Dynamic task allocation method fusing task urgency and sensor performance

Abstract

The invention discloses a dynamic task allocation method fusing task urgency and sensor performance. The method comprises the following steps that S1, a task publisher publishes information on a platform; s2, after receiving the task information broadcasted by the platform, the workers upload own information to the platform, and compete the qualification of participating in the task; the platform measures and calculates an equilibrium factor according to the owned information; and S3, task allocation: substituting the calculated equilibrium factor into an allocation algorithm to perform task allocation, and performing task allocation on workers based on the equilibrium factor. According to the method, the performance levels of different sensors carried by different workers in the mobile crowd sensing system are distinguished; the problem that the data quality of a high-precision task is poor due to wrong task allocation caused by homogenization of the performance of the equipment sensor in a traditional allocation strategy is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a dynamic task allocation method that integrates task urgency and sensor performance. Background Art

[0002] Mobile crowd sensing, as an emerging data collection and processing mode, is widely applied in various fields. It can efficiently complete various tasks such as environmental monitoring, traffic monitoring, and public security monitoring by using the sensors carried by ordinary users' smart devices. In the field of mobile crowd sensing, the design of task allocation strategies directly determines the data collection efficiency and quality of the system in complex scenarios. Traditional task allocation methods are mostly optimized based on user geographical location, device availability, or simple reputation models, but it is difficult to meet the requirements of vertical scenarios with high time urgency or sensitive data quality (such as disaster emergency response, medical health monitoring, real-time diagnosis of industrial equipment, etc.). The existing technologies have the following significant problems: First, in the task allocation strategy, the differences in the time sensitivity and sensor performance requirements of different tasks are not fully considered, and most of the existing inventions adopt a unified task allocation strategy. This will lead to a problem that the task allocation scheme applicable to time-critical tasks may not be applicable to sensing tasks with high requirements for data accuracy. The current development not only requires screening out workers to perform tasks, but also requires workers to meet the requirements of vertical scenarios. Second, most current systems usually assume that the sensing capabilities of all worker devices are homogeneous, but this is not the case. With the increasing number and variety of smart devices, if the sensor performance of devices is still assumed to be homogeneous, it will inevitably deviate from the actual situation in the current world and may cause some unnecessary errors. For example, in the task of air quality monitoring, users with low-precision sensors may be wrongly assigned tasks, resulting in insufficient data accuracy and credibility. Therefore, refined sensing performance in mobile crowd sensing is necessary.

[0003] Moreover, the working habits and levels of many workers on the mobile crowd sensing system are also uneven. For time-critical sensing tasks, the platform pays more attention to completing work tasks according to strict time limits. For most existing platforms, although some workers have high-performance sensors, they may be eliminated by the platform because they cannot absolutely meet the time limit requirements of sensing tasks. This is a waste of resources for the mobile crowd sensing system. In fact, this part of workers can be retained in the mobile crowd sensing system by designing a new task allocation mechanism and using their advantages to perform sensing tasks. If there is a dynamic task allocation mechanism that can integrate sensor performance and time urgency, by quantifying the coupling relationship between device sensor capabilities and task urgency constraints, the collaborative optimization of sensor performance and task urgency can be achieved. In this way, workers with different characteristics can be retained in the same mobile crowd sensing system, increasing the resource volume of the mobile crowd sensing system and improving the task allocation rate and completion rate. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a dynamic task allocation method that integrates task urgency and sensor performance, including the following steps: Step S1: The task publisher publishes task information on the platform according to the publishing model, and preset the task urgency index and the sensing performance index; Step S2: The worker receives the task information broadcast by the platform, uploads the worker information to the platform, and competes for the qualification to participate in this task; obtain the balance factor based on the task urgency index and the sensing performance index in Step S1; Step S3: Task allocation, the task publisher allocates tasks to the workers based on the balance factor through the allocation algorithm; optimize the task allocation process based on the improved genetic algorithm to obtain the optimal task allocation result.

[0005] Further, the publishing model described in Step S1 is expressed as: ; ; ; Among them, represents the task information, is the sensor performance index, is the task urgency index, is the task set, is the task budget, is the th task published by this task publisher, and are respectively the start time and the end time of the th task published by this task publisher, is the location information of the th task published by this task publisher, respectively represent the abscissa and the ordinate of the th task published by this task publisher, represents the sensor type of the th task published by this task publisher.

