actively uploads the target task set and the quotation price application of the bidding document to the crowd sensing platform to execute the task. After receiving the application of the participating users, the crowd sensing platform filters the application of the quotation exceeding the task budget top, and a reverse auction mechanism is adopted to match the tasks with the participants. Meanwhile, a certain credit is issued to the unsuccessful participating users participating in the auction as cost compensation for the next auction to stimulate the participating users. The specific process is as follows:
step 3.1: filtering participating users pat h Task bidding quotation
Set of excess tasks (pat) h ) Budget top (set (pat) h ) ) a label.
Step 3.2: computing a participatory user pat h Average quote for submitted tasks:
step 3.3: push button
Is ranked from low to high, and the benchmarks are evaluated according to the following utility function:
wherein RT is a task set which is remained and not distributed after the first stage of the assignment of the opportunistic user and the opportunistic vehicle tasks, and CT is a candidate task set of the selected bidding document.
Step 3.4: selecting a utility function Unity (pat) h )>And a label of 0, discarding the label of the repeated task set.
Step 3.5: and continuously iterating, and repeating the step 3.3 and the step 3.4 until all the bidding documents submitted by the participatory users are traversed.
Step 3.6: the participatory users upload the initial positions, the crowd sensing platform calculates the mobile cost of the participatory users, and after the participatory users return the sensing result, the crowd sensing platform issues rewards to the participatory users:
wherein, gamma is 1 And gamma 2 Respectively representing the price coefficients; dis (pat) h ,set(pat h ) Represents a participating user pat h Task set (pat) for completing application h ) The moving distance of (c);
representing participating users pat h The points obtained by auction failure are accumulated after each auction failure and are reset after being used.
And 4, step 4: and the crowd sensing platform distributes the rest tasks to the special unmanned vehicle in the third stage. And the crowd-sourcing perception platform clusters the residual unassigned task sets RTF in the second stage according to the geographic coordinates of the tasks and the types of the tasks, and after the task clustering is completed, the unmanned vehicles respectively select the clusters closest to the current distance and then sequentially search the clusters closest to the tasks to complete the clustering. The method comprises the following specific steps:
step 4.1: selecting the number K of the unmanned vehicles as the cluster number after clustering, taking the position of the unmanned vehicles as the initial cluster center, and recording as the cluster center
Step 4.2: calculating perception task t in residual task set RTF of second stage w Distance to cluster center and assigning perceptual tasks to distances from cluster centerCluster center nearest to the cluster center.
Step 4.3: recalculating the mean value of the task points in each cluster, taking the mean value as a new cluster center, and recording as the new cluster center
Where l =1,2, \ 8230, is the number of iterations.
Step 4.4: and taking the sum of the square error of each perception task distance and the center of the cluster as a loss function, and sequentially calculating:
wherein, c w As task t w The cluster to which the cluster belongs to is,
represents a cluster c w M is the number of tasks in the residual task set RTF.
Step 4.5: repeating the steps 4.2 to 4.4 until the loss function is converged, outputting stable cluster centers, and recording as
Step 4.6: sequentially selecting the current unmanned car v Closest cluster:
wherein, the first and the second end of the pipe are connected with each other,
representative participation type unmanned vehicle car v And the cluster center output in step 4.5
J =1,2, \8230, K.
Step 4.7: and sequentially and respectively selecting the cluster centers closest to the current distance by the unmanned vehicles, distributing the unmanned vehicles to the clusters, and sequentially selecting the tasks closest to the cluster centers to complete.
The invention has the beneficial effects that: the crowd sourcing task allocation method based on the man-machine cooperation fully utilizes various movable resources in the city, effectively overcomes the limitation of single body perception, improves the perception task completion rate, saves the perception task completion cost, and can better adapt to various perception task requests in the city.
Drawings
FIG. 1 is a hierarchical structure diagram based on human-machine cooperation provided by the present invention.
FIG. 2 is a diagram of a task assignment scenario based on human-machine collaboration provided by the present invention.
FIG. 3 is a flow chart of a task matching algorithm provided by the present invention.
Fig. 4 is a diagram of a specific case of the task matching algorithm provided by the present invention.
FIG. 5 is a flow chart of a user bid selection algorithm provided by the present invention.
