Space-time crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching

Abstract

The invention relates to the technical field of man-machine cooperation, in particular to a space-time crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching, and the method comprises the steps: determining the priorities of all to-be-allocated tasks based on a nonlinear supply-demand ratio, a Shapley value method and Monte Carlo optimization; for all distributed unmanned aerial vehicles, carrying out dispersed unmanned aerial vehicle task allocation on all to-be-allocated tasks; taking all workers as participants, taking spatial distribution of daily tracks of the participants as participant preferences, taking importance of tasks to be distributed as task preferences, and establishing a bilateral preference driven incentive mechanism; for all participants, based on a bilateral preference driven incentive mechanism, dynamic task allocation with a Top-k buffer area is carried out, and tasks in a delay acceptance window and an execution window of each participant are determined; when the unmanned aerial vehicle has a new task assigned to the participant, the new task is incorporated into a delayed acceptance window of the participant. The method can provide an optimal solution for man-machine cooperation.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of human-machine collaboration technology, and in particular to a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching. Background Art

[0002] Mobile crowd sensing (MCS) is a technology that recruits human participants to complete various tasks, either active or passive. In terms of task allocation and scheduling, MCS needs to consider multi-dimensional diversity, including task type, location, time, and participant preferences and activity. Cai et al. designed an auction model centered on task publishers and mobile users, allocating tasks through a bidding mechanism to optimize resource utilization. Wang et al. leveraged the spatiotemporal activity of workers and proposed a method based on an allocation graph to minimize worker detour costs and enhance system robustness. Zhang et al. proposed a multi-task allocation method based on activity prediction, using a fuzzy logic system to predict user historical trajectories and improve task completion efficiency. Traditional optimization algorithms such as the KM algorithm and the density-aware greedy algorithm have also been used in MCS task allocation. Some studies have proposed task-worker matching algorithms based on congestion game theory or stable matching theory to optimize task allocation. In recent years, an increasing number of researchers have attempted to use deep learning-based methods to solve complex MCS task allocation problems. For example, Zhu et al. designed a crowdsourcing-assisted delivery system that uses a hand-perceived reinforcement learning model to learn market recruitment challenges and formulate recruitment plans.

[0003] Patent publication number CN118488056A proposes a TS-TDDS-based parallel task allocation method for mobile crowd sensing. It uses a two-stage twin delayed deep stochastic policy gradient TS-TDDS to allocate parallel tasks for the MCS system. Patent publication number CN115629885A proposes a multi-source task allocation method and system for crowd sensing based on local budget sharing. By adopting the concept of aggregation, similar and closely located sensing tasks are aggregated and made to share a budget, achieving the task allocation goals of maximizing task completion quality and minimizing total travel distance.

[0004] With the advancement of IoT and communication technologies, research on drone-assisted mobile crowdsourcing has made significant progress, greatly expanding its application scenarios and implementation methods. Drones' aerial mobility and extensive sensing capabilities play a key role in data collection tasks. Bai et al. proposed a drone-supported intelligent truth discovery scheme to provide low-cost communication for real-world data collection in MCS. Qi et al. proposed a drone-assisted data collection framework by jointly optimizing communication costs and drone trajectories through an iterative algorithm based on continuous convex approximation (SCA). Zhao et al. designed a tensor-based trusted incentive mechanism (TensorBC) to enhance space-air-ground integrated vehicle crowdsourcing. Xie et al. proposed a blockchain-based crowdsourcing incentive framework (BCFR) that combines reputation incentives and blockchain technology to select drones. Drones, as a means of transportation, can also be used in collaborative delivery applications, such as the approximate optimal algorithm proposed by Xiang et al. based on equivalent objective function construction and alternating iteration techniques.

[0005] While various research teams have achieved success in conducting MCS studies using either drones or human participants alone, these approaches have their limitations. Considering the complementary capabilities of humans and drones, some research has focused on human-robot teaming. Gao et al. proposed a drone-assisted multi-task allocation method (UMA), which uses deep reinforcement learning to schedule the trajectories and perception activities of drones and humans and arranges for drones to calibrate the human participant's data. Their subsequent work introduced the U-TAM method, which assigns tasks to human participants, prioritizing minimizing privacy leakage while maximizing perception coverage. Patent publication number CN112306654A transforms the human-robot collaborative task allocation problem into two subproblems: task allocation for the robot and task allocation for the human participant. For the robot-based task allocation, a method combining task and participant bidirectional selection with a greedy approach is used for task allocation. A state compression dynamic programming algorithm is employed to plan the order in which the robot participants execute tasks to minimize the total distance traveled. For the human-based task allocation, a dual-chromosome genetic algorithm is designed to achieve the optimization goal of minimizing the total distance traveled by the participants to complete the task. However, these methods have limited adaptability in dynamic environments and fail to fully consider the dynamic changes of tasks and participants (including workers and drones), such as the real-time updates of task priorities and the uncertainty of human participants' trajectories.

[0006] In summary, MCS has several deficiencies in task allocation. First, task priorities are fixed and cannot be dynamically adjusted based on key factors such as the amount of remaining task data, sudden task demands, or data freshness. This results in delayed responses to high-value tasks and poor data timeliness. Second, task allocation methods only consider the unidirectional attractiveness of tasks to workers, such as task rewards or urgency, but ignore the costs of accepting tasks, such as the cost of path deviation, and the workers' actual mobility capabilities, including path range and time constraints. This leads to a disconnect between task allocation and workers' actual capabilities, resulting in high matching failure rates and low execution rates. Finally, proposed solutions have significant deficiencies in dynamic adaptability, making it difficult to adapt to dynamic changes in task, drone, and worker status in real time. This leads to information isolation, delayed policy adjustments, and rigid resource allocation, resulting in duplicated tasks, long response times, and backlogs during peak periods, which seriously affect the overall efficiency and robustness of the system. Summary of the Invention

[0007] In view of this, the embodiments of the present application propose a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC (Human Agent Collaboration) and dynamic priority matching, aiming to combine the advantages of drones and humans to complete high-quality MCS tasks through human-machine collaboration, effectively improving the response capability to sudden tasks, improving the success rate and execution rate of task matching, and providing the optimal solution for human-machine collaboration through a dynamic task priority matching algorithm.

