Global optimal Nash equilibrium task allocation method for double targets

Through the game theory model and NSGA-II algorithm, combined with simulated annealing and PCA methods, the balance problem between worker satisfaction and platform profit in space-time crowdsourcing is solved, Nash equilibrium task allocation is achieved, and worker satisfaction and platform services are improved. quality.

CN118917587APending Publication Date: 2024-11-08ZHEJIANG UNIV OF SCI & TECH +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410949832.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Existing space-time crowdsourcing task allocation methods usually aim at maximizing the overall profit of the system, paying less attention to worker satisfaction, which makes it difficult to balance worker satisfaction and profit, affecting user satisfaction and platform service quality.

Method used

The game theory model is used combined with the NSGA-II algorithm, and the Pareto optimal frontier is identified through simulated annealing algorithm and PCA method, and the dual-objective optimal strategy between worker satisfaction and platform profit is found to achieve Nash equilibrium task allocation.

Benefits of technology

Find a balance between worker satisfaction and system profit, improve workers' satisfaction with tasks and overall service quality of the platform, reduce storage consumption of dual-objective queries, and verify the effectiveness of the dual-objective task allocation method.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917587A_ABST
    Figure CN118917587A_ABST
Patent Text Reader

Abstract

The invention relates to a double-target-oriented global optimal Nash equilibrium task allocation method, which comprises the following steps of: 1, collecting tasks and worker resources in a timestamp, and calculating specific preferences of workers, and 2, by taking maximization of worker satisfaction as a target, obtaining a Nash equilibrium point under a single target by adopting a simulated annealing algorithm, finding out a global optimal strategy combination; 3, taking maximization of the overall profit of the platform as a target, adopting a simulated annealing algorithm to obtain a Nash equilibrium point under a single target, and finding out a global optimal strategy combination; 4, checking whether overlapping exists in the optimal strategy combination under the single target, if yes, obtaining a Nash equilibrium point with the maximum sum value, and taking the combination as a finally selected task allocation scheme; and step 5, if not, using an NSGA-II algorithm to obtain a Pareto optimal frontier strategy, and selecting an optimal strategy under an overall dual-target game according to a Pareto optimal frontier curve, and solving a dual-target task allocation problem of obtaining balance between worker satisfaction and profit improvement.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a dual-objective oriented global optimal Nash equilibrium task allocation method, belonging to the technical field of spatiotemporal task crowdsourcing allocation. Background Art

[0002] Spatiotemporal crowdsourcing refers to a new crowdsourcing model that combines crowdsourcing and geographic information technology, that is, the spatial information tasks of the task publisher are sent to the general public through the network platform, and the crowdsourcing workers use their mobile devices to collect and upload geographic information data, complete the task and receive corresponding rewards. Therefore, spatiotemporal crowdsourcing can be regarded as a crowdsourcing model based on spatial location and time. It utilizes the distribution characteristics of crowdsourcing workers at different times and places to realize the collection, processing and analysis of large-scale, distributed and heterogeneous geographic information data.

[0003] Spatial Crowdsourcing combines the traditional crowdsourcing model with geographic information technology, opening up a new field of crowdsourcing applications. This model publishes tasks related to spatial information to a wide range of unspecified people through an online platform. Participants use their mobile devices, such as smartphones or tablets, to collect geographic information data and upload them to the platform. Once the task is completed, the participants will receive corresponding rewards.

[0004] The core of the spatiotemporal crowdsourcing problem lies in the matching between tasks and workers, that is, how to distribute the right work to the right workers.