FIG. 6 is a flowchart of a task clustering algorithm provided by the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
The technical scheme of the invention is a crowd sourcing task allocation method based on man-machine cooperation. After the crowd sourcing platform analyzes the task request uploaded by the customer, the following task allocation method shown in fig. 1 mainly comprises four task allocation sub-methods of opportunistic user task allocation, opportunistic vehicle task allocation, participatory user task allocation and dedicated unmanned vehicle task allocation.
Fig. 2 shows a task allocation scenario, in which after receiving a task request uploaded by a customer, the crowd sensing platform performs task allocation in three stages, and finally returns a sensing result to the customer. The task allocation work mainly comprises three stages, wherein the first stage analyzes historical routes of an opportunistic user and an opportunistic vehicle to predict the possibility of moving a destination in the future for task matching; the crowd sensing platform issues a task notification in the second stage, the participating users actively apply for tasks through the bidding documents, and the crowd sensing platform carries out auction on the bidding documents to complete task allocation of the participating users; and clustering the remaining unallocated tasks by the crowd sensing platform in the third stage, allocating the unmanned vehicles to the closest clusters, and sequentially allocating the tasks to the unmanned vehicles closest to the cluster. The method comprises the following specific steps:
step 1: the crowd sensing platform analyzes the task request submitted by the customer, extracts a task set T, and each task T i Location with task loc i And task budget top i And so on. Then the task allocation work is carried out in the following three stages in sequence, and the results are returned one by one after the task is finished.
Step 2: the crowd-sourcing aware platform performs task allocation work on the set of opportunistic users OPP and the set of opportunistic vehicles OPV in a first phase. Assuming that historical routes of opportunistic users meet Poisson distribution, and an opportunistic vehicle has a GPS historical driving record, analyzing past driving places of the opportunistic users and the opportunistic vehicle to calculate the possibility that different opportunistic users appear at task positions, and performing task distribution through a designed bidirectional expected value matching algorithm, wherein the matching process is shown in FIG. 3; the method comprises the following specific steps:
step 2.1: suppose an opportunistic user passes a certain task t in one day i Position loc of i The Poisson distribution is satisfied:
wherein s is the number of passes, λ i For task location loc i The coefficient of (c).
Assessing opportunistic user traversal in one day by task location loc i Has a probability of
Counting the number of each opportunistic participant opp at the task t i Past week before start at task location loc i Average probability of occurrence to predict the future occurrence of an opportunistic user at task location loc i Possibility of (2) is denoted as E opp (loc i )。
Step 2.2: through the history GPS data of the unmanned vehicles, the history driving records of each opportunistic vehicle opv in one week are counted, and the opportunistic vehicles are predicted to appear at a certain task position loc i Is calculated as follows:
where N represents the number of tasks, count (loc) i ) Indicating the presence of an opportunistic vehicle at a mission location loc i The number of times.
Step 2.3: each task t i All have a desired value
Will appear at the task location loc i Possibility of (2)
By an opportunistic user or
The vehicle is deemed to be able to successfully complete the task. First filtering out non-conforming task expectations
Assuming that the maximum number of tasks matched by the opportunistic users and the opportunistic vehicles is q, sequentially traversing an opportunistic user set OPP and an opportunistic vehicle set OPV respectively according to E opp (loc i ) And E opv (loc i ) From big to small to task t i And marking the tasks after sorting.
Step 2.4: during traversal, if an already flagged task occurs, a backtrack is made to see if an unmarked task exists for the previously traversed opportunistic user or the opportunistic vehicle. If the unmarked task is successfully found in the backtracking, the previously traversed opportunistic user or the opportunistic vehicle marks the unmarked task, cancels the marking of the current task, and simultaneously marks the current task by the currently traversed opportunistic user or the opportunistic vehicle.
Step 2.5: if the backtracking does not find a task that has not yet been flagged, the currently traversed opportunistic user or opportunistic vehicle skips the flagging of that task while selecting the remaining tasks that have not yet been flagged for flagging.
Step 2.6: repeating steps 2.3 through 2.5 until the end of the opportunistic user and opportunistic vehicle traversal.
Step 2.7: the crowd sensing platform assigns tagged tasks to the opportunistic users and the opportunistic vehicles and waits for the opportunistic users and the opportunistic vehicles to upload task results.