[0008] In order to achieve the above-mentioned purpose, an embodiment of the present application proposes a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching, the method comprising the following steps: based on the nonlinear supply-demand ratio, the Shapley value method in game theory, and Monte Carlo optimization, measuring the importance of all tasks to be assigned and determining the priority of each task to be assigned; performing decentralized drone task allocation for all tasks to be assigned for all distributed drones; treating all workers as participants, taking the spatial distribution of the participants' daily trajectories as the participant preference, and taking the importance of the tasks to be assigned as the task preference, thereby establishing a bilateral preference-driven incentive mechanism; performing dynamic task allocation with a Top-k buffer for all participants based on the bilateral preference-driven incentive mechanism, determining the delayed acceptance window and the tasks in the execution window of each participant; when the drone has a new task assigned to the participant, the new task is included in the participant's delayed acceptance window, and the priority of each task in the delayed acceptance window is updated based on the priority of the new task, and re-sorted.

[0009] In order to achieve the above-mentioned purpose, an embodiment of the present application also proposes an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching as described above.

[0010] In order to achieve the above objectives, an embodiment of the present application also proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching as described above.

[0011] Optionally, based on the nonlinear supply-demand ratio, the Shapley value method in game theory, and Monte Carlo optimization, the importance of all tasks to be assigned is measured and the priority of each task to be assigned is determined, including: Calculate the nonlinear supply-demand ratio based on the relative relationship between the number of workers and the number of tasks within the drone's sensing range , The calculation formula is: ; in, represents the number of workers, Indicates the number of tasks, Indicates the total amount of data, represents the amount of data that the drone can collect, represents the amount of data that a worker is able to collect, Indicates the number of workers required to collect the remaining data, with Indicates in time slots, Indicates the sensing range of the drone, When, if ,but Follow decrease and increase; The task The attraction As the The floating reward for determining PoI in a time slot is based on , combined with the Shapley value method in game theory, considering the task The marginal contribution to a subset of all possible task combinations, using the utility function To express, The calculation formula is: ; in, Indicates the AOI data density of the area within the sensing range of the drone. represents the unit area, Indicates the amount of data required; In order to reduce the time complexity, Monte Carlo optimization is used to optimize , approximate the Shapley value by random sampling, balance the accuracy and performance of the calculation by adjusting the sample size, and based on the optimized Determine the priority of each task to be assigned, and optimize The larger the value, the higher the priority.

[0012] Optionally, the spatial distribution of participants' daily trajectories is used as participant preferences, and the importance of the tasks to be assigned is used as task preferences, thereby establishing a bilateral preference-driven incentive mechanism, including: For participants , using basic rewards Express ideal fixed costs, using variable rewards Participants are encouraged to complete tasks in their region that are short of workers, have large amounts of data, or are more urgent, thereby and Establish participant preferences, which are expressed by the formula: ; ; in, is a preset constant, Indicates the importance of the current point of interest. express and The distance between Indicates established participant preferences; For tasks , being collected and executed as soon as possible is the most important preference, let the participant’s trajectory be , A set of trajectory points Composed of Get three consecutive trajectory points on the participant's shortest working path 、 and ,based on 、 and , to extract the mobility trends of participants’ daily mobility patterns and calculate and The distance between them is used to reduce the extra distance participants need to walk to complete the task, thereby establishing task preference. The task preference is expressed by the formula: ; in, express and The distance between represents the established task preference, A small positive number is preset; Introducing the barrier coefficient Quantify the difficulty of a drone encountering collision risk while moving, The larger it is, the more likely the drone will encounter a collision. The larger the task, the more priority participants should be given to complete, based on 、 and Establish a bilateral preference-driven incentive mechanism, which is expressed by the formula: ; in, express The maximum value in express The maximum value in is the preset weight coefficient, Represents the established bilateral preference-driven incentive mechanism.

[0013] Optionally, dynamic task allocation with a Top-k buffer is performed for all participants based on a bilateral preference-driven incentive mechanism to determine the tasks within each participant's delayed acceptance window and execution window, including: Based on the tasks assigned to each drone and the set of candidate participants, a task allocation table is constructed and sorted; Adaptive synchronization of information between drones and adaptive synchronization of information within drones based on a bilateral preference-driven incentive mechanism; The drone submits its assigned tasks in each time slot and assigns them to the corresponding participants. Each participant maintains a priority double-ended queue. Tasks are sorted according to priority in the priority double-ended queue. The priority double-ended queue is divided into two parts: a delayed acceptance window and an execution window. The tasks in the execution window are the top k tasks.

[0014] Optionally, based on the tasks assigned to each UAV and the set of candidate participants, a task allocation table is constructed and sorted, including: Each drone will assign its corresponding Stored in table middle, It is expressed by the formula: ; in, Represents the set of candidate participants, points of interest Index used as The primary key, Each worker in satisfies the distance constraint, time constraint, profit constraint and task constraint; Distance constraints are specific to participants Default denial exceeds Mission , that is, requiring ; The time constraint is specifically the remaining effective time that can support participants Going around to collect data, i.e. ; Profit constraints are specifically for participants Execute the task The profit obtained is greater than 0, that is ; Task constraints are specifically tasks in the current time slot Corresponding points of interest The amount of data is not 0.

[0015] Optionally, adaptive synchronization of information between drones is performed, including: Given a corresponding point of interest Mission , and two drones and , For drones Maintaining the task assignment table, For drones Maintain the task allocation table in the time slot , if present 、 and , then the adaptive synchronization of information between UAVs is performed, and The ordered merging of and ; Adaptive synchronization of information within the UAV is performed based on a bilateral preference-driven incentive mechanism, including: By controlling the batch size, resources are ensured to be evenly distributed among different participants. When the number of tasks of a participant reaches the limit, the next priority participant is automatically selected to take over the task. The strategy is dynamically adjusted based on the participant's preference for the task. to ensure that tasks are assigned to the most appropriate participants.

[0016] Optionally, the UAV submits its assigned tasks in each time slot and assigns them to the corresponding participants. Each participant maintains a priority double-ended queue. Tasks are sorted according to their priorities in the priority double-ended queue. The priority double-ended queue is divided into two parts: a delayed acceptance window and an execution window. The tasks in the execution window are the top k tasks, including: Tasks are sorted in a priority double-ended queue from highest to lowest priority and dynamically packaged into batches based on the size of the execution window for batch processing. After a participant finishes processing a batch in the execution window, it will accept the next batch of tasks in the delayed acceptance window and send them to the execution window at the same time; The execution window adopts the Top-k strategy. Participants only focus resources on processing the top k tasks. The size of the execution window k is dynamically adjusted according to the time constraints of the working nodes and the task arrival rate. Its calculation formula is expressed as: ; ; in, Priority double-ended queue The capacity size, express The amount of tasks in is the basic Top-k value, indicating the number of executable tasks, are the preset hyperparameters.

