CN108304266A - A kind of mobile multiple target intelligent perception method for allocating tasks - Google Patents

Abstract

The invention discloses a kind of mobile multiple target intelligent perception method for allocating tasks, include the following steps：S1, employer issue location-based query task to task distribution system；S2, task distribution system consider maximum awareness coverage and task completion rate to establish mobile multiple target perception task distribution model under the requirement of employer, solve best worker, and query task is distributed to worker；S3, worker execute query task, and the automatic sensing in the way for going to query task position, query task result and automatic sensing data are finally returned to task distribution system after receiving query task.The method is by considering two targets of automatic sensing task and location-based query task come algorithm for design, sensing capability and passive sensing capability so that worker takes the initiative simultaneously in task, the efficiency of perception task completion is improved, and further reduces the budget needed for employer.

Description

A mobile multi-target crowd sensing task assignment method

technical field

The invention relates to the field of crowd sensing, in particular to a mobile multi-target crowd sensing task assignment method.

Background technique

Crowdsensing is an emerging problem solution that collects data (such as sound, location, noise, GPS) from a large number of ordinary mobile phone users to complete sensing tasks. Using the collected perception data, researchers can realize a variety of large-scale perception applications, such as noise detection, parking space detection, environment detection, etc.

There are two main types of tasks in crowd sensing that have received extensive attention. The first category emphasizes the automatic perception of ordinary mobile phone users when collecting data. For example, in road traffic detection applications, mobile devices automatically sense data and record it for subsequent processing. Another class of tasks are location-based query tasks that require workers to actively respond.

Previous techniques tend to consider these two tasks separately, but there are actually some connections between them. When the user goes to complete some query tasks, he can also obtain the sensing data to complete the automatic sensing tasks on the way. In this way, if there is a dual-objective task assignment system for task combination, it will make full use of each worker's active perception ability and passive perception ability, thereby improving the completion efficiency of perception tasks.

Contents of the invention

The purpose of the present invention is to address the deficiencies in the prior art and provide a mobile multi-objective group sensing task assignment method. The method models a novel multi-objective group sensing task assignment system and designs a A greedy algorithm-based solution method improves the efficiency of perception task completion and further reduces the budget required by employers.

The purpose of the present invention can be achieved through the following technical solutions:

A mobile multi-objective crowd sensing task assignment method, said method comprising the following steps:

Step S1, the employer publishes a location-based query task to the task assignment system;

Step S2. The task allocation system establishes a mobile multi-objective perception task allocation model based on the employer's requirements, taking into account the maximum sensing coverage and task completion rate, solves and selects the best worker, and assigns the query task to the worker ;

Step S3. After receiving the query task, the worker who is assigned the query task executes the query task, and automatically perceives it on the way to the location of the query task, and finally returns the query task result and the automatic sensing data to the task assignment system.

Further, the mobile multi-target sensing task allocation model in step S2 includes a location-based query task target allocation model and an automatic sensing task target allocation model.

Further, the target allocation model of the location-based query task is aimed at maximizing the completion success rate of the query task, and the formula of the objective function max F(S) is as follows:

Among them, t _i represents the i-th query task, T represents the set of all query tasks, m represents the size of the query task set, W _i represents the set of all workers who accept the query task t _i , w _j represents the worker set W _i a worker, Indicates the completion success rate of the i-th query task, and p _j indicates the completion success rate of the j-th worker's historical task;

The automatic perception task target assignment model, its goal is to maximize the perception coverage, the objective function max G (S) formula is as follows:

in, represents the set of employed workers, Indicates the expected path of each worker w _j ∈ W _i ;

Considering the above two task target allocation models, and at the same time requiring that the total cost of constrained workers is not greater than the budget given by the employer, the mobile multi-target perception task allocation model is established as follows:

Among them, W represents the set of all workers, S represents the set of employed workers, c _j represents the cost of the hired workers to complete the task, and B represents the budget given by the employer.

According to the mobile multi-target perception task allocation model (MBC) built above, it can be proved that the model is an NP-hard problem, and the proof is as follows:

The submodular maximization problem under cardinality constraints (submodular maximization problem with cardinality constraints, SMCC) is an NP-hard problem. The problem is described as follows: Given a set U={u ₁ , u ₂ ,...,u _|U| } , a monotone submodular function f defined on U, and a base value K, the goal of this problem is to maximize f(U′), where And |U′|≤K, we prove the lemma by showing that the MBC problem is an instance of the SMCC problem. Assume that there is only one location-based sensing task, and all workers who can accept the task pass the target location of the task. Assuming that the reputation value and quotation of each worker are the same, the problem is equivalent to maximizing the union of the worker's perceived coverage under the constraint of task funding. which is We can prove that G(S) is a monotone submodular function, for any And w∈W\S ₂ , we have therefore:

This inequality is obtained according to the nature of set operations, we can find that S(C ₂ ∪{a})-S(C ₂ )≥0, which proves that S(C) is a monotone submodular function, so it can be explained that MBC The problem is an instance of the SMCC problem, so the MBC problem is an NP-hard problem, and the proof is completed. Further, the MBC-Greedy algorithm design process of the solution method of the model is as follows:

First, the completion probability of the query task has the following properties:

For each query task t _i , the task completion probability function is non-subtractive, as shown below:

Based on the above property, it is obtained that for the worker set S, the average task completion probability function F(S) is non-decreasing, and then for the worker set S, the perception coverage function G(S) is non-decreasing.

Further, solving the mobile multi-target perception task allocation model adopts a solving method based on a greedy algorithm, that is, the MBC-greedy algorithm, and the specific process is as follows:

Step 110: Initialize the set S of hired workers, set W _i of all workers who accept the query task t _i as an empty set, and the cost C required by the hired workers to complete the task is 0;

Step 120, combine all valid task-worker pairs (t _i , w _j ) and assign them to M;

Step 130, if M is not empty, then execute step 140, otherwise end the algorithm;

Step 140, eliminating the workers whose expenses required to complete the task are too high and will exceed the given budget of the employer;

Step 150. For each remaining valid task-worker pair (t _i , w _j ), calculate the weighted increment of query task completion probability and perception coverage respectively

in

Step 160. Eliminate inferior task-worker pairings. The pairing (t _i , w _j ) is better than (t′ _i , w′ _j ) or the pairing (t′ _i , w′ _j ) is worse than the pairing (t _i , w _j ), means at the same time or at the same time

Step 170, sorting the remaining task-worker pairings in descending order according to their superiority over other task-worker pairs;

Step 180. Select the top-ranked pair (t _i , w _j ) as the result of an iteration, and remove all pairs {(t _i , _{w j )} |t _i ∈ T before the next iteration }, and then return to step 130, and loop iteratively until the optimal number of workers selected by the solution meets the requirements of the employer.

Compared with the prior art, the present invention has the following advantages and beneficial effects:

Compared with the previous single-target task matching algorithm designed for a single type of perceptual task, a mobile multi-target crowd sensing task allocation method of the present invention comprehensively considers the two targets of automatic perceptual tasks and location-based query tasks to design the algorithm , so that workers can simultaneously exert active sensing ability and passive sensing ability in tasks, improve the efficiency of sensing task completion, and further reduce the budget required by employers.

Description of drawings

FIG. 1 is a flow chart of a mobile multi-target crowd sensing task assignment method according to an embodiment of the present invention.

Fig. 2 is a schematic diagram of the solutions of the MBC-greedy algorithm and the C-greedy and Q-greedy algorithms in the embodiment of the present invention when the budget and the total number of query tasks are fixed.

FIG. 3 is a comparison diagram of the MBC-greedy algorithm and the C-greedy and Q-greedy two comparison algorithms according to the embodiment of the present invention, as the number of query tasks increases, the perception coverage changes.

FIG. 4 is a comparison chart of the change in success rate of completion of query tasks between the MBC-greedy algorithm and the C-greedy and Q-greedy algorithms according to the embodiment of the present invention as the number of query tasks increases.

Detailed ways

The present invention will be further described in detail below in conjunction with the embodiments and the accompanying drawings, but the embodiments of the present invention are not limited thereto.

Example:

This embodiment provides a mobile multi-target crowd sensing task assignment method, the flow chart of the method is shown in Figure 1, including the following steps:

Step S1, the employer publishes a location-based query task to the task assignment system;

Step S2. The task allocation system establishes a mobile multi-objective perception task allocation model based on the employer's requirements, taking into account the maximum sensing coverage and task completion rate, solves and selects the best worker, and assigns the query task to the worker ;

Step S3. After receiving the query task, the worker who is assigned the query task executes the query task, and automatically perceives it on the way to the location of the query task, and finally returns the query task result and the automatic sensing data to the task assignment system.

Taking the noise monitoring task in various areas of the city and completing the specified location query task as an example, the employer releases the query task to the task distribution system, and the system queries the current location of the worker and the movement path in the future; the task distribution system judges whether the target location of the query task is On the path forward of workers, workers who can accept tasks are screened out. Then, based on the idea of the greedy algorithm, a certain number of workers are iteratively selected to complete the query task. In each iteration, the workers whose work costs are too high and will cause the task budget to exceed are eliminated, and then according to the task completion probability after adding the user Increment and noise perception range area increment to select the best worker. The probability of a worker completing a task is calculated based on its historical task completion status, and the increment of the task completion probability refers to the part that increases the task completion probability after adding this user. The noise perception range is the union of the areas of the circles with a radius of r and each point on the user's moving track as the center. The noise perception range increment refers to the size of the area where the noise perception range of the entire system is expanded after adding this user.

The goal of this task allocation method is to optimize the location-based query task objective and the noise-aware task objective. This can be reflected by the objective function. The goal of location-based tasks is to ensure the high-quality completion of the task as much as possible, and requires workers to be able to respond positively, requiring the workers' active perception capabilities; the goal of noise perception tasks is to cover the entire urban area as much as possible, requiring The most important thing is the worker's passive perception ability.

In this embodiment, the modeling method for the entire multi-target perception task allocation model is as follows: first, a location-based query task target allocation model is established, and its goal is to maximize the completion success rate of the query task. The formula of the objective function max F(S) is as follows :

Among them, t _i represents the i-th query task, T represents the set of all query tasks, m represents the size of the query task set, W _i represents the set of all workers who accept the query task t _i , w _j represents the worker set W _i a worker, Indicates the completion success rate of the i-th query task, and p _j indicates the completion success rate of the j-th worker's historical task;

Then establish a noise automatic perception task target allocation model, the goal of which is to maximize the sensing coverage. The formula of the objective function max G(S) is as follows:

in, W _i represents the set of employed workers, Indicates the expected path of each worker w _j ∈ W _i ;

Considering the above two task target allocation models, and at the same time requiring that the total cost of constrained workers is not greater than the budget given by the employer, the mobile multi-target perception task allocation model is established as follows:

Among them, W represents the set of all workers, S represents the set of employed workers, c _j represents the cost of the hired workers to complete the task, and B represents the budget given by the employer.

The solution to the established model is an NP-hard problem. At present, there is no algorithm with polynomial time complexity that can be solved to obtain an optimal solution. In this embodiment, a solution method based on a greedy algorithm is used to solve the mobile multi-target perception task allocation model, that is, MBC- The greedy algorithm can efficiently solve approximate solutions. The specific process is:

Step 110: Initialize the set S of hired workers, set W _i of all workers who accept the query task t _i as an empty set, and the cost C required by the hired workers to complete the task is 0;

Step 120, combine all valid task-worker pairs (t _i , w _j ) and assign them to M;

Step 130, if M is not empty, then execute step 140, otherwise end the algorithm;

Step 140, eliminating the workers whose expenses required to complete the task are too high and will exceed the given budget of the employer;

Step 150. For each remaining valid task-worker pair (t _i , w _j ), calculate the weighted increment of query task completion probability and perception coverage respectively

in

Step 160. Eliminate inferior task-worker pairings. The pairing (t _i , w _j ) is better than (t′ _i , w′ _j ) or the pairing (t′ _i , w′ _j ) is worse than the pairing (t _i , w _j ), means at the same time or at the same time

Step 170, sorting the remaining task-worker pairings in descending order according to their superiority over other task-worker pairs;

Step 180. Select the top-ranked pair (t _i , w _j ) as the result of an iteration, and remove all pairs {(t _i , _{w j )} |t _i ∈ T before the next iteration }, and then return to step 130, and loop iteratively until the optimal number of workers selected by the solution meets the requirements of the employer.

As shown in Figure 2, the MBC-Greedy algorithm represents the algorithm designed in this embodiment. In order to verify the performance of this algorithm, this embodiment is described by comparing it with the C-Greedy algorithm and the Q-Greedy algorithm. The C-Greedy algorithm tries to maximize the size of the noise-aware coverage area under the budget constraint. The algorithm also uses a weighted greedy strategy, each time selecting the worker that can make the weighted increment of the noise-aware coverage area the largest; Maximize the probability of completion of the location-based query task under the constraints. After selecting a worker, calculate the increment of the new worker's coverage to update the total noise-aware coverage.

Figure 2 shows the solutions obtained by the three algorithms in the simulation task allocation system. From the perspective of multi-objective optimization, the solutions obtained by the three algorithms are roughly located at the non-dominated frontier, but it can be seen that the algorithm designed in this embodiment It performs relatively well in both goals, which is more in line with the actual application situation. The total number of query tasks m used in this simulation task allocation system is 200, and the total budget B is 400.

Figure 3 shows that as the number of query tasks increases, the noise-aware coverage of all algorithms decreases. This trend is inevitable under a certain budget, but it can be seen from the figure that MBC-Greedy is compared with C- Greedy coverage is slightly smaller (less than 5%) and larger than Q-Greedy (over 15%).

Figure 4 shows that as the number of query tasks increases, the probability of query task completion also decreases, but it can be seen that the task completion probability of MBC-Greedy is slightly smaller (less than 4%) compared with Q-Greedy, and larger than that of C-Greedy ( from 15% to 38%).

The above is only a preferred embodiment of the patent of the present invention, but the scope of protection of the patent of the present invention is not limited thereto. The equivalent replacement or change of the technical solution and its invention patent concept all belong to the protection scope of the invention patent.

Claims (4)

1. A mobile multi-target crowd sensing task allocation method is characterized by comprising the following steps:

step S1, the employer issues a location-based query task to a task distribution system;

step S2, under the requirement of an employer, the task allocation system simultaneously considers the maximum perception coverage and the task completion rate to establish a mobile multi-objective perception task allocation model, solves and selects the best worker, and allocates the query task to the worker;

and step S3, after receiving the query task, the worker assigned with the query task executes the query task, automatically senses the query task in the way of the position of the query task, and finally returns the query task result and the automatic sensing data to the task assignment system.

2. The method as claimed in claim 1, wherein the mobile multi-objective crowd sensing task assignment model in step S2 includes a location-based query task object assignment model and an auto-sensing task object assignment model.

3. The method as claimed in claim 2, wherein the objective of the location-based query task objective assignment model is to maximize the success rate of the query task, and the objective function max F (S) is as follows:

wherein, t_iRepresents the ith query task, T represents all the query task sets, m represents the size of the query task set, W_iRepresenting the accepting of a query task t_iAll worker sets of w_jRepresenting a set of workers W_iOne of the workers in the group of workers,indicating the completion success rate of the ith query task,_jindicating the historical task completion success rate of the jth worker;

the goal of the automatic perception task target distribution model is to maximize the perception coverage, and the objective function max G (S) is as follows:

wherein,representing the set of workers employed,represents each worker w_j∈W_iA predicted path;

considering the two task objective allocation models described above, and requiring that the total cost of the constraint workers is not greater than the budget given by the employer, a mobile multi-objective aware task allocation model is established as follows:

where W represents the set of all workers, S represents the set of workers hired, c_jIndicating the cost of the hired worker to complete the task and B indicating the budget given by the hirer.

4. The method for distributing the mobile multi-target crowd sensing task according to claim 3, wherein solving the mobile multi-target sensing task distribution model adopts a solving method based on a greedy algorithm, and the concrete process is as follows:

step 110, initialize the set of workers employed S, accept the query task t_iAll worker set W of_iFor an empty set, the cost C required for the hired worker to complete the task is 0;

step 120, combine all valid task-worker pairs (t)_i,w_j) Assigning a value to M;

step 130, if M is not empty, executing step 140, otherwise ending the algorithm;

at step 140, eliminating workers whose cost is too high to complete the task can result in exceeding the given budget of the employer;

step 150, for each task-worker pairing that remains valid (t)_i,w_j) Respectively calculating the weighted increment of the completion probability and the perception coverage range of the query task

Wherein

Step 160, eliminating the task-worker pairs at the disadvantage, and pairing (t)_i,w_j) Is superior to (t)_i′,w_j') or pairings (t)_i′,w_j') is inferior to the pairing (t)_i,w_j) Means thatAt the same time haveOr isAt the same time have

Step 170, sorting the remaining task-worker pairs in a descending order according to the quantity superior to other task-worker pairs;

step 180, select the top ranked pair (t)_i,w_j) As a result of one iteration, and all worker-containing w are removed before the next iteration_jPair of { (t)_i,w_j)|t_iE.t, then returning to step 130, and looping the iteration until the optimal number of workers selected to solve the employer's requirement is reached.