[0005] When allocating tasks in spatiotemporal crowdsourcing, we must not only consider the overall profit of the platform, but also the preferences of workers. Fully considering the preferences of workers in task allocation is conducive to improving user satisfaction and the overall service quality of the platform. Workers' preferences are in opposition to the overall profit of the platform. Although workers accepting and performing multiple tasks at the same time may bring higher economic returns, various constraints in reality will hinder workers from accepting and performing multiple tasks at the same time, which has become the main challenge of task allocation. For example, in a specific time period, traffic congestion may cause workers to be unable to complete tasks on schedule, thereby affecting the on-time delivery of subsequent tasks. This situation is particularly prominent during peak hours and in large cities, so workers have personal preferences for the number of tasks they receive. Although accepting tasks that are far away may bring higher economic benefits in some cases, for the sake of time and cost-effectiveness, workers usually prefer to choose tasks that are closer to their current location. Workers have personal preferences for the average spatial and temporal cost of tasks. Based on considerations such as distance, safety, and long-term career development, workers may be dissatisfied with some tasks assigned by the system. Users have personal preferences for the rewards of tasks and the difficulty of tasks. When faced with unsatisfactory tasks, workers’ work enthusiasm and efficiency will drop significantly. Workers may fail to complete tasks on time for various reasons, or show a perfunctory attitude in the process of performing tasks, which ultimately leads to the task failing to achieve the expected quality. This will directly affect customer satisfaction, causing customers to be dissatisfied with workers and platforms and give negative reviews.

[0006] Although workers' preferences are important for task allocation, existing studies usually aim to maximize the overall profit of the system and pay less attention to the dual-objective task allocation problem of striking a balance between addressing worker satisfaction and improving profits. Summary of the invention

[0007] In view of the shortcomings of the existing allocation schemes, the present invention proposes a global optimal Nash equilibrium task allocation method for the dual objectives of worker satisfaction and platform profit. Because there is no linear promotion between the worker's task satisfaction and the overall platform profit, that is, blindly focusing on the worker's satisfaction will affect the overall profit of the platform. Therefore, the present invention proposes a new balance mechanism, that is, to reconcile the relationship between overall satisfaction and overall profit by proposing a game theory model. In this game model, "overall profit" and "overall satisfaction" are regarded as opposing players, and their strategy set is based on all single-objective strategy combinations that match workers and tasks. The present invention identifies the Pareto optimal frontier by implementing the NSGA-II algorithm, thereby finding the optimal strategy under the dual objectives.

[0008] The specific steps of a dual-objective global optimal Nash equilibrium task allocation method are as follows:

[0009] Step 1: Collect the tasks to be assigned and the available worker resources within the timestamp, and calculate the specific preferences of each worker. The tasks to be assigned include the tasks that were not successfully assigned earlier due to insufficient satisfaction and the tasks that have newly arrived at the platform system. The available worker resources include the old workers who have not accepted the tasks that have reached their personal capacity limit before the current timestamp and the new workers who have recently joined the platform system.

[0010] Step 2: To maximize the worker satisfaction goal, use the simulated annealing algorithm to obtain the Nash equilibrium point in the multi-worker multi-task selection under a single goal, and find the global optimal strategy combination;

[0011] Step 3: With the goal of maximizing the overall profit of the platform, the simulated annealing algorithm is used to obtain the Nash equilibrium point in the multi-worker multi-task selection under a single objective, and find the global optimal strategy combination;

[0012] Step 4: Check whether there are overlapping strategies in the optimal strategy combinations under the single objective of step 2 and step 3. If so, obtain the Nash equilibrium point with the largest total value and use this combination as the final selected task allocation plan;

[0013] Step 5: In the absence of overlapping single-objective optimal strategies, use the NSGA-II algorithm to obtain the Pareto optimal frontier strategy, and select the optimal strategy under the overall dual-objective game according to the Pareto optimal frontier curve.

[0014] Step 6: Conduct experimental tests on the dataset to verify the effect of dual-objective task allocation.

[0015] The step 1 uses the PCA method to calculate the specific preferences of workers, specifically including:

[0016] Step 1.1: Take the total time spent on the task and the number of tasks received as negative indicators; take the task reward and the space time between tasks as positive indicators and construct an initialization matrix;

[0017] Step 1.2: Through normalization, the differences of various factors can be scaled to a relatively consistent range;

[0018] Step 1.3: Calculate the covariance matrix and calculate the eigenvalues ​​and eigenvectors, calculate the variance proportion of each principal component, determine the number of principal components based on the criterion that the eigenvalue is greater than 1 and the cumulative variance contribution rate is greater than the preset threshold, select the number of principal components determined above, and construct the projection matrix with the corresponding eigenvectors, where each column represents a principal component;

[0019] Step 1.4: Calculate the worker's preference as:

[0020] Sa(D(w))=α*U(w)+β*Spa(w)+χ*Time(w)+δ*Len(w)

[0021] Where (δ, χ, β, α)∈(0,1), U(w) represents the total reward that worker w can obtain after completing the scheduled tasks. If α is set relatively high, it means that the worker's satisfaction mainly comes from the task reward; Spa(w) represents the average idle time of worker w completing the scheduled tasks. If β is set relatively high, it means that the worker chooses a task set with more reserved idle time in order to avoid not being able to complete the task; Time(w) represents the time spent by worker w after completing the scheduled tasks. If χ is set relatively high, it means that the worker likes to receive nearby task sets; Len(w) represents the number of tasks scheduled by worker w. If δ is set relatively high, it means that the worker prefers to complete a task before receiving a new task.

[0022] The simulated annealing algorithm in step 2 and step 3 is to find the global optimal solution under a given single objective, and its input parameters include the parameters involved in the simulated annealing algorithm abstracted from the spatiotemporal crowdsourcing problem:

[0023] Initial temperature, cooling rate, number of iterations, and strategy matrix.

[0024] The simulated annealing algorithm in step 2 and step 3 specifically includes:

[0025] Step 2.1: At the beginning of the algorithm, a task allocation scheme is initialized by random generation. This scheme is a potential optimal solution in the current iteration process. The optimization process of the algorithm is tracked. Two variables, best_assignment and best_score, are used to record the currently found optimal task allocation scheme and its corresponding total satisfaction score respectively.

[0026] Step 2.2: As the algorithm enters the iteration phase, the temperature parameters are first adjusted according to the preset cooling rate in each iteration. This step simulates the annealing phenomenon in the physical process, that is, the system temperature gradually decreases. The algorithm randomly selects a worker and randomly modifies its task allocation strategy to generate a new task allocation scheme new_assignment;

[0027] Step 2.3: Calculate the total satisfaction score of the new solution and compare it with the currently known optimal score. If the satisfaction score of the new solution is higher than the currently recorded optimal score, accept the new solution with a certain probability and update best_assignment and best_score to record this new optimal solution. If the newly generated task assignment solution fails to improve the total satisfaction, the algorithm will not update the current optimal solution but continue to the next iteration.

[0028] Step 2.4: The simulated annealing algorithm can avoid the trap of local optimal solutions while exploring the solution space, and eventually tends to find the global optimal solution or a solution close to the global optimal solution. The iterative process of the algorithm continues until the predetermined number of iterations is reached, or the temperature drops below the preset threshold. At this time, the algorithm is considered to be stable and outputs the currently recorded optimal task allocation plan and the corresponding satisfaction score. This result is regarded as the optimal Nash equilibrium solution to the problem.

[0029] The fourth step uses an iterative traversal method to find out whether the single-objective optimal strategies obtained in the second and third steps overlap. In the scenario where multiple strategy overlap points are identified, the strategy with the highest reward value is selected as the final solution.

[0030] In the case where there is no overlap in strategies, step 5 constructs the Pareto frontier and identifies a series of solutions. Based on the Pareto curve, the optimal strategy choice that balances the two major goals is revealed to the decision maker, including:

[0031] Step 5.1: At the initial stage of the algorithm execution, the step of randomly generating the initial population is performed. Each individual in the population represents a potential task allocation solution;

[0032] Step 5.2: For each individual, determine its quality in the solution space by evaluating its fitness function. The metrics of the fitness function include profit and satisfaction.

[0033] Step 5.3: Perform non-dominated sorting on the population by performing T iteration cycles, that is, classifying the individuals in the population according to the principle of non-dominated sorting so as to divide them into different frontiers;

[0034] Step 5.4: Determine which individuals will be retained by calculating the crowding degree of individuals in each frontier;

[0035] Step 5.5: Generate new solutions through crossover and mutation operations. When the preset maximum number of iterations is reached, the algorithm stops, otherwise it will continue to iterate and eventually return to the Pareto frontier.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention relates to a dual-objective task allocation method that achieves a balance between worker satisfaction and increased profits. The present invention also solves two core problems: improving worker satisfaction with tasks to improve the overall service of the platform; and finding an appropriate balance between increasing worker satisfaction and improving system profits.