[0017] Optionally, when a new task is assigned to a participant by the drone, the new task is included in the participant's delayed acceptance window, and the priorities of the tasks in the delayed acceptance window are updated based on the priority of the new task, and the tasks are reordered, including: When a new task is assigned to a participant by a drone, the new task is included in the participant's delayed acceptance window, and the priority of the new task is compared with the priorities of the tasks in the delayed acceptance window. If the priority of the new task is greater than the priorities of the tasks in the delayed acceptance window and the delayed acceptance window is not full, the priorities of the tasks in the delayed acceptance window are updated and reordered together with the new task. If the priority of the new task is greater than the priority of each task in the delayed acceptance window and the delayed acceptance window is full, the task with the lowest priority in the delayed acceptance window will be removed to accommodate the new task, and the task priority and batch allocation will be dynamically adjusted.

[0018] The present application proposes a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching, which has at least the following advantages and beneficial effects.

[0019] First, data timeliness is improved. This application establishes a dynamic task importance assessment mechanism to monitor indicators such as the remaining task data, sudden task requirements, and data freshness in real time. It dynamically adjusts task priorities to ensure that high-value tasks are assigned priority in a timely manner, thereby improving data timeliness and the system's responsiveness to sudden tasks.

[0020] Second, it improves task matching efficiency. This application considers both the attractiveness of tasks to workers and the cost of accepting them, building a two-way preference model between tasks and workers. On this basis, it dynamically detects workers' mobility constraints, including path range and time limits, to ensure that task assignments are consistent with workers' actual capabilities while also meeting the urgency and reward requirements of the task, thereby improving the success and execution rates of task matching.

[0021] Third, the global allocation strategy is optimized. This application introduces an adaptive synchronization mechanism and a dynamic resource scheduling algorithm, enabling the system to perceive dynamic changes in task, drone, and worker status in real time and adjust the task allocation strategy accordingly. By optimizing the information synchronization and resource allocation process, problems such as information isolation, delayed policy adjustments, and rigid resource allocation are resolved, improving the overall efficiency and robustness of the system and ensuring high-quality task allocation and data collection in a dynamic and changing environment.

[0022] Fourth, it combines the advantages of cost and scenario adaptability. This application uses dynamic incentive mechanisms and resource scheduling strategies to reduce manpower and operation and maintenance costs while adapting to complex scenarios such as sudden surges in tasks. This application balances task demands and resource inputs through floating reward adjustments and path planning optimization, ensuring task coverage while reducing ineffective resource consumption. This is particularly suitable for dynamic and changing scenarios such as smart cities and emergency monitoring.

[0023] Fifth, it improves the flexibility of task execution. This application uses a dynamic batching strategy and priority update mechanism to dynamically adjust the execution window size based on the worker's time constraints and task arrival rate, and updates task priorities in real time to ensure that the most important and urgent tasks are executed first. This makes task execution more flexible, better able to adapt to dynamic changes in tasks, and improves the efficiency and quality of task execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the related technologies, the following is a brief introduction to the drawings required for use in the embodiments of the present application or the description of the related technologies. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0025] Figure 1 This is a flowchart of a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching provided in one embodiment of the present application; Figure 2 This is a framework diagram of a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching provided in one embodiment of the present application; Figure 3 A flowchart of an embodiment of the present application provides a method for dynamically allocating tasks with a Top-k buffer based on a bilateral preference-driven incentive mechanism, and determining tasks within each participant's delayed acceptance window and execution window. Figure 4 This is a schematic diagram of a data synchronization and internal allocation adjustment process between distributed drones, provided in one embodiment of the present application; Figure 5 is a schematic diagram of a dynamic batch processing and priority update process with Top-k buffer limitations provided in one embodiment of the present application; Figure 6 It is a structural diagram of an electronic device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. In the various embodiments of the present application, many technical details are proposed to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in the present application can be implemented. The division of the following embodiments is only for the convenience of description and should not constitute any limitation on the specific implementation of the present application. The various embodiments can be combined with each other and referenced to each other under the premise of no contradiction.

[0027] Traditional MCS technology faces the following problems.

[0028] First, the static nature of task priorities in human-machine collaboration leads to low data timeliness. Traditional methods use a fixed priority allocation strategy, which prevents task priorities from being dynamically adjusted after initialization based on remaining task data, unexpected task demands, or changes in data freshness. When high-value tasks arise, delayed response times due to delayed priority updates significantly degrade data timeliness, preventing high-value tasks from being executed in a timely manner.

[0029] Second, the lack of unidirectional preference modeling and the neglect of mobility constraints lead to inefficient task matching. Traditional approaches only consider the unidirectional attractiveness of tasks to workers (such as task rewards or urgency), ignoring the costs workers in accepting tasks (such as the cost of path deviation) and their actual mobility capabilities (such as path range and time constraints). For example, high-reward tasks are assigned to workers who cannot afford the cost of path deviation, resulting in high matching failure rates. Furthermore, when tasks exceed workers' mobility capabilities, execution rates drop significantly.

[0030] Third, the lack of dynamic adaptability in human-machine collaboration leads to ineffective global allocation strategies. Traditional technical solutions are unable to adapt to the dynamic changes in tasks, drones, and worker status in real time. This manifests as information isolation (duplication of tasks), delayed policy adjustments (long response times), and rigid resource allocation (backlogs of tasks during peak periods). For example, task allocation in a centralized architecture relies on global information synchronization, and communication delays lead to slow policy updates.

[0031] In order to solve these technical problems, an embodiment of the present application proposes a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching, which is applied to a server. The following is a specific description of the implementation details of the spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching proposed in this embodiment. The following content is only the implementation details provided for the convenience of understanding and is not necessary for the implementation of this solution.

[0032] The specific process of the spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching proposed in this embodiment can be as follows: Figure 1 As shown, its specific framework can be as follows Figure 2 As shown in the figure, the framework mainly consists of three parts: task importance assessment, decentralized UAV task allocation, and human-machine collaborative task allocation. The method integrates these components to create a dynamic, hierarchical system for human-machine collaborative crowdsourcing to optimize data freshness, energy efficiency, and task coverage while adapting to changing environmental conditions and the participants involved. The method specifically includes the following steps: Step 101 : Based on the nonlinear supply-demand ratio, the Shapley value method in game theory, and Monte Carlo optimization, the importance of all tasks to be assigned is measured, and the priority of each task to be assigned is determined.

[0033] In specific implementations, task importance assessment is an essential part of decentralized drone scheduling and human-machine task allocation. The server combines nonlinear supply-demand ratios, the Shapley value method in game theory, and Monte Carlo optimization to measure task importance and determine the priority of each task to be assigned, thereby ensuring that areas with more data but higher information age (AoI areas) can receive more attention.

