Task scheduling method and system based on space-time voxelization and unmanned equipment self-evaluation

Abstract

The invention discloses a task scheduling method and system based on space-time voxelization and unmanned equipment self-evaluation. The task scheduling method comprises the following steps: dynamically scheduling a voxelization unit according to the demand intensity of a task area; evaluating the unmanned equipment in real time through the multi-dimensional parameters to obtain an adaptability score; adjusting the weight coefficient of the voxel unit based on the adaptability score, and carrying out task allocation and execution; the task execution state is fed back in real time, the weight coefficient is dynamically adjusted, and the subsequent scheduling strategy is optimized. According to the method, accurate allocation of task area resources and improvement of equipment cooperation efficiency are realized through a multi-dimensional dynamic optimization mechanism, a three-dimensional voxel unitization dynamic division model is constructed based on task demand intensity, equipment real-time states and environmental factors, fine scheduling of task areas is realized, resource waste in task execution is avoided, and the task execution efficiency is improved. And it is ensured that each device can execute a proper task according to the ability of the device, and the flexibility and response speed in a dynamic environment are improved.

Description

Technical Field

[0001] This invention relates to the field of collaborative task scheduling technology for unmanned equipment, and more specifically, to a task scheduling method and system based on spatiotemporal voxelization and self-evaluation of unmanned equipment. Background Technology

[0002] With the rapid development of unmanned equipment technology, especially the widespread application of intelligent unmanned equipment such as drones and automated ground robots in various fields, how to efficiently and intelligently schedule multiple unmanned devices to collaboratively execute tasks, particularly in highly dynamic and uncertain environments, has become a research hotspot in the field of unmanned equipment. These unmanned devices can not only complete simple independent tasks, but also perform more complex tasks through coordination and cooperation, such as military reconnaissance, environmental monitoring, and disaster relief. The core challenge of collaborative scheduling of unmanned devices lies in how to effectively and dynamically adjust resource allocation and task execution strategies based on task requirements, unmanned device capabilities, and environmental changes to improve the efficiency and accuracy of task completion.

[0003] Most existing scheduling methods are based on preset rules or statically divided task regions, failing to fully consider the dynamic changes in task requirements, unmanned equipment status, and environmental factors. Therefore, they are ill-suited to complex and ever-changing application scenarios. Traditional methods often rely on fixed grids or regular spatial partitioning for task region division. This approach ignores the spatiotemporal variations in task requirements and the adaptability of unmanned equipment, leading to uneven resource scheduling and low execution efficiency. Furthermore, the collaborative scheduling of unmanned equipment often lacks adaptive assessment of its capabilities, resulting in uneven load distribution, resource waste, and task delays.

[0004] Therefore, the existing technology has problems and needs further improvement and development. Summary of the Invention

[0005] (I) Purpose of the invention: In order to solve the problems existing in the prior art, the purpose of the present invention is to provide a task scheduling method and system based on spatiotemporal voxelization and unmanned equipment self-evaluation, so as to solve the problems of static region division and insufficient dynamic adaptability in the prior art.

[0006] (II) Technical Solution: To address the aforementioned technical problems, this technical solution provides a task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation, including:

[0007] Step 1: Dynamically adjust voxel units according to the demand intensity of the task area;

[0008] Step 2: Evaluate the unmanned equipment in real time using multi-dimensional parameters to obtain an adaptability score;

[0009] Step 3: Adjust the weight coefficients of voxel units based on the adaptive score, assign tasks, and execute them; dynamically update the voxel units based on task feedback, unmanned equipment status, and real-time environmental information.

[0010] V′ i =V i ×(1+ΔC i )×(1+ΔE i ),

[0011] Among them, V′ i It is an updated voxel unit; V i It is the voxel unit before the update; ΔC i It is the effect of changes in the state of unmanned equipment on voxel units; ΔE i It is the effect of environmental changes on voxel units;

[0012] Step 4: Provide real-time feedback on task execution status, dynamically adjust weight coefficients, and optimize subsequent scheduling strategies.

[0013] Preferably, step 1 specifically includes:

[0014] Step 101: Obtain the task demand intensity based on the input task information;

[0015] Step 102: Dynamically divide the task area into voxel units according to the intensity of task requirements.

[0016] Preferably, the task demand intensity is dynamically calculated based on task type, collection frequency, and priority:

[0017] D i =w1×T i +w2×F i +w3×S i ,

[0018] Among them, D i T is the task demand intensity of region i; i It is the weight of the task type; F i It is the task acquisition frequency; S i It represents the strategic importance of the region; w1, w2, and w3 are weighting coefficients.

[0019] Preferably, the task region is divided into multiple three-dimensional voxel units based on the intensity of task demand. Each three-dimensional voxel unit corresponds to one operation unit in the task region. Smaller voxel units are used in high-demand regions, while larger voxel units are used in low-demand regions.

[0020]

[0021] Among them, V min D is the side length of the smallest voxel unit.high The threshold for high demand; D medium The threshold for the required amount.

[0022] Preferably, step 2 includes:

[0023] Step 201: Collect multi-dimensional parameters of the unmanned equipment;

[0024] Step 202: Calculate the adaptability in real time based on the multidimensional parameters of the unmanned equipment;

[0025] Step 203: Determine whether to accept the task based on the adaptability of the unmanned equipment.

[0026] Preferably, each unmanned device performs an adaptive assessment before executing a task, evaluating its ability to perform the current task based on its battery level, load, sensor health, and computing power.

[0027] A device =w 11 ×f(B current B max )+w 12 ×f(L current ,L max )+w 13 ×f(S current ,S max )+w 14 ×f(C current C required ),

[0028] Among them, B current and B max These represent the current battery level and maximum battery level of the unmanned equipment, respectively; L current and L max These represent the current load and maximum load of the unmanned equipment, respectively; S current and S max These represent the sensor's current health status and maximum health status, respectively; C current and C required These represent the current computing power of the unmanned equipment and the computing power required for the mission, respectively; w 11 w 12 w 13 w 14 These are dynamic weights, representing the impact of power consumption, load, sensor health, and computing power on the adaptability of unmanned equipment.

[0029] Preferably, each voxel unit is weighted according to the intensity of task requirements, environmental factors, and the adaptability of unmanned equipment to determine the priority of resource allocation:

[0030] W i =w4×D i +w5×Ci +w6×E i ,

[0031] Among them, W i It is the weighted value of the voxel unit in the i-th region; D i C is the task demand intensity of region i; i For the capability assessment of unmanned equipment in region i; E i denoted as the environmental factors of region i; w4, w5, and w6 are weighting coefficients, representing the intensity of task requirements, the capabilities of unmanned equipment, and the impact of the environment.

[0032] Preferably, through multi-objective optimization decision-making, comprehensively considering factors such as task priority, latency, and resource consumption, tasks are rationally scheduled and equipment resources are allocated:

[0033] O total =w8·P priority +w9·(1-T delay )+w 10 ·C cost ,

[0034] Among them, P priority Task priority is assigned; higher priority tasks will receive higher weight. delay The delay time is the time required for the task; the greater the delay, the greater the impact. (C) cost The resource consumption cost of the task; the more resources consumed, the smaller the objective function value; w8, w9, w 10 These are weighting coefficients, representing the impact of task priority, latency, and resource consumption, respectively.

[0035] Preferably, during task execution, based on historical task data and real-time feedback data on device power consumption, load, and environmental changes, the weight coefficients and adaptability are updated through an incremental learning algorithm to dynamically optimize the scheduling strategy.

[0036] w new =w old +α×(A feedback -A old ),

[0037] Among them, w new For the updated weight coefficients, w old For historical weighting coefficients, A feedback It is an adaptive score derived from historical feedback and task results, A. old It is the previous adaptive score, and α is the learning rate, which controls the adjustment speed.

[0038] The task scheduling system based on spatiotemporal voxelization and unmanned equipment self-evaluation is applicable to task scheduling methods based on spatiotemporal voxelization and unmanned equipment self-evaluation, including a spatiotemporal voxel unitization unit, an equipment adaptive evaluation unit, a task scheduling allocation unit, and a feedback optimization unit.

[0039] The spatiotemporal voxel unit dynamically adjusts the voxel unit according to the demand intensity of the task area;

[0040] The equipment adaptive evaluation unit evaluates the unmanned equipment in real time using multi-dimensional parameters to obtain an adaptive score;

[0041] The task scheduling and allocation unit adjusts the weight coefficients of voxel units based on adaptive scores, allocates tasks, and executes them; it dynamically updates voxel units based on task feedback, unmanned equipment status, and real-time environmental information.

[0042] V′ i =V i ×(1+ΔC i )×(1+ΔE i ),

[0043] Among them, V′ i It is an updated voxel unit; V i It is the voxel unit before the update; ΔC i It is the effect of changes in the state of unmanned equipment on voxel units; ΔE i It is the effect of environmental changes on voxel units;

[0044] The feedback optimization unit provides real-time feedback on task execution status, dynamically adjusts weight coefficients, and optimizes subsequent scheduling strategies.

[0045] (III) Beneficial Effects: This invention, based on a task scheduling method and system using spatiotemporal voxelization and unmanned equipment self-assessment, firstly divides the task area into multiple voxel units by real-time calculation of the demand intensity of the task area, and dynamically adjusts the voxel units according to changes in task demand, thereby achieving more precise resource allocation. Secondly, through a weighted voxel unit allocation strategy, each voxel unit is weighted to ensure that tasks in high-load areas or complex environments are prioritized, significantly improving the accuracy of task execution and resource utilization. Thirdly, through an unmanned equipment adaptive self-assessment mechanism, the equipment can quantify its own state parameters in real time and autonomously decide on task acceptance, significantly improving load balancing and task execution reliability. Combining multi-objective optimization algorithms and dynamic path planning strategies, task scheduling optimization in complex environments is achieved by comprehensively considering task priority, execution timeliness, and resource consumption. Finally, relying on an incremental learning feedback mechanism, the scheduling strategy can be continuously iterated based on task execution data, effectively enhancing environmental adaptability and collaborative operation efficiency in dynamic scenarios, fundamentally solving technical problems such as resource allocation imbalance, equipment overload risk, and dynamic response lag in traditional scheduling methods. Attached Figure Description

[0046] Figure 1 This is a flowchart of the steps of the task scheduling method based on spatiotemporal voxelization and self-evaluation of unmanned equipment in this invention;

[0047] Figure 2 This is a schematic diagram of the task scheduling system based on spatiotemporal voxelization and self-evaluation of unmanned equipment according to the present invention;

[0048] Figure 3 This is a schematic diagram of voxel unit division according to an embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of the weighted voxel unit according to an embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of the device adaptability score according to an embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram of task allocation according to an embodiment of the present invention. Detailed Implementation

[0052] The present invention will be further described in detail below with reference to preferred embodiments. More details are set forth in the following description in order to provide a full understanding of the present invention. However, the present invention can obviously be implemented in many other ways different from those described herein. Those skilled in the art can make similar extensions and derivations based on actual application situations without departing from the spirit of the present invention. Therefore, the scope of protection of the present invention should not be limited by the content of this specific embodiment.

[0053] The accompanying drawings are schematic diagrams of embodiments of the present invention. It should be noted that these drawings are for illustrative purposes only and are not drawn to scale, and should not be construed as limiting the actual scope of protection of the present invention.

[0054] Task scheduling methods based on spatiotemporal voxelization and self-evaluation of unmanned equipment, such as Figure 1 As shown, the specific steps include:

[0055] Step 1: Dynamically adjust voxel units according to the demand intensity of the task area.

[0056] Step 2: Evaluate the unmanned equipment in real time using multi-dimensional parameters to obtain an adaptability score.

[0057] Step 3: Adjust the weight coefficients of voxel units based on the adaptive score, assign tasks, and execute them.

[0058] Step 4: Provide real-time feedback on task execution status, dynamically adjust weight coefficients, and optimize subsequent scheduling strategies.

[0059] The task scheduling system based on spatiotemporal voxelization and unmanned equipment self-evaluation is applicable to the task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation, such as... Figure 2 As shown, the system includes a spatiotemporal voxel unitization unit, an equipment adaptive evaluation unit, a task scheduling and allocation unit, and a feedback optimization unit. The spatiotemporal voxel unitization unit dynamically adjusts the voxel units based on the demand intensity of the task area. The equipment adaptive evaluation unit evaluates the unmanned equipment in real time using multi-dimensional parameters to obtain an adaptability score. The task scheduling and allocation unit adjusts the weight coefficients of the voxel units based on the adaptability score, allocates tasks, and executes them. The feedback optimization unit provides real-time feedback on the task execution status, dynamically adjusts the weight coefficients, and optimizes subsequent scheduling strategies.

[0060] More specifically, to address the dynamic nature of task requirements and environmental changes, the spatiotemporal voxel unitization unit in step 1 of this invention dynamically divides the task region. Each task region is divided into multiple three-dimensional voxel units according to the intensity of its task requirements. Specific operations include:

[0061] Step 101: Based on the input task information, obtain the task requirement intensity.

[0062] The task information includes task requirements, environmental data, and historical task execution data. Task requirements include task type (meteorology, surveying, navigation, etc.), data collection frequency, task priority, and task time limit. Environmental data includes weather conditions (e.g., sunny, rainy), obstacle density, and communication interference. Historical task execution data includes past task execution status, real-time task feedback, and historical equipment performance.

[0063] The task information described herein consists of dimensionless values ​​obtained after data cleaning and normalization. The sources of the original data include historical data calculations, external system pushes, and expert experience definitions. No specific restrictions are placed on the specific methods used for data cleaning and normalization.

[0064] The intensity of task requirements is dynamically calculated based on task type, collection frequency, priority, etc., as shown in formula (1):

[0065] D i =w1×T i +w2×F i +w3×S i (1),

[0066] Among them, D i T is the task demand intensity of region i; i F is the weight of the task type in the i-th region; i S is the task acquisition frequency for the i-th region; i is the strategic importance of region i; w1, w2, and w3 are weighting coefficients.

[0067] Step 102: Dynamically divide the task area into voxel units according to the intensity of task requirements.

[0068] The task area is divided into multiple three-dimensional voxel units based on the intensity of task demand; the voxel units are dynamically adjusted by combining environmental data and historical task execution data to ensure that more refined voxel units are used in high-demand areas, thereby optimizing resource allocation.

[0069] More specifically, the VoxelGrid technique is used to integrate the task demand intensity scores of all task regions to form a data demand heatmap. During voxel grid partitioning, a three-level attention weighting system is constructed to retain key features of the task regions: task-level attention, data-level attention, and spatial-level attention. Task-level attention assigns high weights to task type and acquisition frequency. Data-level attention extracts key features from sensor data such as heat source points from infrared thermal imaging and obstacle coordinates from LiDAR using an autoencoder, retaining feature dimensions with a variance greater than or equal to 0.8. Spatial-level attention generates heatmap masks for strategically important sub-regions, prioritizing the preservation of voxel unit integrity within the masked area during splitting.

[0070] The data demand heatmap marks the task demand intensity of each region and sorts them according to demand priority. The data demand heatmap helps unmanned equipment identify which regions have higher data demand and prioritize the execution of tasks in these regions. The task region is divided into multiple three-dimensional voxel units, with one three-dimensional voxel unit corresponding to one operation unit of the task region. Smaller voxel units are used for high-demand regions, and larger voxel units are used for low-demand regions, as shown in formula (2):

[0071]

[0072] Among them, V min D is the side length of the smallest voxel unit. i D is the task demand intensity of region i; high The threshold for high demand; d medium The threshold for the required amount.

[0073] The voxel allocation diagram shows the voxel units for each task region. Each point in the diagram represents a voxel unit for a task region, and the voxel units are adjusted according to the intensity of task requirements. Regions with lower requirements have larger voxel units, while regions with higher requirements have smaller voxel units.

[0074] Specifically, voxel units are dynamically split or merged based on the intensity of task requirements, key features are retained through an attention mechanism weighted matrix, and the adjusted voxel units and weighted voxel units are output.

[0075] When the intensity of task demand exceeds the high demand threshold D high When the time comes, the splitting process is initiated, decomposing the current voxel unit into 4 micro-voxel units. For example, a 1km×1km voxel unit is split into 0.5km×0.5km micro-voxel units. Based on the attention weight matrix, high-importance features are assigned to independent micro-voxels, for example, the sub-region where the high-frequency acquisition task is located is divided into separate micro-voxel units.

[0076] When the number of voxels in three adjacent voxel units is all less than the threshold D required by the medium. medium When merging, they are combined into one macro voxel unit, with the size of the voxel unit being 1.5 times the original size. For example, three 0.5km micro voxel units are merged into a 0.75km voxel unit. Low-importance features are dimensionality-reduced and aggregated, and redundant data is compressed through mean pooling.

[0077] More specifically, before executing a task, each unmanned device in the device adaptive evaluation unit of step 2 undergoes an adaptive evaluation, assessing its ability to perform the current task based on multiple dimensions such as battery power, load, sensor health, and computing power. This includes:

[0078] Step 201: Collect multi-dimensional parameters of the unmanned equipment.

[0079] The multidimensional parameters refer to the unmanned equipment's own status information, environmental data, and historical task execution data. The status information includes battery level, load capacity, sensor health, computing power, and current task progress. The environmental data includes weather conditions (e.g., sunny or rainy), obstacle density, and communication interference. The historical task execution data includes past task execution status, real-time task feedback, and the equipment's historical performance.

[0080] Step 202: Calculate the adaptability in real time based on the multidimensional parameters of the unmanned equipment.

[0081] Before each unmanned device performs a task, it first undergoes an adaptive assessment, evaluating its capability to execute the task based on multiple factors such as battery level, load, sensor health, and computing power. The purpose is to better prevent unexpected events caused by the unmanned device's own condition during task execution and ensure the successful completion of the data acquisition task.

[0082] The device's adaptability score considers the influence of multiple parameters. The evaluation of each parameter can be represented as a standardized value and weighted by a weighting coefficient. As shown in formula (3), the weighting coefficient will be adjusted in real time to take into account the dynamic changes during task execution.

[0083]

[0084] Among them, B current and B max These represent the current battery level and maximum battery level of the unmanned equipment, respectively; L current and L max These represent the current load and maximum load of the unmanned equipment, respectively; S current and S max These represent the sensor's current health status and maximum health status, respectively; C current and C required These represent the current computing power of the unmanned equipment and the computing power required for the mission, respectively; w 11 w 12 w 13 w 14 These are dynamic weights, representing the impact of power consumption, load, sensor health, and computing power on the adaptability of unmanned equipment.

[0085] More preferably, A normalized function representing the status information of unmanned equipment, using battery power as an example for explanation. Normalization can be achieved using the following nonlinear function, as shown in formula (4), which fully considers the square relationship of the unmanned equipment's power consumption, ensuring that adaptability decreases significantly when the power consumption is low. Modeling the adaptability of unmanned equipment using nonlinear functions ensures that the evaluation results accurately reflect the actual situation.

[0086]

[0087] The function f(x,y) maps to the interval [0,1], so the value of the adaptability of unmanned equipment is also stable within the interval [0,1].

[0088] Similarly, as shown in formula (5), environmental factors (such as weather, obstacle density, etc.) also affect the mission execution capability of unmanned equipment.

[0089] A environment =w 15 ×f(W weather )+w 16 ×f(O obstacle )+w 17 ×f(C communication (5),

[0090] Among them, W weather It is a weather influencing factor; O obstacle This is the obstacle density factor; high-density obstacle areas affect execution capability, with an impact value ranging from 0 to 1; C communication It is a communication interference factor; high interference areas will reduce the efficiency of task execution; w 15 w 16 w 17 These are weighting coefficients, representing the impact of weather, obstacle density, and communication interference, respectively.

[0091] In this invention, the weather influence factor is preferably 1 on sunny days, 0.7 on rainy days, and 0.4 on stormy days.

[0092] Step 203: Determine whether to accept the task based on the adaptability of the unmanned equipment.

[0093] Specifically, based on the adaptability assessment score of the unmanned device, it is determined whether it has the ability to execute the current task, as shown in formula (6). If the adaptability score of the unmanned device is greater than or equal to the preset task acceptance threshold, the unmanned device will actively select to accept the task and update the task scheduling status. If the adaptability score of the unmanned device is less than the task acceptance threshold, the task will be reassigned to other unmanned devices to ensure the completion of the task. Finally, the unmanned device executes the task assigned to it and feeds back the execution result after the task is completed, updating the task status.

[0094]

[0095] The task reception threshold is preset according to the actual situation, and no specific requirements are made in this invention.

[0096] More specifically, the specific operation of the task scheduling and allocation unit in step 3 is as follows:

[0097] By using multi-objective optimization decision-making, and taking into account factors such as task priority, delay, and resource consumption, tasks are rationally scheduled and equipment resources are allocated.

[0098] Specifically, the spatiotemporal window for each task is optimized based on its time requirements, spatial distribution, and the resource availability of the unmanned equipment. Before task scheduling, the most suitable spatiotemporal window is first evaluated based on the task requirements. This spatiotemporal window should avoid conflicting with the spatiotemporal windows of other tasks as much as possible. This optimization of the spatiotemporal window ensures that tasks are completed on time and to a high standard, while avoiding equipment overload and task conflicts.

[0099] A multi-objective optimization method is employed to comprehensively consider factors such as task priority, latency, and resource consumption, ensuring efficient task scheduling. Based on the power, load, and location status of each device, the system assesses whether each device can complete the task within its time window. If a device cannot complete the task, its time window is adjusted or the task is reassigned. Tasks are prioritized, with urgent, high-priority tasks executed first. Lower-priority tasks are scheduled when available devices and resources are available. The task scheduling order is dynamically adjusted based on the task time window, device status, and task priority, ensuring that high-priority tasks are completed first. During device scheduling, the task's data collection area is considered to minimize idle or redundant paths and optimize resource utilization.

[0100] Based on equipment capabilities and environmental factors, voxel units are weighted and allocated. Each voxel unit is weighted according to task demand intensity, environmental factors (such as obstacle density, weather, etc.), and the adaptability of the unmanned equipment to determine resource allocation priority, as shown in formula (7). If the voxel unit partitioning meets resource allocation requirements, a voxel unit region is generated, and the weighted voxel unit region is output to step 3 for task scheduling and allocation. If the voxel unit partitioning does not meet resource allocation requirements, the size and weight coefficients of the voxel units are readjusted according to task demand intensity, and a new voxel unit region is generated.

[0101] W i =w4×D i +w5×C i +w6×E i (7),

[0102] Among them, W i It is the weighted value of the voxel unit in the i-th region; Di C is the task demand intensity of region i; i For the capability assessment of unmanned equipment in region i; E i denoted as the environmental factors of region i; w4, w5, and w6 are weighting coefficients, representing the relative importance of task demand intensity, unmanned equipment capabilities, and environmental impact in the total weighted value.

[0103] More specifically, the requirements and environmental conditions of the task area will change during task execution. Therefore, the voxel unit division needs to be updated based on real-time data. The process of dynamically updating the voxel unit division is adjusted based on task feedback, unmanned equipment status, and real-time environmental information. As shown in formula (8):

[0104] V′ i =V i ×(1+ΔC i )×(1+ΔE i (8),

[0105] Among them, V′ i It is an updated voxel unit; V i It is the voxel unit before the update; ΔC i This refers to the impact of changes in the state of unmanned equipment on voxel units (e.g., a decrease in the battery power of the unmanned equipment, leading to a reduction in voxel unit partitioning); ΔE i This refers to the impact of environmental changes on voxel units (e.g., weather changes leading to changes in regional demand).

[0106] The weighted voxel allocation diagram illustrates how voxel units are weighted based on equipment capabilities and environmental factors. It displays the voxel unit weighting for each task area, with the color intensity of the voxel units indicating the weighted resource allocation. Darker colors indicate higher weighted voxel unit areas and stronger resource requirements, while lighter colors indicate lower weighted voxel unit areas and weaker resource requirements.

[0107] During task execution, if the device status changes (such as insufficient power, network problems, etc.), it will be fed back to the scheduling system in real time. The scheduling system will adjust the time and space window of the task and the device scheduling according to the feedback. The specific task scheduling optimization objective function is shown in formula (9):

[0108] O total =w8·P priority +w9·(1-T delay )+w 10 ·C cost (9),

[0109] Among them, P priority Task priority is assigned based on priority; higher priority tasks will receive higher weight. delayThe longer the delay, the greater the impact and the lower the objective function value; C cost The resource consumption cost of the task; the more resources consumed, the smaller the objective function value; w8, w9, w 10 These are weighting coefficients, representing the impact of task priority, latency, and resource consumption, respectively.

[0110] The numerical values ​​used for evaluation in the formula (9), namely task priority, task delay time, and task resource consumption cost, are dimensionless values ​​obtained after data cleaning, normalization and other processing. The sources of the original data include historical data calculation, external system push, expert experience definition, etc., and there are no restrictions on the specific sources.

[0111] The path planning diagram shows the design results of the task execution path, demonstrating how unmanned equipment plans its path within the task area to avoid task conflicts and resource waste.

[0112] The path planning diagram for the unmanned equipment shows how the equipment enters each voxel unit. Red lines in the diagram represent the equipment's path, as it starts from its initial position and follows the planned path to enter different voxel unit regions.

[0113] Upon failure, a bidirectional feature mapping migration process is triggered to generate a high-dimensional feature vector and match candidate voxel units. When the device adaptability score is lower than the threshold, it is determined to be a "sudden failure". At this time, the task scheduling and allocation unit immediately starts the load migration process, and at the same time, the feedback optimization unit freezes the task progress of the original failed voxel unit and generates a feature snapshot containing task type, remaining duration, and environmental parameters.

[0114] The key data of the original fault voxel, including the intensity of task requirements, equipment capability assessment, and environmental factors, are encoded into a high-dimensional feature vector through a cross-voxel correlation matrix. The matrix dimension is N×M, where N represents the number of voxel units in the task region division, that is, the total number of independent task units in three-dimensional space, and M represents the feature dimension contained in each voxel unit, that is, the number of key parameters describing the attributes of the voxel unit.

[0115] Based on weighted voxel unit allocation, the top three candidate voxel units among neighboring voxel units are selected. The feature vector matching degree is calculated using cosine similarity, and target voxel units with a similarity greater than or equal to 0.85 are selected for neighborhood matching. After receiving the feature vector, the target voxel unit recovers the original task progress through reverse mapping and automatically compensates for environmental changes during the migration process for bidirectional decoding.

[0116] If the original task is of high priority, the target voxel's task priority weight coefficient is increased through a multi-objective optimization function, thereby achieving dynamic priority adjustment and ensuring priority resource allocation. If there are task conflicts among the target voxels, the tasks are split through a time-slice round-robin mechanism, for example, the original task is split into core subtasks and auxiliary subtasks, with the core subtasks being executed first.

[0117] More specifically, the specific operation of the feedback optimization unit in step 4 is as follows:

[0118] During task execution, based on historical task data and real-time feedback data on power consumption, load, and environmental changes from the equipment, the weight coefficients and adaptability are updated through incremental learning algorithms to dynamically optimize the scheduling strategy.

[0119] Specifically, the unmanned equipment provides feedback on status and environmental changes during task execution. The feedback optimization unit combines historical data and real-time feedback to perform incremental learning and continuously optimize the task scheduling strategy to adapt to the dynamically changing environment and the status of the unmanned equipment.

[0120] Furthermore, this invention introduces an intelligent feedback mechanism to adjust the evaluation model based on historical task execution data. As task execution progresses, the weights of evaluation factors are updated in real time to ensure that the evaluation results remain consistent with the dynamic changes in the environment, equipment, and task. This allows the unmanned equipment to adjust its task execution strategy based on real-time feedback during execution. The real-time task execution data and environmental changes of the unmanned equipment are used as inputs, and the task scheduling is optimized through incremental learning, as shown in formula (10):

[0121] w new =w old +α×(A feedback -A old (10),

[0122] Among them, w new For the updated weight coefficients, w old For historical weighting coefficients, A feedback It is an adaptive score derived from historical feedback and task results, A. old It is the previous adaptive score, and α is the learning rate, which controls the adjustment speed.

[0123] Decision optimization is carried out by combining task requirements (such as task priority, equipment load, task completion time, etc.) to ensure that task scheduling not only meets the adaptability of the equipment, but also efficiently meets the time limit, resource consumption and other requirements of the task.

[0124] The final output includes optimized task scheduling and device scheduling schemes. The task scheduling scheme includes information such as the assigned device, start and end times, and execution area for each task. The device scheduling scheme includes information such as task arrangement, time and space window, and load for each device.

[0125] The following is a description with reference to specific embodiments:

[0126] Example 1:

[0127] This embodiment applies a collaborative task scheduling method based on spatiotemporal hotspot demand voxel unitization and unmanned equipment adaptive self-evaluation to a complex emergency rescue mission. It assumes multiple unmanned devices are performing search and rescue missions in a disaster area, where the mission area exhibits varying demand intensities, complex environmental conditions, multiple unmanned device states, and task priorities. This embodiment demonstrates how to dynamically evaluate mission requirements and unmanned equipment adaptability for real-time task scheduling and execution.

[0128] The calculation rules for the numerical values ​​of each algorithm are shown in Table 1, and the calculated numerical values ​​are shown in Table 2.

[0129] Table 1. Numerical Evaluation Principles of Algorithms

[0130]

[0131]

[0132] Table 2 Algorithm Numerical Values

[0133]

[0134]

[0135]

[0136]

[0137] Assume the mission area is a 10km x 10km disaster zone, divided into multiple smaller zones, each 1km x 1km. The mission requirements for each smaller zone are determined by the following factors:

[0138] Task type T i Such as search and rescue missions, meteorological monitoring missions, etc.

[0139] Task acquisition frequency F i This refers to the frequency at which a task needs to collect data; tasks with higher frequency require more unmanned equipment support.

[0140] Regional importance S i Different areas within the disaster zone have varying degrees of importance to the search and rescue efforts, with important areas receiving higher priority.

[0141] Based on these factors, the task demand intensity D for each region can be calculated. i The intensity of task requirements is shown in Table 3:

[0142] Table 3 Task Demand Intensity

[0143]

[0144] like Figure 3 As shown, based on the task requirement intensity D i The task area is divided into multiple voxel units. The size of each voxel unit is inversely proportional to the task demand intensity; smaller voxel units are used in high-demand areas, while larger voxel units are used in low-demand areas. Assume the smallest voxel unit has a side length of 1 km and the largest voxel unit has a side length of 3 km. The voxel units are dynamically adjusted according to the task demand intensity, as shown in Table 4.

[0145] Table 4 Voxel Unit Adjustment

[0146]

[0147] Weighted voxel unit allocation is a process of allocating task areas based on the intensity of task area demand, the capabilities of unmanned equipment, and environmental factors. As shown in Table 5, by weighting each voxel unit area, the system can more accurately allocate tasks to unmanned equipment, ensuring that tasks in high-demand areas and complex environments are prioritized, thereby improving task execution efficiency. The weighting value of each voxel unit is determined by the following factors:

[0148] Task requirement intensity D i The demand intensity of each region reflects the urgency, task type, collection frequency, and importance of the region. Regions with higher demand intensity will receive a weighted allocation of more resources to voxel units.

[0149] Unmanned equipment capability C i The capability assessment of unmanned equipment takes into account multiple factors such as the equipment's power, load, and sensor health. Areas with strong unmanned equipment capabilities will be prioritized for task allocation to ensure efficient task execution.

[0150] Environmental Factor E i Environmental factors such as weather conditions and obstacle density can also affect mission execution. Areas with complex environments (such as areas with dense obstacles or severe weather) will have access to more resources to ensure the successful completion of the mission.

[0151] The weighted value W for each task region i It is divided into three parts: Task requirement intensity D i Unmanned equipment capability C i and environmental factors E i .like Figure 4As shown, the bar chart for each region is divided into three different color segments: blue represents the intensity of task requirements, indicating the intensity of requirements in the task region; green represents equipment capabilities, reflecting the assessment of equipment capabilities; and red represents environmental factors, indicating the impact of the environment on task execution.

[0152] Table 5 Weighted Voxel Unit Allocation Table

[0153]

[0154]

[0155] Based on the aforementioned weighted values, resources are prioritized for allocation to areas with higher weighted values, such as Area 1 and Area 8, which have stronger task demands and greater unmanned equipment capabilities. This effectively avoids resource waste and ensures that high-demand tasks can be completed in a timely and efficient manner.

[0156] In the adaptive self-evaluation phase of the unmanned equipment, it is assumed that there are 5 unmanned devices participating in the task, each with a different state, and therefore each unmanned device has a different adaptability to the task. The adaptability of the unmanned devices is calculated, as shown in Table 6.

[0157] Table 6. Example of adaptability assessment for unmanned equipment:

[0158]

[0159] like Figure 5 As shown, the adaptability score C for each device is... i The assessment is divided into three parts: device battery level, device load, and device sensor health. Each bar chart is divided into three different color segments: blue for battery level, representing the contribution of battery power to the adaptability score; green for load, representing the contribution of load to the adaptability score; and red for sensor health, representing the contribution of sensor health to the adaptability score.

[0160] According to the adaptability score C of unmanned equipment i The intensity of the task requirements in the task area D i The system automatically selects suitable unmanned equipment to perform tasks. As shown in Table 7, tasks are preferentially assigned to unmanned equipment with higher adaptability scores to ensure task execution efficiency.

[0161] Table 7 Task Allocation Table

[0162]

[0163] like Figure 6The diagram illustrates how the task requirement intensity, task weighting, and equipment adaptability score are allocated and executed for each task region. The adaptability allocation for each voxel unit is determined through the assessment of equipment capabilities and environmental factors. Each bar represents the impact of equipment capabilities or environmental factors on task scheduling. The blue bars represent the task requirement intensity D. i The green bars represent the task weighting value W. i The red bars represent the equipment adaptability score C. i Each region's bars are divided into three parts to allow for comparison of different dimensions of the task.

[0164] During task execution, the unmanned equipment also provides real-time feedback on the task execution status, including information such as battery level and load changes. As shown in Table 8, task allocation is dynamically adjusted based on this feedback information, and the task scheduling strategy is optimized through an incremental learning mechanism, making the scheduling results more in line with changes in the environment and task requirements, as shown in Table 9.

[0165] Table 8 Feedback Data and Incremental Learning Mechanism

[0166]

[0167] Table 9 shows the task scheduling strategy output.

[0168]

[0169] The incremental learning mechanism adjusts the task scheduling strategy through real-time feedback data. It updates the task allocation strategy and the unmanned equipment adaptability assessment model by combining feedback data from the unmanned equipment with historical execution data. This dynamically adjusts task allocation and optimizes scheduling decisions.

[0170] This invention presents a task scheduling method and system based on spatiotemporal voxelization and unmanned equipment self-evaluation. First, by calculating the demand intensity of the task area in real time, the task area is divided into multiple voxel units, and the voxel units are dynamically adjusted according to changes in task demand, thus achieving more precise resource allocation. Second, a weighted voxel unit allocation strategy is used to weight each voxel unit, ensuring that tasks in high-load areas or complex environments are prioritized, significantly improving task execution accuracy and resource utilization. Third, through an unmanned equipment adaptive self-evaluation mechanism, the equipment can quantify its own state parameters in real time and autonomously decide on task acceptance, significantly improving load balancing and task execution reliability. Combining multi-objective optimization algorithms and dynamic path planning strategies, task scheduling optimization in complex environments is achieved by comprehensively considering task priority, execution timeliness, and resource consumption. Finally, relying on an incremental learning feedback mechanism, the scheduling strategy can be continuously iterated based on task execution data, effectively enhancing environmental adaptability and collaborative operation efficiency in dynamic scenarios, fundamentally solving technical problems such as resource allocation imbalance, equipment overload risk, and dynamic response lag in traditional scheduling methods.

[0171] In summary, this invention achieves precise allocation of resources in the task area and improves equipment collaboration efficiency through a multi-dimensional dynamic optimization mechanism. It breaks through the limitations of traditional static grid partitioning and constructs a three-dimensional voxel-based dynamic partitioning model based on task demand intensity, real-time equipment status, and environmental factors. This enables refined scheduling of the task area, avoids resource waste during task execution, ensures that each device can perform appropriate tasks according to its capabilities, and improves flexibility and response speed in dynamic environments.

[0172] The above description illustrates preferred embodiments of the present invention and helps those skilled in the art to more fully understand the technical solution of the present invention. However, these embodiments are merely illustrative and should not be construed as limiting the specific implementation of the present invention to these embodiments. For those skilled in the art, several simple deductions and modifications can be made without departing from the inventive concept, and all such modifications should be considered within the protection scope of the present invention.

Claims

1. A task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation, characterized in that, include: Step 1: Dynamically adjust voxel units according to the demand intensity of the task area; Step 2: Evaluate the unmanned equipment in real time using multi-dimensional parameters to obtain an adaptability score; Step 3: Adjust the weight coefficients of voxel units based on the adaptive score, assign tasks, and execute them; dynamically update the voxel units based on task feedback, unmanned equipment status, and real-time environmental information. V′ i =V i ×(1+ΔC i )×(1+ΔE i ), Among them, V′ i It is an updated voxel unit; V i It is the voxel unit before the update; ΔC i It is the effect of changes in the state of unmanned equipment on voxel units; ΔE i It is the effect of environmental changes on voxel units; Step 4: Provide real-time feedback on task execution status, dynamically adjust weight coefficients, and optimize subsequent scheduling strategies.

2. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 1, characterized in that, Step 1 specifically includes: Step 101: Obtain the task demand intensity based on the input task information; Step 102: Dynamically divide the task area into voxel units according to the intensity of task requirements.

3. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 2, characterized in that, The task demand intensity is dynamically calculated based on task type, collection frequency, and priority. D i =w1×T i +w2×F i +w3×S i , Among them, D i T is the task demand intensity of region i; i F is the weight of the task type in the i-th region; i S is the task acquisition frequency for the i-th region; i is the strategic importance of region i; w1, w2, and w3 are weighting coefficients.

4. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 2, characterized in that, The task region is divided into multiple three-dimensional voxel units based on the intensity of task requirements. Each three-dimensional voxel unit corresponds to one operation unit in the task region. Smaller voxel units are used in high-demand regions, while larger voxel units are used in low-demand regions. Among them, V min D is the side length of the smallest voxel unit. i D is the task demand intensity of region i; high The threshold for high demand; D medium The threshold for the required amount.

5. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 1, characterized in that, Step 2 includes: Step 201: Collect multi-dimensional parameters of the unmanned equipment; Step 202: Calculate the adaptability in real time based on the multidimensional parameters of the unmanned equipment; Step 203: Determine whether to accept the task based on the adaptability of the unmanned equipment.

6. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 5, characterized in that, Before executing a task, each unmanned device undergoes an adaptive assessment, evaluating its ability to perform the current task based on its battery level, load, sensor health, and computing power. Among them, B current and B max These represent the current battery level and maximum battery level of the unmanned equipment, respectively; L current and L max These represent the current load and maximum load of the unmanned equipment, respectively; S current and S max These represent the sensor's current health status and maximum health status, respectively; C current and C required These represent the current computing power of the unmanned equipment and the computing power required for the mission, respectively; w 11 w 12 w 13 w 14 These are dynamic weights, representing the impact of power consumption, load, sensor health, and computing power on the adaptability of unmanned equipment.

7. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 1, characterized in that, For each voxel unit, a weighted allocation is performed based on the intensity of task requirements, environmental factors, and the adaptability of unmanned equipment to determine the priority of resource allocation: IN i =w4×D i +w5×C i +w6×E i , Among them, W i It is the weighted value of the voxel unit in the i-th region; D i C is the task demand intensity of region i; i For the capability assessment of unmanned equipment in region i; E i denoted as the environmental factors of region i; w4, w5, and w6 are weighting coefficients, representing the intensity of task requirements, the capabilities of unmanned equipment, and the impact of the environment.

8. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 7, characterized in that, Through multi-objective optimization decision-making, comprehensively considering task priority, latency, and resource consumption, tasks are rationally scheduled and equipment resources are allocated: O total =w8·P priority +w9·(1-T delay )+w 10 ·C cost , Among them, P priority Task priority is assigned; higher priority tasks will receive higher weight. delay The delay time is the time required for the task; the greater the delay, the greater the impact. (C) cost The resource consumption cost of the task; the more resources consumed, the smaller the objective function value; w8, w9, w 10 These are weighting coefficients, representing the impact of task priority, latency, and resource consumption, respectively.

9. The task scheduling method based on spatiotemporal voxelization and unmanned equipment self-evaluation according to claim 1, characterized in that, During task execution, based on historical task data and real-time feedback data on device power consumption, load, and environmental changes, the weight coefficients and adaptability are updated through incremental learning algorithms to dynamically optimize the scheduling strategy. w new =w old +α×(A feedback -A old ), Among them, w new For the updated weight coefficients, w old For historical weighting coefficients, A feedback It is an adaptive score derived from historical feedback and task results, A. old It is the previous adaptive score, and α is the learning rate, which controls the adjustment speed.

10. A task scheduling system based on spatiotemporal voxelization and unmanned equipment self-evaluation, applicable to task scheduling methods based on spatiotemporal voxelization and unmanned equipment self-evaluation, characterized in that, It includes a spatiotemporal voxel unit, an equipment adaptive evaluation unit, a task scheduling and allocation unit, and a feedback optimization unit; The spatiotemporal voxel unit dynamically adjusts the voxel unit according to the demand intensity of the task area; The equipment adaptive evaluation unit evaluates the unmanned equipment in real time using multi-dimensional parameters to obtain an adaptive score; The task scheduling and allocation unit adjusts the weight coefficients of voxel units based on adaptive scores, allocates tasks, and executes them; it dynamically updates voxel units based on task feedback, unmanned equipment status, and real-time environmental information. V′ i =V i ×(1+ΔC i )×(1+ΔE i ), Among them, V′ i It is an updated voxel unit; V i It is the voxel unit before the update; ΔC i It is the effect of changes in the state of unmanned equipment on voxel units; ΔE i It is the effect of environmental changes on voxel units; The feedback optimization unit provides real-time feedback on task execution status, dynamically adjusts weight coefficients, and optimizes subsequent scheduling strategies.

Images