[0038] The present invention obtains single-objective global optimal game combinations under the two objectives of worker satisfaction and system total profit, and the simulated annealing algorithm can find the global optimal single-objective strategy combination.

[0039] The present invention uses the NSGA-II algorithm to calculate the Pareto curve, which not only finds the optimal strategy combination of the dual-objective game, but also accelerates the overall dual-objective optimal search process and reduces the storage consumption of the dual-objective query.

[0040] The present invention verifies the effectiveness of the proposed dual-objective game task allocation method through experiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 A dual-objective task allocation flow chart of a dual-objective global optimal Nash equilibrium task allocation method of the present invention;

[0043] Figure 2 It is a flow chart of the simulated annealing algorithm under a single objective of the present invention;

[0044] Figure 3 The present invention is a flow chart of the NSGA-II algorithm for calculating the Pareto curve;

[0045] Figure 4 It is a variation core operation diagram in the NSGA-II algorithm of the present invention;

[0046] Figure 5 is a cross-core operation diagram in the NSGA-II algorithm of the present invention;

[0047] Figure 6 This is a graph showing the experimental results of the dual-objective task allocation of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Reference Figure 1The dual-objective task allocation flow chart is as follows: A global optimal Nash equilibrium task allocation method for dual objectives (worker preference and platform system profit) has the following specific steps:

[0050] Step 1: Initialize worker and task data and calculate worker preferences.

[0051] Collect the tasks to be assigned and the available worker resources within the timestamp. The tasks include the tasks that were not successfully assigned earlier due to insufficient satisfaction and the tasks that have just arrived at the platform system. The available worker resources are divided into two categories: one is the old workers who have not yet accepted the upper limit of their personal capacity before the current timestamp; the other is the new workers who have recently joined the platform system. The PCA method is used to calculate the specific preferences for each worker.

[0052] The PCA method is used to calculate the specific preferences for each worker, including:

[0053] Step 1.1: Take the total time spent on the task and the number of tasks received as negative indicators; take the task reward and the space time between tasks as positive indicators and construct an initialization matrix.

[0054] Step 1.2: Through normalization, the differences between various factors can be scaled to a relatively consistent range, thereby preventing some factors from having an overly significant or insufficient impact on the model.

[0055] Step 1.3: Calculate the covariance matrix and calculate the eigenvalues ​​and eigenvectors. Calculate the variance proportion of each principal component, and determine the number of principal components based on the criteria that the eigenvalue is greater than 1 and the cumulative variance contribution rate is greater than the preset threshold. Select the number of principal components determined above and construct the projection matrix with the corresponding eigenvectors, where each column represents a principal component.

[0056] Step 1.4: Calculate the worker's preference as

[0057] Sa(D(w))=α*U(w)+β*Spa(w)+χ*Time(w)+δ*Len(w)

[0058] Among them, (δ, χ, β, α)∈(0,1), U(w) represents the total reward that worker w can obtain by completing the scheduled tasks. If α is set to a high value, it means that the worker's satisfaction mainly comes from the task reward; Spa(w) represents the average idle time of worker w to complete the scheduled tasks. If β is set to a high value, it means that the worker chooses a task set with more idle time to avoid not being able to complete the task; Time(w) represents the time it takes for worker w to complete the scheduled tasks. If χ is set to a high value, it means that the worker likes to receive nearby task sets; Len(w) represents the number of tasks scheduled by worker w. If δ is set to a high value, it means that the worker prefers to complete a task before receiving a new task.

[0059] Step 2: Under the goal of maximizing worker interest satisfaction, find all the strategy combinations under the global optimal Nash equilibrium. The process of the simulated annealing algorithm is as follows Figure 2 As shown, specifically:

[0060] Step 2.1: At the beginning of the algorithm, a task allocation scheme is initialized by random generation. This scheme is a potential optimal solution in the current iteration. In order to track the optimization process of the algorithm, two variables, best_assignment and best_score, are used to record the currently found optimal task allocation scheme and its corresponding total satisfaction score.