[0034] In one example, a nonlinear supply-demand ratio Reflects the sensing range of the drone The server can calculate the nonlinear supply-demand ratio based on the relative relationship between the number of workers and the number of tasks within the sensing range of the drone. .

[0035] In one example, The calculation formula is: ; in, represents the number of workers, Indicates the number of tasks, Indicates the total amount of data, represents the amount of data that the drone can collect, represents the amount of data that a worker is able to collect, Indicates the number of workers required to collect the remaining data, with Indicates in time slots, Indicates the sensing range of the drone. When, if ,but Follow decrease and increase. The value range is 0 to 1. If the number of workers required is not less than the number of tasks or , indicating that there are enough workers or there are no tasks, and the supply and demand are balanced. .when When , it means there are no workers available, Get the maximum value.

[0036] It should be noted that the nonlinear supply and demand ratio It is a process of nonlinear adjustment and smooth transition.

[0037] In one example, a nonlinear supply-demand ratio is achieved After the calculation, the server needs to The attraction As the The floating reward for determining PoI in a time slot is based on , combined with the Shapley value method in game theory, considering the task The marginal contribution to a subset of all possible task combinations, using the utility function This is because some rarely perceived PoIs are more important, and the attractiveness of PoIs plays an important guiding role in the incentive mechanism of drones and participants.

[0038] In one example, The calculation formula is expressed as: ; in, Indicates a task In all possible combinations of tasks.

[0039] In one example, the utility function The calculation formula is: ; in, Indicates the AOI data density of the area within the sensing range of the drone, reflecting the intensity of the demand for information update per unit area. represents the unit area, Indicates the data demand.

[0040] In one example, to reduce time complexity, the server uses Monte Carlo optimization to optimize , approximate the Shapley value by random sampling, balance the accuracy and performance of the calculation by adjusting the sample size, and based on the optimized Determine the priority of each task to be assigned, and optimize The larger the value, the higher the priority. The cost is calculated for each sample, and the time complexity of this method is .

[0041] In one example, priority calculation can also be achieved by collecting user satisfaction after task execution, or data quality feedback (such as data integrity score), thereby dynamically correcting the task priority weight.

[0042] In one example, a supervised learning model can also be used to predict the potential value of a task.

[0043] Step 102 : performing decentralized drone task allocation for all tasks to be assigned to all distributed drones.

[0044] In practice, after completing the importance assessment, the server needs to distribute all pending tasks to all distributed drones. This process involves multiple drones interacting with the environment, determining whether the experience array exceeds H entries, and then performing LSTM+N-step learning, followed by MADDPG path planning.

[0045] In step 103, all workers are regarded as participants, the spatial distribution of the participants' daily trajectories is used as the participant preference, and the importance of the tasks to be assigned is used as the task preference, thereby establishing a bilateral preference-driven incentive mechanism.

[0046] In a specific implementation, after completing decentralized drone task assignments, the server needs to treat all workers as participants, using the spatial distribution of participants' daily trajectories as participant preferences and the importance of the assigned tasks as task preferences, thereby establishing a bilateral preference-driven incentive mechanism. This mechanism aims to effectively incentivize workers to complete tasks with high quality. This mechanism considers the spatial distribution of participants' daily trajectories and the importance of tasks. Regarding participant preferences, participants generally prefer to complete tasks without deviating from their original path. Regarding task preferences, higher-priority tasks should be completed first, and all tasks should be completed with the lowest cost within time and resource constraints.

[0047] In one example, for participants , if the requested amount of data is collected under the budget constraint, the server uses the basic reward represents the ideal fixed cost ( is a constant), using a floating reward Participants are encouraged to complete tasks in their region that are short of workers, have large amounts of data, or are more urgent, thereby and Establish participant preferences.

[0048] In one example, participant preferences are expressed as: ; ; in, is a preset constant, Indicates the importance of the current point of interest. express and The distance between Indicates the established participant preferences.

[0049] In one example, the server can introduce a third-party resource leasing platform (such as a shared drone network) to temporarily call external resources according to task requirements, instead of relying solely on the resource pool expansion method of the system's internal executors.

[0050] In one example, the server can dynamically adjust the reward distribution weight based on the executor's historical task completion rate, data quality and other indicators, replacing the floating incentive model based on the real-time task attractiveness.

[0051] In one example, for the task , being collected and executed as soon as possible is the most important preference. The server considers taking charge of those areas where the drone cannot work effectively in fixed mode without affecting the daily trajectory of human participants. Suppose the trajectory of the participant is , A set of trajectory points Composed of Get three consecutive trajectory points on the participant's shortest working path 、 and ,based on 、 and , to extract the mobility trends of participants’ daily mobility patterns and calculate and The distance between the two groups was reduced to reduce the extra distance participants needed to walk to complete the task, more accurately assessing participants' task efficiency in performing the MSC, and thus establishing task preference.

[0052] In one example, task preference is expressed as: ; in, express and The distance between represents the established task preference, A small positive number is preset.

[0053] In one example, the server needs to introduce a barrier factor Quantify the difficulty of a drone encountering collision risk while moving, , The larger the value is, the greater the possibility that the UAV will encounter a collision, which may cause the path planning process to fall into a local optimal solution or generate an invalid path. The larger the task, the more priority participants should be given to complete. It is an important risk assessment indicator in drone navigation. The server can be based on 、 and Establish an incentive mechanism driven by bilateral preferences.

[0054] In one example, given a task and participants , is a set of worker and task pairs, which is the result of task allocation. The bilateral preference of is defined as the sum of the weighted satisfaction of the task and the worker. The overall satisfaction (two-sided preference-driven incentive mechanism) is calculated as follows: ; in, express The maximum value in express The maximum value in is the preset weight coefficient, Represents the established bilateral preference-driven incentive mechanism.

[0055] It should be noted that if the capacity or distance limit is exceeded, is an empty set, indicating that the allocation at this time is invalid, that is, and .

[0056] In one example, the server can introduce a game negotiation process (such as a bargaining game or auction mechanism) between the task and the executor, allowing both parties to dynamically adjust preference parameters and reach consensus, replacing the static two-way preference scoring rules.

[0057] In step 104 , dynamic task allocation with a Top-k buffer is performed for all participants based on a bilateral preference-driven incentive mechanism to determine the tasks within the delayed acceptance window and execution window of each participant.