[0006] Further, the mathematical model of the worker information in Step S2 is as follows: ; Among them, is the th worker information, is the sensor information of the th worker, represents the index as The type of sensor Indicates the index is The performance of the sensor Indicates the index is The type of sensor Indicates the index is The performance of the sensor; Indicates the on-time completion status of the most recent task of the th worker, either 1 or -1. 1 indicates that the task is completed on time; -1 indicates that the task is not completed on time, Indicates the on-time completion status of the th historical task; Indicates the expected bid of the th worker for the current round of tasks.

[0007] Furthermore, in step S2, when workers compete for tasks, the platform will read the information of each worker participating in the current round of competition and comprehensively calculate the equivalent value of the on-time completion status of the workers' most recent tasks to obtain the balance factor between the task urgency index and the sensor performance, The process of obtaining the balance factor is specifically as follows: According to the task urgency index set by the task publisher and the sensing performance index to determine the time urgency threshold in the allocation algorithm as the balance factor, expressed as: ; ; ; ; Among them, Indicates the equivalent value of the on-time completion status of the most recent task of the th worker, Indicates the average on-time completion status of the recent tasks of the current round of candidate workers; is the total number of tasks, is the total number of candidate workers, is the influence coefficient; is the efficiency coefficient; Determine the task preference of the task publisher according to the magnitudes of the task urgency index and the sensing performance index, and make allocations according to the task preference.

[0008] Furthermore, step S3 includes: Step S31: Shuffle the order of the tasks and assign eligible workers to each task through an improved genetic algorithm; Calculate the work level index for each worker meeting the requirements, sort them, and initially select the worker with the largest work level index value as the worker to execute the task; repeat this process several times to obtain the initial population; Work level index It is expressed as: ; Among them, is the type of sensor required for task and represents the performance level equivalent value calculation function.

[0009] Furthermore, step S32: Adjust and replace workers who do not meet the requirements, check the individuals in the population, and adjust the individuals who do not meet the time urgency threshold; the conditions to be met are expressed as: ; Among them, represents the cost of task ; represents the set of sensor types extracted from the information of the replaced worker ; Obtain the result subset as an individual in the population, represents that task is assigned to worker , represents that task is assigned to worker ; Step S33: Replace the replaceable workers; the replacement conditions are expressed as: ; Among them, represents the replaced worker, is the budget after replacement , represents the type of sensor required extracted from task , represents the set of sensor types extracted from the information of the replaced worker , represents the performance level equivalent value of the required sensor type of the original worker w , represents the of the replaced worker required sensor type performance level equivalent value; Obtain the result subset.

[0010] Furthermore, step S3 also includes: Step S34: Calculate the population fitness , according to the fitness Output the best solution; expressed as: ; ; Among them, is the balance factor calculation function; is the weight of the average value of the sensor horizontal equivalent value, is the weight occupied by the average value of is the index of the worker, is the index of the task, represents the total number of tasks, represents the worker the performance level equivalent value of the corresponding sensor required for the carried task, represents the index range of the worker, is the number of groups corresponding to tasks and workers in the individuals of the population; After obtaining the fitness values of all individuals in the population, select the individual with the maximum fitness as the best allocation solution.

[0011] Compared with the existing technology, the present invention has the following beneficial effects:

[0012] (1) The present invention improves the task allocation mechanism. By introducing the task urgency index and the sensor performance index to fuse the sensor performance and the task urgency, it realizes collaborative optimization. Compared with other inventions, the task allocation mechanism of the present invention can be dynamically adjusted according to the task urgency index and the sensor performance index of each round of tasks, and the task allocation mechanism has a wider applicability.

[0013] (2) The present invention differentiates the performance levels of different sensors carried by different workers in the mobile crowd sensing system; solves the problem that the traditional allocation strategy homogenizes the performance of device sensors and wrongly allocates tasks. And introduces a balance factor to achieve the balance between the task urgency and the sensor performance of the task allocation mechanism. Enables the crowd sensing platform to maximize the advantages of platform workers according to the task requirements.