In order to better understand the task matching algorithm provided in the first stage, a specific example is provided below for explanation, as shown in fig. 4, assuming that the task has been completed and the maximum matching number q of the task is 2. According to step 2.3 of the above algorithm, we first traverse the opportunistic user opp 1 To task { t } 1 ,t 2 Mark and then traverse the user opp 2 In-process discovery task t 2 Has been opportunistic user opp 1 Marking, backtracking according to the step 2.4, and finding the task t 3 Has not been yet opportunistic user opp 1 Tagging, so opportunistic user opp 1 Delete Pair task t 2 Is added to the task t 3 Of the opportunistic user opp 1 The marked task sequence is t 1 ,t 3 }, opportunistic user opp 2 Increase to task t 2 Is marked, then the user opp is defined opportunistically 2 The marked task sequence is t 2 }. Subsequent traversals according to step 2.6, opportunisticHousehold opp 3 For task t 4 And (5) marking. On traversing opportunistic vehicle opv 1 Then, the discovery task t 4 Has been opportunistic user opp 3 Mark, and find user opp in backtracking 3 There are no unmarked tasks already, so the opportunistic vehicle opv 1 No task can be flagged. Last opportunistic vehicle opv 2 For task t 5 And marking, and when the traversal is finished, the crowd sensing platform distributes tasks to the opportunistic users and the vehicles and waits for the subsequent return of task results.
And step 3: and (3) the crowd sensing platform issues the residual unallocated task set RT in the first stage, and the participating user set PAT in the city actively uploads a target task set and a bid price application execution task of the bidding document to the platform. As shown in FIG. 5, after receiving the application of participating users, the crowd sourcing platform filters out applications with bids exceeding the task budget top, and a reverse auction mechanism is used for matching tasks and participants. And simultaneously, a certain credit is issued to the participatory users who participate in the auction but are not successful as cost compensation in the next auction to stimulate the participatory users.
Step 3.1: firstly, filtering out participating users pat by the crowd sensing platform h Task bidding price of path Set of excess tasks (pat) h ) Budget top (set (pat) h ) ) a label.
Step 3.2: computing a participatory user pat h Average quote for submitted tasks:
step 3.3: push button
Is ranked from low to high, and the benchmarks are evaluated according to the following utility function:
wherein RT is a task set which is remained and not distributed after the first stage of the assignment of the opportunistic user and the opportunistic vehicle tasks, and CT is a candidate task set of the selected bidding document.
Step 3.4: selecting a utility function Unity (pat) i )>The bid of 0, otherwise, is considered as auction failure. And discarding the tasks repeated with the candidate task set CT, assigning the rest tasks to the participatory users, and updating the rest task set RT and the candidate task set CT. And issues a certain credit card to the users who failed in the auction as compensation.
Step 3.5: and continuously iterating, and repeating the steps 3.3 and 3.4 until all the bidding documents submitted by the participatory users are traversed.
Step 3.6: uploading the initial position by the participatory user, calculating the mobile cost of the participatory user by the crowd sensing platform, and after the participatory user returns a task result, providing rewards to the participatory user by the crowd sensing platform:
wherein, γ 1 And gamma 2 Respectively representing the price coefficients; dis (pat) h ,set(pat h ) Represents a participating user pat h Task set (pat) for completing application h ) The moving distance of (a);
representing participating users pat h The integral obtained when the auction fails can be accumulated after each auction failure and can be reset after use.
And 4, step 4: and the crowd sensing platform distributes the rest tasks to the special unmanned vehicle in the third stage. As shown in fig. 6, the crowd sensing platform firstly clusters the residual task sets RTF in the second stage according to the geographic coordinates of the tasks and the types of the tasks, and after the task clustering is completed, the unmanned vehicles respectively select the clusters closest to the current distance, and then sequentially search the clusters for the tasks closest to the current distance to complete the task clustering.
Step 4.1: firstly, selecting the number K of the unmanned vehicles as the number of clusters after clustering, and taking the positions of the unmanned vehicles as initial cluster centers and recording the initial cluster centers as the initial cluster centers
Step 4.2: calculating perception task t in residual unallocated task set RTF in second stage w Distance to the cluster center, and assigning the perception task to the cluster to which the cluster center closest to the cluster center belongs.