[0058] In specific implementations, tasks and participants change dynamically over time. On the one hand, data from points of interest is continuously generated, and task priorities are constantly evolving. On the other hand, drone energy and storage, as well as human mobility, also evolve over time. Therefore, several issues should be considered in dynamic task allocation. First, distributed drones' information needs to be mutually adaptive and synchronized to ensure a reasonable task allocation strategy in a global environment. Second, the optimal allocation strategy needs to be dynamically adjusted to accommodate real-time changes in tasks, drones, and workers. Finally, the uneven distribution of tasks in a task sequence has a significant impact on task allocation strategies and completion quality. Therefore, the server needs to implement dynamic task allocation with a top-k buffer for all participants (including drones) based on a bilateral preference-driven incentive mechanism to determine the tasks within each participant's delayed acceptance window and execution window.

[0059] In one example, the server performs dynamic task allocation with a Top-k buffer based on a bilateral preference-driven incentive mechanism, determines the delayed acceptance window and the tasks in the execution window of each participant, and can be done by Figure 3 The implementation of each sub-step shown includes: Step 1041 : construct and sort a task allocation table based on the tasks assigned to each drone and the set of candidate participants.

[0060] Specifically, each drone will assign its corresponding Stored in table (task assignment table) middle, It is expressed by the formula: .in, Represents the set of all candidate participants, points of interest Index used as The primary key, Each worker in satisfies the distance constraint, time constraint, profit constraint and task constraint.

[0061] Distance constraints are specific to participants Default denial exceeds Mission , that is, requiring ; The time constraint is specifically the remaining effective time that can support participants Going around to collect data, i.e. ; Profit constraints are specifically for participants Execute the task The profit obtained is greater than 0, that is ; Task constraints are specifically tasks in the current time slot Corresponding points of interest The amount of data is not 0.

[0062] In order to check whether the distance constraint and time constraint are met, it is necessary to calculate and Therefore, it is necessary to find participants Daily shortest path trajectory Optimal offset point and detour distance To get participants Task Preference ,calculate Each task in scope The attraction According to the above constraints, each task is obtained Candidate set Calculate the overall bilateral preference .right according to Sort in non-increasing order. in In the current time slot , build each drone of .

[0063] Step 1042 : Adaptively synchronize information between drones, and adaptively synchronize information within drones based on a bilateral preference-driven incentive mechanism.

[0064] In one example, the server first performs adaptive synchronization of information between drones, given a corresponding point of interest Mission , and two drones and , For drones Maintaining the task assignment table, For drones Maintain the task allocation table in the time slot , if present 、 and , then adaptive synchronization of information between drones is required. Current point of interest It is fixed and in different tables Corresponding 、 All of them are participants who meet the distance constraint, time constraint and profit constraint. 、 Both are based on bilateral preferences An ordered list in reverse order, in this case you need to and The ordered merging of and .

[0065] Next, the adaptive synchronization of information within the UAV is performed based on the bilateral preference-driven incentive mechanism. In the execution process, there is a positive scale economy effect. Under the constraint of meeting the task requirements, The more participants there are, the higher the task completion efficiency will be. Set of tasks to be completed The following limitations and adjustments apply. Identify tasks that meet the requirements Participants who completed the task within a single time interval and met the time constraints of their daily trajectory. However, when considering participants Execute a set of tasks at multiple consecutive time intervals The accumulated task completion time may exceed the time limit. Cancel tasks from the bottom of the table one by one ,until To prevent a single participant from affecting the quality of task completion due to task overload, it is necessary to Submitted by participants The server controls the batch size to ensure that resources are evenly distributed among different participants. When the number of tasks of a participant reaches the limit, the server automatically selects the next priority participant to take over the task. The server dynamically adjusts the strategy based on the participant's preference for the task. to ensure that tasks are assigned to the most appropriate participants.

[0066] In one example, the data synchronization and internal allocation adjustment process between distributed drones is as follows: Figure 4 shown.

[0067] In one example, the server can adopt a decentralized task scheduling architecture (such as blockchain consensus mechanism or edge node collaboration) to replace the synchronization mechanism based on a centralized coordinator, and achieve global task allocation optimization through local information exchange between nodes.

[0068] In step 1043, the UAV submits its assigned task in each time slot and assigns it to the corresponding participant. Each participant maintains a priority double-ended queue. Tasks are sorted according to priority in the priority double-ended queue. The priority double-ended queue is divided into a delayed acceptance window and an execution window. The tasks in the execution window are the top k tasks.

[0069] In the specific implementation, Figure 5 As shown, the drone is Submit their task assignments For the corresponding participants, each participant maintains a capacity of Priority double-ended queue The queue is divided into a delayed acceptance window and an execution window. Tasks are sorted in descending order of priority in the priority double-ended queue and dynamically packaged into batches based on the size of the execution window for batch processing. After a participant completes processing a batch in the execution window, it accepts the next batch of tasks in the delayed acceptance window and sends them to the execution window. This operation extends the time that tasks remain in the delayed acceptance window, providing greater flexibility for dynamic updates to subsequent time slots.

[0070] Subsequently, the execution window adopts the Top-k strategy, where participants only concentrate resources to process the top k tasks, and the size of the execution window k is dynamically adjusted according to the time constraints of the working nodes and the task arrival rate.

[0071] Considering the uneven distribution of tasks over time, the number of tasks in the task queue also affects Top-k. When too many tasks arrive in a given time period, the number of frozen tasks should be reduced. This facilitates subsequent priority updates and prevents participants from executing too many tasks at a given location. Conversely, when fewer tasks are available in a given time period, the distance between the drone platform and the participant may prevent them from being assigned. In this case, the k value should be increased to ensure that tasks are executed.

[0072] The calculation formula of the execution window size k is expressed as: ; ; in, Priority double-ended queue The capacity size, express The amount of tasks in is the basic Top-k value, indicating the number of executable tasks, are the preset hyperparameters.

[0073] Tasks within the execution window must satisfy all constraints and are unaffected by environmental changes. They are frozen, ensuring they will be executed. The algorithm also accounts for the dynamic changes in task importance and allocation results over time. Tasks not in the Top-k range are delayed and can be updated before being sent to the next batch for execution. However, in later time slots, participants may reject previous tasks in order to accept new ones when new tasks arrive.

[0074] In one example, the server can dynamically trigger batch processing windows based on the arrival rate of real-time tasks (such as starting batch allocation when the task queue length exceeds a threshold), replacing fixed time windows or load-adaptive dynamic batch processing rules to reduce scheduling delays.