[0014] (3) The present invention designs a task allocation algorithm according to the characteristics of the proposed task allocation mechanism. This algorithm is improved with reference to the genetic algorithm. This algorithm designs a two-stage optimization scheme to replace the relatively complex optimization processes of crossover and mutation in the genetic algorithm. This algorithm can adapt to the dynamic adjustment characteristics of the task allocation mechanism proposed by the present invention and obtain the optimal task allocation solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is the system architecture diagram of the present invention.

[0016] Figure 2 It is a comparative experimental diagram of the situation where the tasks of the present invention are completed on time. Specific implementation manners

[0017] Reference Figure 1 , which is the system architecture diagram of mobile crowd sensing of the present application. It mainly consists of a task publisher, a mobile crowd sensing platform, and workers for sensing data. The task publisher is the demanding party and needs to upload relevant information of the task to the platform. After receiving the information, the platform will broadcast the task to the worker group to recruit the workers to execute this task. Then the workers decide whether to participate in this competition, and the workers participating in this competition will form a candidate worker set. Finally, the task matching mechanism of the platform finds the best worker group from the candidate worker set and gives the best task allocation plan. After the task allocation is completed, the workers will perform the sensing task according to the agreement and upload the data to the platform. The platform will issue rewards to each worker according to the execution situation of the task, and then perform some simple processing on the data and send it to the task publisher.

[0018] A dynamic task allocation method that fuses task urgency and sensor performance includes the following steps: Step S1: The task publisher publishes task information on the platform according to the publishing model, and preset the task urgency index and the sensing performance index; Step S2: The worker receives the task information broadcast by the platform, uploads the worker information to the platform, and competes for the qualification to participate in this task; obtain the balance factor based on the task urgency index and the sensing performance index in Step S1; Step S3: Task allocation. The task publisher allocates tasks to the workers based on the balance factor through the allocation algorithm; optimize the task allocation process based on the improved genetic algorithm to obtain the optimal task allocation result.

[0019] Further, the publishing model described in Step S1 is expressed as: ; ; ; Among them, represents the task information, is the sensor performance index, is the task urgency index, is the task set, is the task budget, is the th task published by the task publisher this time, and are respectively the The start time and end time of a task, is the location information of the th task published by the task publisher for this time, respectively representing the abscissa and ordinate of the th task published by the task publisher for this time, represents the sensor type of the

[0020] th task published by the task publisher for this time. ; where, is the th worker information, is the th worker's sensor information, represents the sensor type with index , represents the sensor performance with index , represents the sensor type with index , represents the sensor performance with index ; represents the on-time completion status of the nearest task of the th worker, both being 1 or -1, 1 indicating on-time task completion; -1 indicating uncompleted task on time, represents the on-time completion status of the th historical task; represents the expected quote of the th worker for this round of tasks.

[0021] Furthermore, in step S2, when workers compete for tasks, the platform will read the information of each worker participating in this round of competition and comprehensively calculate the equivalent value of the on-time completion status of the workers' nearest tasks to obtain the balance factor between the task urgency index and the sensor performance. The process of obtaining the balance factor is specifically as follows: According to the task urgency index set by the task publisher and the sensing performance index to determine the time urgency threshold in the allocation algorithm as the balance factor, expressed as: ; ; ; ; where, Represents the equivalent value of the on-time completion of the most recent tasks of the th worker, indicating the average on-time completion of the recent tasks of the candidate workers in this round; is the total number of tasks, is the total number of candidate workers, is the influence coefficient; is the efficiency coefficient; Determine the task preference of the task publisher according to the magnitude of the task urgency index and the sensing performance index, and allocate according to the task preference.

[0022] Furthermore, step S3 includes: Step S31: Shuffle the order of the tasks, and allocate workers meeting the requirements to each task through an improved genetic algorithm; Calculate the work level index of each worker meeting the requirements, sort them, and initially select the worker with the largest work level index value as the worker to execute this task; and so on, repeat several rounds to obtain the initial population; Work level index is expressed as: ; where is the type of sensor required for task , represents the performance level equivalent value calculation function.