[0061] Step 2.2: As the algorithm enters the iteration phase, the temperature parameters are first adjusted according to the preset cooling rate in each iteration. This step simulates the annealing phenomenon in the physical process, that is, the system temperature gradually decreases. Next, the algorithm randomly selects a worker and randomly modifies its task assignment strategy to generate a new task assignment scheme new_assignment.

[0062] Step 2.3: Calculate the total satisfaction score of the new solution and compare it with the currently known best score. If the satisfaction score of the new solution is higher than the currently recorded best score, then accept the new solution with a certain probability and update best_assignment and best_score to record this new best solution. If the newly generated task assignment fails to improve the total satisfaction, the algorithm will not update the current best solution, but continue to the next iteration.

[0063] Step 2.4: The simulated annealing algorithm can avoid the trap of local optimal solutions while exploring the solution space, and eventually tends to find the global optimal solution or a solution close to the global optimal solution. The algorithm's iteration process continues until the predetermined number of iterations is reached, or the temperature drops below the preset threshold. At this time, the algorithm is considered to be stable and outputs the currently recorded optimal task allocation plan and the corresponding satisfaction score. This result is regarded as the optimal Nash equilibrium solution to the problem.

[0064] Step 3: Under the goal of maximizing overall profit, use the simulated annealing algorithm to find the global optimal Nash equilibrium solution. The overall idea is the same as that of step 2, except that the value obtained by each worker's strategy set under different goals is different from that under the strategy, so different results will be obtained.

[0065] Step 4: By comparing the two sets of Nash equilibrium points, explore the overlap of strategies. In the scenario where multiple strategy overlaps are identified, the strategy with the highest reward value is selected as the final solution. The following is an example to explain:

[0066] Step 4.1: Set the worker's capacity to receive tasks to 1. For worker w1, the tasks he can choose at this time are t1, t2 and t3; the tasks that worker w2 can choose are t4, t5 and t6. Table 1 shows the profits and satisfaction that the two workers can obtain when choosing different tasks. For example, when worker w1 chooses task t2 and worker w2 chooses task t4, the profit that worker w1 can obtain is 3 and the satisfaction with this task is 2; the profit that worker w2 can obtain is 2 and the satisfaction with this task is 4.

[0067] Table 1 Example of strategy matrix under dual objectives

[0068]

[0069] Step 4.2: Table 1 can be divided into Table 2 and Table 3 according to the goal, which represent the strategy choices with profit as the goal and satisfaction as the goal respectively. When worker w1 chooses task t2 and worker w2 chooses task t5, the strategy choice is a Nash equilibrium point. Because worker w2's choice is fixed, if worker w1 chooses task t1, the profit obtained will change from 4 to 2; if worker w2 chooses task t3, the profit obtained will change from 4 to 3, so worker w1 will not change its strategy; if worker w1's choice is fixed, if worker w2 chooses task t4, the profit obtained will still be 4; if worker w2 chooses task t4, the profit will change from 4 to 2; so worker w2 will not change its strategy. By the same token, there are two Nash equilibrium results for task allocation with satisfaction as the goal, namely: (1) worker w1 chooses task t1 and worker w2 chooses task t6; (2) worker w1 chooses task t2 and worker w2 chooses task t5.

[0070] Table 2 Profit-oriented strategy matrix

[0071]

[0072] Table 3 Strategy matrix with worker interest satisfaction as the goal

[0073]

[0074] Step 4.3: If the sets of Nash equilibrium points under these two objectives overlap, that is, worker w1 chooses task t1 and worker w2 chooses task t5, the overlapping part can be directly used as the final result; if there is more than one Nash equilibrium point in the overlapping part, the Nash equilibrium point with the largest sum can be taken.

[0075] Step 5: In the case of no overlapping points, use the NSGA-II algorithm to find the Pareto frontier to determine the optimal strategy combination under the dual objectives. The process of the NSGA-II algorithm to find the Pareto frontier is as follows: Figure 3 As shown, the core operations of NSGA-II algorithm design include selection, crossover, mutation and sorting. Figure 4 shows a cross-core operation diagram, Figure 5 The core operation diagram of the mutation is shown below. Figure 3 , Figure 4 and Figure 5 Introduce the specific process of NSGA-II algorithm to find Pareto frontier:

[0076] Step 5.1: At the initial stage of the algorithm execution, the step of randomly generating the initial population is performed. Each individual in the population represents a potential task allocation solution.