[0075] In one example, the server can adopt a hybrid architecture of "centralized coordination + edge autonomy", allowing some high-priority tasks to be directly assigned by the central node, and the remaining tasks to be scheduled by the edge nodes based on local policies, instead of relying entirely on the centralized coordinator.

[0076] In one example, the server breaks down a complex task into multiple subtasks, each adapted to the capabilities and resource status of different executors (such as drones handling high-altitude data collection and ground equipment handling local data processing), replacing a unified task allocation logic.

[0077] Step 105: When the drone has a new task assigned to a participant, the new task is included in the participant's delayed acceptance window, and the priorities of the tasks in the delayed acceptance window are updated based on the priority of the new task and reordered.

[0078] In a specific implementation, when a drone has a new task assigned to a participant, the new task is included in the participant's delayed acceptance window, and the priorities of each task in the delayed acceptance window are updated based on the priority of the new task and reordered.

[0079] In one example, when a drone has a new task assigned to a participant, the new task is included in the participant's delayed acceptance window, and the priority of the new task is compared with the priority of each task in the delayed acceptance window. If the priority of the new task is greater than the priority of each task in the delayed acceptance window and the delayed acceptance window is not full, the priority of each task in the delayed acceptance window is updated and reordered together with the new task.

[0080] If the priority of the new task is greater than the priority of each task in the delayed acceptance window and the delayed acceptance window is full, the task with the lowest priority in the delayed acceptance window will be removed to accommodate the new task, and the task priority and batch allocation will be dynamically adjusted.

[0081] When resources are limited, participants may reject low-priority tasks to make room for new tasks.

[0082] The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching proposed in this embodiment has at least the following advantages and beneficial effects.

[0083] First, data timeliness is improved. This embodiment establishes a dynamic task importance assessment mechanism to monitor indicators such as the remaining task data, sudden task requirements, and data freshness in real time. It dynamically adjusts task priorities to ensure that high-value tasks are promptly prioritized, thereby improving data timeliness and the system's responsiveness to sudden tasks.

[0084] Second, it improves task matching efficiency. This embodiment considers both the attractiveness of a task to a worker and the cost of accepting the task, building a two-way preference model between tasks and workers. Based on this, it dynamically detects workers' mobility constraints, including path range and time limits, to ensure that task assignments are consistent with the actual capabilities of human workers while also meeting the urgency and reward requirements of the task, thereby improving the success rate and execution rate of task matching.

[0085] Third, the global allocation strategy is optimized. This embodiment introduces an adaptive synchronization mechanism and a dynamic resource scheduling algorithm, enabling the system to perceive dynamic changes in task, drone, and worker status in real time and adjust task allocation strategies accordingly. By optimizing information synchronization and resource allocation processes, it addresses issues such as information isolation, delayed policy adjustments, and rigid resource allocation, improving overall system efficiency and robustness and ensuring high-quality task allocation and data collection in a dynamic and changing environment.

[0086] Fourth, it combines the advantages of cost and scenario adaptability. This embodiment uses dynamic incentive mechanisms and resource scheduling strategies to reduce manpower and operation and maintenance costs while adapting to complex scenarios such as sudden surges in tasks. This embodiment balances task requirements and resource inputs through floating reward adjustments and path planning optimization, ensuring task coverage while reducing ineffective resource consumption. This is particularly suitable for dynamic and changing scenarios such as smart cities and emergency monitoring.

[0087] Fifth, it effectively improves the flexibility of task execution. This embodiment utilizes a dynamic batching strategy and priority update mechanism. This dynamically adjusts the execution window size based on the worker's time constraints and the rate at which tasks arrive, and updates task priorities in real time, ensuring that the most important and urgent tasks are prioritized. This makes task execution more flexible, better adapting to dynamic changes in tasks, and improving both efficiency and quality.

[0088] The step division of the above various methods is only for the purpose of clear description. During implementation, they can be combined into one step or some steps can be split and decomposed into multiple steps. As long as they include the same logical relationship, they are all within the scope of protection of this application; adding insignificant modifications or introducing insignificant designs to the algorithm or process without changing the core design of the algorithm and process are all within the scope of protection of this application.

[0089] In one example, we conducted simulation experiments to validate the effectiveness of the proposed method. The experiments were conducted on a resource-constrained device equipped with an Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz and an NVIDIA GeForce GTX 1660 SUPER graphics card. The software environment was based on Python 3.7.0 and the TensorFlow 1.15.0 framework.

[0090] 250 points of interest are randomly deployed in a 16 x 16 working area, and the data volume is Each point of interest corresponds to a task At the initial moment, the AOI of all tasks is 0, that is, Each point of interest is initialized with an obstacle coefficient according to its geographical location . The larger the value, the higher the risk of collision. There are 5 charging stations in the area to charge the drones. At the beginning of the experiment, each drone carries 50 units of energy and releases it from the center of the map. When performing data collection tasks, set the energy consumption per unit of movement to , energy consumption per unit data collected Set the drone's sensing range and charging station's charging range If the drone Collision with obstacles, set the corresponding penalty coefficient The performance of the method is verified through iterations of the training phase, and the results are as follows.

[0091] First, in terms of improving data collection rates, compared to traditional algorithms (such as random allocation and greedy algorithms) and algorithms using only drones, this application achieved a significantly higher data collection rate, reaching 94.5%. Traditional algorithms had lower data collection rates, such as 36.7% for random algorithms and 31.5% for greedy algorithms. This demonstrates that this application can collect data more efficiently, improving data integrity and coverage.

[0092] Second, in terms of improving geographic coverage fairness, experimental results show that this application achieves a geographic coverage fairness index of 95.8%, while other algorithms have relatively lower geographic coverage fairness indicators, such as 68.1% for the TRAP method and 75.3% for the UMA method. This demonstrates that this application can better balance task allocation across different regions, avoiding excessive concentration of tasks in certain areas while neglecting tasks in other areas.

[0093] Third, in terms of reducing energy consumption, this application achieved an energy consumption ratio of 0.295, significantly lower than other algorithms, such as TRAP (0.405) and UMA (0.475). This demonstrates that this application can more efficiently utilize energy and reduce system operating costs while ensuring data collection quality.

[0094] Fourth, in terms of improving the average age of information in the system, the average AoI of this application is 13.648, which is much lower than other algorithms, such as the random algorithm with an average AoI of 26.321 and the greedy algorithm with an average AoI of 32.874. This shows that this application can update data more promptly, improving the freshness and timeliness of data.

[0095] Fifth, in terms of improving system efficiency, this application comprehensively considers multiple indicators, including data collection ratio, geographic coverage fairness, energy consumption ratio, and system average AoI. By optimizing the task allocation strategy, it achieves improved overall system performance. In experiments, this application achieved a system efficiency of 3.075, significantly higher than other algorithms, such as TRAP's system efficiency of 1.142 and UMA's system efficiency of 1.294. This demonstrates that this application can improve overall system efficiency and resource utilization while maintaining high data collection quality.

[0096] Sixth, in terms of enhancing the system's scalability and robustness, this application demonstrated good adaptability and stability across experimental setups of varying scales (e.g., varying the number of drones and workers, and the number and distribution of tasks). As the number of drones increased, the application's data collection ratio and geographic coverage fairness gradually improved and stabilized, while system efficiency decreased slightly but remained within a reasonable range. Furthermore, this application maintained good performance despite complex situations such as an increase in the number of tasks and an uneven distribution of tasks, demonstrating its strong scalability and robustness, adapting to practical application scenarios of varying scale and complexity.

[0097] In one embodiment, the present application can be applied to smart city human-machine collaborative traffic diversion, that is, it can be applied to urban traffic diversion scenarios to achieve rapid response to emergencies through human-machine collaboration. Traffic command center personnel view road condition data (such as accident points and congestion levels) in real time through the system. The system automatically marks high-priority areas and generates disposal tasks. The drone is controlled to fly to the accident scene to collect high-definition images and transmit them back to the command center's large screen at the same time. At the same time, the system pushes navigation paths and diversion instructions to the nearest traffic police handheld terminal, guiding the traffic police and drones to work together. The drone monitors traffic flow changes in real time, and the traffic police manually adjust temporary roadblocks and traffic lights based on dynamic data to achieve precise diversion through human-machine collaboration.

[0098] In one embodiment, this application can be applied to human-machine collaborative life-saving search and rescue scenarios during disaster relief. At an earthquake disaster site, rescuers receive system-assigned tasks via mobile terminals and collaborate with drones and search and rescue robots to detect life. The drone swarm performs thermal imaging scans of the debris area, automatically marking coordinates upon detecting signs of life and notifying ground rescue teams. Search and rescue robots enter confined spaces to locate trapped victims and simultaneously push structural risk data (such as wall cracks) to rescuers' helmet displays in real time. Rescuers develop excavation plans based on the robot's feedback and control the robot through voice commands to clear obstacles, forming an efficient closed loop of "machine detection, human decision-making, and collaborative rescue."

[0099] In one embodiment, this application can be applied to human-machine collaborative environmental law enforcement and evidence collection scenarios. To address the challenge of noise pollution control, urban management law enforcement officers use smart body cameras to receive system-driven information on noise-exceeding locations and collaborate with drones to collect evidence. The drone automatically cruises over the noise source, recording video evidence and displaying a real-time decibel fluctuation curve on the body camera screen. Law enforcement officers quickly arrive at the scene based on the drone's location and, using voice commands, control the drone to capture multiple angles of the construction site in question. The system automatically uploads the evidence chain (images and data) to the law enforcement platform, enabling precise law enforcement.

[0100] In one embodiment, the present application can be applied to the scenario of human-machine collaborative sorting in warehousing and logistics. In e-commerce warehousing, sorters and Automated Guided Vehicle (AGV) robots collaborate to complete order processing. The system dynamically dispatches AGVs to transport shelves to the sorting platform based on order hot zones, and sorters need to scan the goods and confirm the quantity through handheld terminals. If the AGV detects that the goods on the shelves do not match the system records (such as out of stock), it immediately triggers an alarm and prompts the sorter to manually check. For orders for fragile items, sorters can mark special packaging requirements, and the AGV automatically adjusts the transportation route to the reinforced packaging area, realizing a flexible collaborative mode of "machine handling, manual sorting, and collaborative quality inspection."

[0101] The above embodiments all focus on human-machine interaction and collaborative decision-making, highlighting the core role of humans in task planning, exception handling, and final decision-making. At the same time, they emphasize the auxiliary capabilities of machines in data collection, automatic execution, and real-time feedback, reflecting the collaborative innovation value of "machines amplifying human power and humans dominating key links."

[0102] Another embodiment of the present application provides an electronic device, the specific structure of which is as follows: Figure 6 As shown, it includes: at least one processor 201; and a memory 202 communicatively connected to the at least one processor 201; wherein the memory 202 stores instructions that can be executed by the at least one processor 201, and the instructions are executed by the at least one processor 201 to enable the at least one processor 201 to execute a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching as described in the above method embodiment.

[0103] The memory and processor can be connected using a bus, which can include any number of interconnected buses and bridges. The bus connects various circuits of one or more processors and memories. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits. These are all well known in the art and will not be described further in this article. The bus interface is responsible for providing an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices on a transmission medium. Data processed by the processor is transmitted over a wireless medium via an antenna. Furthermore, the antenna also receives data and transmits it to the processor.

[0104] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory can be used to store data used by the processor when performing operations.

[0105] Another embodiment of the present application proposes a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it can implement a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching as described in the above method embodiment.

[0106] That is, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a program. The program is stored in a storage medium and includes a number of instructions for causing a device (such as a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes: a USB flash drive, a mobile hard drive, ROM (Read-Only Memory), RAM (Random Access Memory), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0107] Those skilled in the art will appreciate that the above embodiments are specific embodiments for implementing the present application, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present application.

Claims

1. A spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching, suitable for mobile crowd intelligence perception, characterized by: The method comprises: Based on the nonlinear supply-demand ratio, the Shapley value method in game theory, and Monte Carlo optimization, the importance of all tasks to be assigned is measured and the priority of each task is determined; For all distributed drones, decentralized drone task allocation is performed for all tasks to be assigned; All workers are considered as participants, the spatial distribution of participants' daily trajectories is used as participant preferences, and the importance of the tasks to be assigned is used as task preferences, thus establishing a bilateral preference-driven incentive mechanism. For all participants, a bilateral preference-driven incentive mechanism is used to dynamically allocate tasks with a Top-k buffer, determining the tasks within each participant's delayed acceptance window and execution window. When the UAV has a new task assigned to a participant, the new task is included in the participant's delayed acceptance window, and the priorities of each task in the delayed acceptance window are updated based on the priority of the new task and reordered.