[0023] Furthermore, step S32: Adjust and replace the workers who do not meet the requirements, check the individuals in the population, and adjust the individuals who do not meet the time urgency threshold; the conditions to be met are expressed as: ; where represents the cost of task , represents the set of sensor types extracted from the information of the replaced worker , Obtain the result subset as an individual in the population, represents that task is assigned to worker , represents that task is assigned to worker ; Step S33: Replace the replaceable workers; the replacement conditions are expressed as: ; where represents the replaced worker, is after replacement The budget, represents extracting the required sensor types from the task and represents extracting the set of sensor types from the information of the replaced worker ; represents the equivalent value of the performance level of the required sensor types of the original worker w ; represents the equivalent value of the performance level of the required sensor types of the replaced worker ; ; Obtain the result subset.

[0024] Furthermore, step S3 further includes: Step S34: Calculate the population fitness , and output the optimal solution according to the fitness ; expressed as: ; ; where is the equilibrium factor calculation function; is the weight of the mean of the sensor level equivalent values, is the weight of the mean of is the index of the worker, is the index of the task, represents the total number of tasks, represents the equivalent value of the performance level of the sensors required for the corresponding task carried by the worker ; represents the index range of the workers, is the number of groups corresponding to tasks and workers in the individuals of the population; After obtaining the fitness values of all individuals in the population, select the individual with the maximum fitness as the optimal allocation plan.

[0025] Experimental results

[0026] In step S2, calculate the equilibrium factor and obtain the threshold for task allocation according to the equilibrium factor, the time urgency index, and the sensor performance index. The results of different task allocations tend to be different. For example Figure 2As shown, under the condition that other factors are the same, task allocation is carried out with different thresholds for the task allocation algorithm. The result is the on-time completion of the average tasks of the finally matched workers and the performance level of their sensors. The solid line is the right vertical coordinate corresponding to the on-time completion of the workers, and the dashed line is the average performance level of the workers' sensors, corresponding to the left vertical coordinate. The abscissa is the value of the threshold. It can be seen from the figure that as the threshold continues to increase, the on-time completion of the workers' tasks will get better and better, but some workers with excellent sensor performance but not very good on-time performance will be eliminated. It can be seen from the figure that when the task requires as many workers with high-performance sensors as possible and there is no requirement for the task completion time limit, the threshold should be set smaller, and vice versa, the threshold can be larger. In this way, the resource utilization rate can be maximized according to the requirements of different tasks.

[0027] As a further illustration, mobile crowd sensing is a technology that uses intelligent devices carried by workers for data collection and environmental perception. This solution targets two types of workers. The first type of sensing workers are like local residents and the general public. The characteristics of this type of workers are that they have free time and can mostly start and complete tasks on time strictly according to the time limit of the sensing tasks, but the sensors they possess are of relatively ordinary performance. The other type is practitioners engaged in some professional work, such as logistics workers and couriers. They are equipped with high-quality GPS modules, RFID readers, and temperature and humidity sensors (for goods that require environmental control). The sensors they are equipped with are not available to most of the first type of workers and their performance is also better. For some environmental detection tasks that require high-performance sensors to execute, this type of workers can be assigned to execute, and because of the high-quality sensors they are equipped with, the sensed data will be more accurate. There are also maintenance technicians. They are usually equipped with professional tools such as voltage detectors and infrared thermal imagers. They can help relevant departments obtain the status information of power lines distributed in different regions in a timely manner and assist relevant departments in power maintenance. There are also devices such as vibration sensors and ultrasonic leak detectors that can capture the usage conditions of gas supply pipes and water supply pipes, and can remind relevant departments to carry out maintenance according to the provided information. The characteristics of this type of workers are that the sensors they carry are relatively rare and have better performance, but since they are industry professionals, their time is not as free as that of the first type of workers, and this group of people cannot execute strictly time-pressured sensing tasks. If the platform abandons this type of workers with high-quality sensors because of this, it will be a very big loss. Therefore, in this application, the balance factor of the task urgency index U and the sensor performance index P is calculated for each round of tasks, and the enabling of this type of workers is controlled through the time urgency threshold of the task, so that the platform can select workers with higher-performance sensors as much as possible under the condition of meeting the task time urgency as much as possible when allocating each round of tasks.

[0028] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A dynamic task allocation method that integrates task urgency and sensor performance, characterized in that, It includes the following steps: Step S1: The task publisher publishes task information on the platform according to the publishing model, and preset the task urgency index and sensing performance index; Step S2: The worker receives the task information broadcast by the platform, uploads the worker information to the platform, and competes for the qualification to participate in this task; Obtain the balance factor based on the task urgency index and sensing performance index in Step S1; Step S3: Task allocation. The task publisher allocates tasks to workers based on the balance factor through the allocation algorithm; Optimize the task allocation process based on the improved genetic algorithm to obtain the optimal task allocation result.

2. The dynamic task allocation method that integrates task urgency and sensor performance according to claim 1, wherein The publishing model described in Step S1 is expressed as: ； ； ； Among them, represents task information, is the sensor performance index, is the task urgency index, is the task set, is the task budget, is the th task issued by the current task publisher, and are respectively the start time and end time of the th task issued by the current task publisher, is the location information of the th task issued by the current task publisher, respectively represent the abscissa and ordinate of the th task issued by the current task publisher, represents the sensor type of the th task issued by the current task publisher.

3. The dynamic task allocation method integrating task urgency and sensor performance according to claim 1, characterized in that The mathematical model of the worker information in Step S2 is as follows: ； Among them, is the th worker information, is the th worker's sensor information, represents the sensor type with index , represents the sensor performance with index , represents the sensor type with index , represents the sensor performance with index ; represents the on-time completion status of the most recent task of the th worker, all being 1 or -1, where 1 indicates on-time task completion; -1 indicates non-on-time task completion, represents the on-time completion status of the th historical task; represents the expected bid of the th worker for the current round of tasks.

4. A dynamic task allocation method that integrates task urgency and sensor performance according to claim 3, characterized in that In Step S2, when the worker competes for the task, the platform will read the information of each worker participating in this round of competition, and comprehensively calculate the equivalent value of the on-time completion of the worker's recent tasks, and obtain the balance factor of the task urgency index and the sensor performance. The process of obtaining the balance factor is specifically as follows: According to the task urgency index set by the task publisher and the sensing performance index to determine the time urgency threshold in the allocation algorithm as the balance factor, expressed as: ； ； ； ； Among them, represents the equivalent value of the on-time completion of the most recent tasks of the th worker, represents the average on-time completion of the recent tasks of the candidate workers in this round; is the total number of tasks, is the total number of candidate workers, is the influence coefficient; is the efficiency coefficient; Determine the task preference of the task publisher according to the magnitudes of the task urgency index and the sensing performance index, and make the allocation according to the task preference.

5. The dynamic task allocation method integrating task urgency and sensor performance according to claim 4, wherein Step S3 includes: Step S31: Shuffle the order of the tasks, and allocate eligible workers to each task through the greedy algorithm; Calculate the work level index of each eligible worker, sort them, and initially select the worker with the largest work level index value as the worker to execute this task; Repeat several rounds in this way to obtain the initial population; Working level index Expressed as: ； Among them, is the type of sensor required for the task, represents the performance level equivalent value calculation function.

6. The dynamic task allocation method integrating task urgency and sensor performance according to claim 5, characterized in that Step S32: Adjust and replace the workers who do not meet the requirements, check the individuals in the population, and adjust the individuals who do not meet the time urgency threshold; The conditions to be met are expressed as: ； Among them, represents the cost of the task ; represents the set of sensor types extracted from the information of the replaced worker ; Obtain a subset of results As an individual in the population, representing a task assigned to a worker , representing a task assigned to a worker ; Step S33: Replace the replaceable workers; The replacement conditions are expressed as: ； Among them, represents the replaced worker, is the replaced budget, represents extracting the required sensor types from the task ; represents the set of sensor types extracted from the information of the replaced worker ; represents the equivalent value of the performance level of the required sensor types of the original worker w ; represents the equivalent value of the performance level of the required sensor types of the replaced worker ; ; Obtain the result subset.

7. A dynamic task allocation method that integrates task urgency and sensor performance according to claim 6, characterized in that Step S3 also includes: Step S34: Calculate the fitness of the population , and output the optimal solution according to the fitness ; It is expressed as: ； ； Among them, is the equilibrium factor calculation function; is the weight of the mean of the sensor horizontal equivalent value, is the weight of the mean of is the index of the worker, is the index of the task, represents the total number of tasks, represents the worker carries the performance level equivalent value of the sensors required for the corresponding tasks, represents the index range of the worker, is the number of groups corresponding to tasks and workers in the individuals of the population; After obtaining the fitness values of all individuals in the population, select the individual with the maximum fitness as the best allocation plan.

Images