Step 4.3: recalculating the mean value of the task points in each cluster, taking the mean value as a new cluster center, and recording as the new cluster center
Where l =1,2, \ 8230, is the number of iterations.
Step 4.4: each perception task t w And the sum of the square errors of the distance and the center of the cluster is used as a loss function, and the following calculation is sequentially carried out:
wherein, c w As task t w The cluster to which the cluster belongs to is,
represents a cluster c w M is the number of tasks in the task set RTF.
Step 4.5: repeating the steps 4.2 to 4.4 until the loss function is converged, and outputting a stable cluster center which is recorded as
Step 4.6: sequentially selecting the current unmanned car v Closest cluster:
wherein the content of the first and second substances,
representative participation type unmanned vehicle car v Cluster center with the above output
Manhattan distance of.
Step 4.7: the unmanned vehicles respectively select the cluster centers closest to the current position, the unmanned vehicles are distributed into the clusters, the unmanned vehicles sequence the tasks in the clusters from near to far according to the distance between the unmanned vehicles and the tasks, and the tasks closest to the current position are sequentially selected from the clusters to complete.
In conclusion, the crowd-sourcing task allocation method based on man-machine cooperation provided by the invention fully utilizes various movable resources in the city, effectively overcomes the limitation of single body perception, improves the perception task completion rate, saves the task completion cost, and can better adapt to various perception task requests in the city.
Claims (5)
1. A crowd-sourcing task allocation method based on man-machine cooperation is characterized by comprising the following steps:
step 1: the crowd sensing platform analyzes a task request of a customer, and extracts a task set T, a task place, task time, a task type and a task budget;
step 2: the crowd sensing platform carries out task allocation on an opportunistic user set OPP and an opportunistic vehicle set OPV in a first stage; setting an opportunistic user historical route to meet Poisson distribution, wherein an opportunistic vehicle has a GPS historical record, analyzing past running places of the opportunistic user and the opportunistic vehicle to calculate the possibility of the future occurrence of a task position, and performing task distribution through a designed bidirectional expected value matching algorithm;
and 3, step 3: the crowd sensing platform issues a task set RT which is left unallocated in the first stage, and a participating user set PAT in a city actively uploads a target task set and a quotation price application execution task of a bidding document to the crowd sensing platform; after receiving the application of the participating user, the crowd sensing platform filters the application that the quotation exceeds the task budget top, and a reverse auction mechanism is adopted to match the task with the participants; meanwhile, points are issued to the participatory users who participate in the auction but are not successful as cost compensation in the next auction to stimulate the participatory users;
and 4, step 4: and clustering the residual unallocated task sets RTF in the second stage by the crowd sensing platform according to the geographic coordinates and the types of the tasks, respectively selecting the clusters closest to the current distance by the unmanned vehicle after the task clustering is finished, and then sequentially searching the tasks closest to the current distance in the clusters to finish the task clustering.
2. The crowd-sourcing task allocation method based on human-computer collaboration as claimed in claim 1, wherein the specific process of step 2 is as follows:
step 2.1: an opportunistic user passes a certain task t in one day i Position loc of i And (3) according with Poisson distribution:
where s is the number of passes, λ i For task location loc i The coefficients of (c);
assessing opportunistic user traversal in one day by task location loc i Has a probability of
Counting the number of each opportunistic user opp in the task t i Past week before start at task location loc i Average probability of occurrence to predict the future occurrence of an opportunistic user at task location loc i Possibility of (1), denoted as E opp (loc i )；
Step 2.2: through the historical GPS data of the unmanned vehicles, the historical driving of each opportunistic vehicle opv within one week is countedRecording and predicting the opportunistic occurrence of a vehicle at a certain mission location loc i The probability of (c) is calculated as follows:
where N represents the number of tasks, count (loc) i ) Indicating the presence of an opportunistic vehicle at a mission location loc i The number of times of (c);
step 2.3: each task t i All have a desired value
Will appear at the task location loc i Possibility of (2)
By an opportunistic user or
The opportunistic vehicle is regarded as successfully completing the task, the maximum number of the tasks matched by the opportunistic user and the opportunistic vehicle is set to be q, the opportunistic user set OPP and the opportunistic vehicle set OPV are sequentially traversed, and the task is respectively completed according to E opp (loc i ) And E opv (loc i ) From big to small to task t i After sequencing, marking the tasks;
step 2.4: in the traversing process, if the tasks which are already marked appear, backtracking is carried out to see whether the previously traversed opportunistic users or opportunistic vehicles have unmarked tasks; if the unmarked task is successfully found in backtracking, the previously traversed opportunistic user or the opportunistic vehicle marks the unmarked task, cancels the marking of the current task, and simultaneously marks the current task by the currently traversed opportunistic user or the opportunistic vehicle;
step 2.5: if the backtracking does not find the tasks which are not marked yet, the currently traversed opportunistic user or the opportunistic vehicle skips the marking of the current task, and selects the remaining tasks which are not marked yet to mark;
step 2.6: repeating the step 2.3 to the step 2.5 until the opportunistic user and the opportunistic vehicle are traversed;
step 2.7: the crowd sensing platform distributes marked tasks to the opportunistic users and the opportunistic vehicles and waits for the opportunistic users and the opportunistic vehicles to upload task results.