[0077] Step 5.2: For each individual, determine its quality in the solution space by evaluating its fitness function. In this study, the metrics of the fitness function include profit and satisfaction.

[0078] Step 5.3: By performing T iteration cycles, the population is non-dominated sorted, that is, the individuals in the population are classified according to the principle of non-dominated sorting so as to be divided into different frontiers.

[0079] Step 5.4: Determine which individuals will be retained by calculating the crowding degree of each individual in the frontier.

[0080] Step 5.5: Generate new solutions through crossover and mutation operations to increase the diversity of the population. When the preset maximum number of iterations is reached, the algorithm stops, otherwise it will continue to iterate. Finally, it returns to the Pareto frontier.

[0081] Step 6: Experimental verification. The experimental results are as follows Figure 6 As shown, the details are as follows:

[0082] Step 6.1: Experimental data. The Ele.me data was used, with a time range from January 31, 2020 to February 1, 2020. The Ele.me dataset covers 1,000 tasks, 1,000 workers, and 532 locations. The data occurred between 121.071954 and 122.236704 east longitude and 122.236704 north latitude.

[0083] Between 38.892721 and 39.6853409999999994.

[0084] Step 6.2: The goal of the experiment is to determine the Pareto frontier and finally determine the task allocation scheme based on the system's preferences. Specifically, first generate a set of all possible task allocation results based on each worker's strategy set. Each element in the result set represents a possible task allocation scheme. In the process of generating the result set, if a conflict is found between worker strategies, the task allocation scheme is considered invalid and is not included in the result set, thereby reducing the computational burden.

[0085] NSGA-II batches the result set through genetic operations such as selection, crossover, and mutation. These genetic operations simulate the natural selection process and gradually optimize the solutions in the result set through mechanisms such as survival of the fittest, gene recombination, and random mutation. Through these genetic operations, the algorithm gradually optimizes the solutions in the result set, guides them to better solutions, and finally determines the Pareto frontier.

[0086] Through the above method, the present invention effectively determines the optimal task allocation scheme in result sets of different sizes, and selects it according to the preference of the system, ensuring that the optimal multi-objective balance is achieved in complex task allocation problems.

[0087] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. A dual-objective global optimal Nash equilibrium task allocation method, characterized by: The following steps are involved: Step 1: Collect the tasks to be assigned and the available worker resources within the timestamp, and calculate the specific preferences of each worker. The tasks to be assigned include the tasks that were not successfully assigned earlier due to insufficient satisfaction and the tasks that have newly arrived at the platform system. The available worker resources include the old workers who have not accepted the tasks that have reached their personal capacity limit before the current timestamp and the new workers who have recently joined the platform system. Step 2: To maximize the worker satisfaction goal, use the simulated annealing algorithm to obtain the Nash equilibrium point in the multi-worker multi-task selection under a single goal, and find the global optimal strategy combination; Step 3: With the goal of maximizing the overall profit of the platform, the simulated annealing algorithm is used to obtain the Nash equilibrium point in the multi-worker multi-task selection under a single objective, and find the global optimal strategy combination; Step 4: Check whether there are overlapping strategies in the optimal strategy combinations under the single objective of step 2 and step 3. If so, obtain the Nash equilibrium point with the largest total value and use this combination as the final selected task allocation plan; Step 5: In the absence of overlapping single-objective optimal strategies, use the NSGA-II algorithm to obtain the Pareto optimal frontier strategy, and select the optimal strategy under the overall dual-objective game according to the Pareto optimal frontier curve.

2. The dual-objective global optimal Nash equilibrium task allocation method according to claim 1, characterized in that: It also includes step six: conducting experimental tests on the data set to verify the effect of dual-objective task allocation.

3. The dual-objective global optimal Nash equilibrium task allocation method according to claim 1, characterized in that: The first step uses the PCA method to calculate the specific preferences of workers.