2. The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to claim 1, characterized in that: Based on the nonlinear supply-demand ratio, the Shapley value method in game theory, and Monte Carlo optimization, the importance of all tasks to be assigned is measured and the priority of each task to be assigned is determined, including: Calculate the nonlinear supply-demand ratio based on the relative relationship between the number of workers and the number of tasks within the drone's sensing range , The calculation formula is: ； in, represents the number of workers, Indicates the number of tasks, Indicates the total amount of data, represents the amount of data that the drone can collect, represents the amount of data that a worker is able to collect, Indicates the number of workers required to collect the remaining data, with Indicates in time slots, Indicates the sensing range of the drone, When, if ,but Follow decrease and increase; The task The attraction As the The floating reward for determining PoI in a time slot is based on , combined with the Shapley value method in game theory, considering the task The marginal contribution to a subset of all possible task combinations, using the utility function To express, The calculation formula is: ； in, Indicates the AOI data density of the area within the sensing range of the drone. represents the unit area, Indicates the amount of data required; In order to reduce the time complexity, Monte Carlo optimization is used to optimize , approximate the Shapley value by random sampling, balance the accuracy and performance of the calculation by adjusting the sample size, and based on the optimized Determine the priority of each task to be assigned, and optimize The larger the value, the higher the priority.

3. The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to claim 2, characterized in that: The spatial distribution of participants' daily trajectories is used as participant preferences, and the importance of the assigned tasks is used as task preferences. This establishes a bilateral preference-driven incentive mechanism, including: For participants , using basic rewards Express ideal fixed costs, using variable rewards Participants are encouraged to complete tasks in their region that are short of workers, have large amounts of data, or are more urgent, thereby and Establish participant preferences, which are expressed by the formula: ； ； in, is a preset constant, Indicates the importance of the current point of interest. express and The distance between Indicates established participant preferences; For tasks , being collected and executed as soon as possible is the most important preference, let the participant’s trajectory be , A set of trajectory points Composed of Get three consecutive trajectory points on the participant's shortest working path 、 and ,based on 、 and , to extract the mobility trends of participants’ daily mobility patterns and calculate and The distance between them is used to reduce the extra distance participants need to walk to complete the task, thereby establishing task preference. The task preference is expressed by the formula: ； in, express and The distance between represents the established task preference, A small positive number is preset; Introducing the barrier coefficient Quantify the difficulty of a drone encountering collision risk while moving, The larger it is, the more likely the drone will encounter a collision. The larger the task, the more priority participants should be given to complete, based on 、 and Establish a bilateral preference-driven incentive mechanism, which is expressed by the formula: ； in, express The maximum value in express The maximum value in is the preset weight coefficient, Represents the established bilateral preference-driven incentive mechanism.

4. The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to claim 3, characterized in that: For all participants, a bilateral preference-driven incentive mechanism is used to dynamically allocate tasks with a Top-k buffer, determining the tasks within each participant's delayed acceptance window and execution window, including: Based on the tasks assigned to each drone and the set of candidate participants, a task allocation table is constructed and sorted; Adaptive synchronization of information between drones and adaptive synchronization of information within drones based on a bilateral preference-driven incentive mechanism; The drone submits its assigned tasks in each time slot and assigns them to the corresponding participants. Each participant maintains a priority double-ended queue. Tasks are sorted according to priority in the priority double-ended queue. The priority double-ended queue is divided into two parts: a delayed acceptance window and an execution window. The tasks in the execution window are the top k tasks.

5. The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to claim 4, characterized in that: Based on the tasks assigned to each drone and the set of candidate participants, the task allocation table is constructed and sorted, including: Each drone will assign its corresponding Stored in table middle, It is expressed by the formula: ； in, Represents the set of candidate participants, points of interest Index used as The primary key, Each worker in satisfies the distance constraint, time constraint, profit constraint and task constraint; Distance constraints are specific to participants Default denial exceeds Mission , that is, requiring ; The time constraint is specifically the remaining effective time that can support participants Going around to collect data, i.e. ; Profit constraints are specifically for participants Execute the task The profit obtained is greater than 0, that is ; Task constraints are specifically tasks in the current time slot Corresponding points of interest The amount of data is not 0.

6. The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to claim 4, characterized in that: Adaptive synchronization of information between drones, including: Given a corresponding point of interest Mission , and two drones and , For drones Maintaining the task assignment table, For drones Maintain the task allocation table in the time slot , if present 、 and , then the adaptive synchronization of information between UAVs is performed, and The ordered merging of and ; Adaptive synchronization of information within the UAV is performed based on a bilateral preference-driven incentive mechanism, including: By controlling the batch size, resources are ensured to be evenly distributed among different participants. When the number of tasks of a participant reaches the limit, the next priority participant is automatically selected to take over the task. The strategy is dynamically adjusted based on the participant's preference for the task. to ensure that tasks are assigned to the most appropriate participants.

7. A spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to any one of claims 4 to 6, characterized in that: The drone submits its assigned tasks in each time slot and assigns them to the corresponding participants. Each participant maintains a priority double-ended queue. Tasks are sorted according to their priorities in the priority double-ended queue. The priority double-ended queue is divided into two parts: the delayed acceptance window and the execution window. The tasks in the execution window are the top k tasks, including: Tasks are sorted in a priority double-ended queue from highest to lowest priority and dynamically packaged into batches based on the size of the execution window for batch processing. After a participant finishes processing a batch in the execution window, it will accept the next batch of tasks in the delayed acceptance window and send them to the execution window at the same time; The execution window adopts the Top-k strategy. Participants only focus resources on processing the top k tasks. The size of the execution window k is dynamically adjusted according to the time constraints of the working nodes and the task arrival rate. Its calculation formula is expressed as: ； ； in, Priority double-ended queue The capacity size, express The amount of tasks in is the basic Top-k value, indicating the number of executable tasks, are the preset hyperparameters.

8. The spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching according to any one of claims 1 to 4, characterized in that: When a new task is assigned to a participant by the drone, the new task is included in the participant's delayed acceptance window. Based on the priority of the new task, the priority of each task in the delayed acceptance window is updated and reordered, including: When a new task is assigned to a participant by a drone, the new task is included in the participant's delayed acceptance window, and the priority of the new task is compared with the priorities of the tasks in the delayed acceptance window. If the priority of the new task is greater than the priorities of the tasks in the delayed acceptance window and the delayed acceptance window is not full, the priorities of the tasks in the delayed acceptance window are updated and reordered together with the new task. If the priority of the new task is greater than the priority of each task in the delayed acceptance window and the delayed acceptance window is full, the task with the lowest priority in the delayed acceptance window will be removed to accommodate the new task, and the task priority and batch allocation will be dynamically adjusted.

9. An electronic device, characterized in that: include: at least one processor; and, a memory communicatively coupled to the at least one processor; In which, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it can implement a spatiotemporal crowdsourcing task allocation method based on hierarchical HAC and dynamic priority matching as described in any one of claims 1 to 8.

Images