3. A crowd-sourcing task allocation method based on human-machine cooperation according to claim 1 or 2, wherein the specific process of step 3 is as follows:
step 3.1: filtering participating users pat h Task ticket quotation price path Set of excess tasks (pat) h ) Budget top (set (pat) h ) A banner of);
step 3.2: computing a participatory user pat h Average quote of submitted tasks:
step 3.3: push button
Is ranked from low to high, the bid is evaluated according to the following utility function:
wherein RT is a residual unallocated task set after the first stage of the allocation of the opportunistic users and the opportunistic vehicle tasks, and CT is a candidate task set of the selected bidding documents;
step 3.4: selecting a utility function Unity9pat h )>A bidding document of 0, discarding the bidding document of the repeated task set;
step 3.5: continuously iterating, and repeating the step 3.3 and the step 3.4 until all the bidding documents submitted by the participatory users are traversed;
step 3.6: uploading the initial position by the participatory user, calculating the mobile cost of the participatory user by the crowd sensing platform, and after the participatory user returns a sensing result, the crowd sensing platform issues a reward to the participatory user:
wherein, γ 1 And gamma 2 Respectively representing the price coefficients; dis (pat) h ,set(pat h ) Represents a participating user pat h Task set (pat) for completing application h ) The moving distance of (a);
representative participating user pat h The points obtained by auction failure are accumulated after each auction failure and are reset after being used.
4. A crowd-sourcing task allocation method based on human-machine cooperation according to claim 1 or 2, wherein the specific process of step 4 is as follows:
step 4.1: selecting the number K of the unmanned vehicles as the cluster number after clustering, taking the position of the unmanned vehicles as the initial cluster center, and recording as the cluster center
And 4.2: calculating perception task t in residual task set RTF of second stage w Distance to the cluster center, and allocating the perception task to the cluster to which the cluster center closest to the sensing task belongs;
step 4.3: recalculating the mean value of the task points in each cluster, taking the mean value as a new cluster center, and recording as the new cluster center
Wherein l =1,2, \ 8230, is the number of iterations;
step 4.4: and taking the sum of the square error of each perception task distance and the center of the cluster as a loss function, and sequentially calculating:
wherein, c w As task t w The cluster to which the cluster belongs to is,
represents a cluster c w M is the number of tasks in the residual task set RTF;
step 4.5: repeating the steps 4.2 to 4.4 until the loss function is converged, outputting stable cluster centers, and recording as
Step 4.6: sequentially selecting the current unmanned car v Closest cluster:
wherein the content of the first and second substances,
representative participation type unmanned vehicle car v Cluster center with step 4.5 output
J =1,2, \ 8230, K;
step 4.7: and the unmanned vehicles sequentially and respectively select the cluster centers closest to the current distance, distribute the unmanned vehicles to the clusters, and sequentially select the tasks closest to the cluster in order to complete the tasks.