4. A dual-objective global optimal Nash equilibrium task allocation method according to claim 1 or 3, characterized in that: In step 1, the PCA method is used to calculate the specific preferences for each worker, which specifically includes: Step 1.1: Take the total time spent on the task and the number of tasks received as negative indicators; take the task reward and the space time between tasks as positive indicators and construct an initialization matrix; Step 1.2: Scale the differences of various factors to a relatively consistent range through normalization; Step 1.3: Calculate the covariance matrix and calculate the eigenvalues ​​and eigenvectors, calculate the variance proportion of each principal component, determine the number of principal components based on the criterion that the eigenvalue is greater than 1 and the cumulative variance contribution rate is greater than the preset threshold, select the number of principal components determined above, and construct the projection matrix with the corresponding eigenvectors, where each column represents a principal component; Step 1.4: Calculate the worker's preference as: Sa(D(w))=α*U(w)+β*Spa(w)+χ*Time(w)+δ*Len(w) Where (δ, χ, β, α)∈(0,1), U(w) represents the total reward that worker w can obtain after completing the scheduled tasks. If α is set high, it means that the worker's satisfaction mainly comes from the task reward; Spa(w) represents the average idle time of worker w completing the scheduled tasks. If β is set high, it means that the worker chooses a task set with more reserved idle time to avoid not being able to complete the task; Time(w) represents the time spent by worker w after completing the scheduled tasks. If χ is set high, it means that the worker likes to receive nearby task sets; Len(w) represents the number of tasks scheduled by worker w. If δ is set high, it means that the worker prefers to complete a task before receiving a new task.

5. The dual-objective global optimal Nash equilibrium task allocation method according to claim 1, characterized in that: The simulated annealing algorithm in step 2 and step 3 is used to find the global optimal solution under a given single objective, and its input parameters include the parameters involved in the simulated annealing algorithm abstracted from the spatiotemporal crowdsourcing problem: initial temperature, cooling rate, number of iterations and strategy matrix.

6. A dual-objective global optimal Nash equilibrium task allocation method according to claim 1 or 5, characterized in that: The simulated annealing algorithm in step 2 and step 3 specifically includes: Step 2.1: At the beginning of the algorithm, a task allocation scheme is initialized by random generation. This scheme is a potential optimal solution in the current iteration process. The optimization process of the algorithm is tracked. Two variables, best_assignment and best_score, are used to record the currently found optimal task allocation scheme and its corresponding total satisfaction score respectively. Step 2.2: As the algorithm enters the iteration phase, the temperature parameter is first adjusted according to the preset cooling rate in each iteration. The algorithm randomly selects a worker and randomly modifies its task assignment strategy to generate a new task assignment scheme new_assignment; Step 2.3: Calculate the total satisfaction score of the new solution and compare it with the currently known optimal score. If the satisfaction score of the new solution is higher than the currently recorded optimal score, accept the new solution according to probability, and update best_assignment and best_score to record this new optimal solution. If the newly generated task assignment solution fails to improve the total satisfaction, the algorithm will not update the current optimal solution, but continue to the next iteration; Step 2.4: The simulated annealing algorithm avoids the trap of local optimal solutions while exploring the solution space, and eventually tends to find the global optimal solution or a solution close to the global optimal solution. The iterative process of the algorithm continues until the predetermined number of iterations is reached, or the temperature drops below the preset threshold. At this time, the algorithm is considered to be stable and outputs the currently recorded optimal task allocation plan and the corresponding satisfaction score. This result is regarded as the optimal Nash equilibrium solution to the problem.

7. The dual-objective global optimal Nash equilibrium task allocation method according to claim 1, characterized in that: The step five specifically includes: Step 5.1: At the initial stage of the algorithm execution, the step of randomly generating the initial population is performed. Each individual in the population represents a potential task allocation solution; Step 5.2: For each individual, determine its quality in the solution space by evaluating its fitness function. The metrics of the fitness function include profit and satisfaction. Step 5.3: Perform non-dominated sorting on the population by performing T iteration cycles, that is, classifying the individuals in the population according to the principle of non-dominated sorting so as to divide them into different frontiers; Step 5.4: Determine which individuals will be retained by calculating the crowding degree of individuals in each frontier; Step 5.5: Generate new solutions through crossover and mutation operations. When the preset maximum number of iterations is reached, the algorithm stops, otherwise it will continue to iterate and eventually return to the Pareto frontier.