5. A crowd-sourcing task allocation method based on human-machine cooperation according to claim 3, wherein the specific process of step 4 is as follows:
step 4.1: selecting the number K of the unmanned vehicles as the cluster number after clustering, taking the position of the unmanned vehicles as the initial cluster center, and recording as the cluster center
Step 4.2: calculating perception task t in residual task set RTF of second stage w Distance from the cluster center, and allocating the sensing task to the cluster to which the cluster center closest to the sensing task belongs;
step 4.3: recalculating the mean value of the task points in each cluster, taking the mean value as a new cluster center, and recording as the new cluster center
Wherein l =1,2, \ 8230, is the number of iterations;
step 4.4: and taking the sum of the square error of each perception task distance and the center of the cluster as a loss function, and sequentially calculating:
wherein, c w As task t w The cluster to which the cluster belongs to is,
represents a cluster c w M is the number of tasks in the residual task set RTF;
step 4.5: repeating the steps 4.2 to 4.4 until the loss function is converged, and outputting a stable cluster center which is recorded as
Step 4.6: sequentially selecting the current unmanned car v Closest cluster:
wherein the content of the first and second substances,
representative participation type unmanned vehicle car v Cluster center with step 4.5 output
J =1,2, \ 8230;, K;
step 4.7: and the unmanned vehicles sequentially and respectively select the cluster centers closest to the current distance, distribute the unmanned vehicles to the clusters, and sequentially select the tasks closest to the cluster in order to complete the tasks.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211480715.2A CN115713213A (en) | 2022-11-24 | 2022-11-24 | Crowd-sourcing task allocation method based on man-machine cooperation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211480715.2A CN115713213A (en) | 2022-11-24 | 2022-11-24 | Crowd-sourcing task allocation method based on man-machine cooperation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115713213A true CN115713213A (en) | 2023-02-24 |
Family
ID=85234767
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211480715.2A Pending CN115713213A (en) | 2022-11-24 | 2022-11-24 | Crowd-sourcing task allocation method based on man-machine cooperation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115713213A (en) |
-
2022
- 2022-11-24 CN CN202211480715.2A patent/CN115713213A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Alemi et al. | What influences travelers to use Uber? Exploring the factors affecting the adoption of on-demand ride services in California | |
| JP6925479B2 (en) | Systems and methods for assigning sharable orders | |
| CN110645983B (en) | Path planning method, device and system for unmanned vehicle | |
| Asghari et al. | Price-aware real-time ride-sharing at scale: an auction-based approach | |
| CN106875066B (en) | Vehicle travel behavior prediction method, device, server and storage medium | |
| Asghari et al. | An on-line truthful and individually rational pricing mechanism for ride-sharing | |
| US20180012153A1 (en) | Order allocation system and method | |
| CN109325671B (en) | Space-time crowdsourcing online task allocation method and system | |
| CN112074845A (en) | Deep reinforcement learning for optimizing car pooling strategies | |
| WO2018081634A1 (en) | Price-aware real-time auction-based ride-sharing system | |
| Friedrich et al. | A modeling approach for matching ridesharing trips within macroscopic travel demand models | |
| US20180314998A1 (en) | Resource Allocation in a Network System | |
| CN110390415A (en) | A kind of method and system carrying out trip mode recommendation based on user's trip big data | |
| US20210117874A1 (en) | System for dispatching a driver | |
| CN112036696A (en) | Task allocation method, task allocation device, storage medium, and electronic apparatus | |
| CN109724587A (en) | Dynamic route planing method, device and medium based on GPS | |
| CN112232596B (en) | Multi-mode intermodal transit city optimization method, system, terminal and storage medium | |
| Luo et al. | Optimal design of intermodal mobility networks under uncertainty: Connecting micromobility with mobility-on-demand transit | |
| CN108288101B (en) | Auction mechanism-based vehicle resource allocation and pricing method for online taxi appointment reservation service | |
| US20230316395A1 (en) | Management methods and systems for affordable housing applications in smart city based on internet of things | |
| WO2018146622A1 (en) | Dynamic selection of geo-based service options in a network system | |
| Nguyen | Fair cost sharing auction mechanisms in last mile ridesharing | |
| CN114372830A (en) | Network taxi booking demand prediction method based on space-time multi-graph neural network | |
| CN110414737A (en) | Public transport stroke processing method, device, electronic equipment and storage medium | |
| Liu et al. | Heuristic approach for the multiobjective optimization of the customized bus scheduling problem |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |