Multilayer semantic venue map

Abstract

The invention discloses a multi-layer semantic field map, belongs to the technical field of unmanned vehicle path planning, and remarkably improves the multi-task planning efficiency in a complex environment by integrating geographic information, task point dynamic perception and priority dynamic evaluation. According to the hierarchical design, multi-dimensional information such as terrains, obstacles, task targets and priorities is presented in a unified manner, so that an unmanned vehicle or a ground troop can quickly identify passable areas, key task points and potential risks, and decision delay caused by information fragmentation is avoided. Meanwhile, the task-path coupling analysis module associates task priorities and path planning in real time to ensure that high-value tasks are executed preferentially, resource allocation is more accurate, invalid path exploration and time loss are reduced, and the efficiency is improved. Through layered display of various key elements such as different geography, passable areas, service targets, service tasks and the like, a specific area can be more comprehensively understood and analyzed in multi-task and multi-path planning, and more efficient task decoupling is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned vehicle path planning, specifically a multi-layer semantic field map. Background Art

[0002] Collaborative mission planning for unmanned vehicles (UAVs) refers to the process by which multiple UAVs, operating in the same environment, collaborate to complete complex tasks through information sharing, interactive communication, and collaborative decision-making. It involves multiple technical areas, including path planning, task allocation, obstacle avoidance, and platoon control. Collaborative mission planning aims to improve the overall efficiency, safety, and robustness of UAV systems, enabling them to efficiently and autonomously perform tasks in complex, dynamic environments, such as logistics delivery, search and rescue, and security patrols. To achieve effective collaboration, UAVs must be equipped with advanced sensors, communication equipment, and computing platforms. They must also utilize technologies such as artificial intelligence, machine learning, and optimization algorithms to perceive, model, and analyze the environment, generating appropriate task allocation plans and driving paths. Collaborative mission planning also needs to consider factors such as vehicle dynamic constraints, communication latency, and information security to ensure system stability and reliability. With continued technological advancement, collaborative mission planning for UAVs will play an increasingly important role in future intelligent transportation systems.

[0003] However, existing technologies in vision and environmental perception mainly provide only single-dimensional planar map information such as targets and obstacles, which makes it impossible to provide effective basic data support for dynamic path planning of unmanned vehicles under dynamic multiple constraints, multiple targets and multiple tasks. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-layer semantic field map in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: a multi-layer semantic field map, wherein the multi-layer semantic field map comprises:

[0006] Basic geographic information and service area service target identification module;

[0007] Building structure, resource and key target identification module;

[0008] Task point and priority identification module;

[0009] The task point and priority identification module is internally configured with: a task point dynamic perception submodule, a priority dynamic evaluation submodule, a task point interactive correction submodule, a multi-level visual identification submodule and a task-path coupling analysis submodule.

[0010] In a preferred embodiment, during use of the basic geographic information and service area service target identification module, the drone uses cameras, airborne synthetic aperture radar interferometry data and multispectral data sensors to build a high-precision first-layer physical map of the environment, displaying terrain features, obstacle distribution and accessible paths.

[0011] In a preferred embodiment, the Building Structure, Resource, and Key Target Identification Module uses a multispectral camera and thermal imager to automatically identify key resources, service targets, and mission execution points in a region of interest, marking the resource locations and status on a map. Using multi-angle photography and the Structure from Motion algorithm, the drone generates a second-level model, displaying detailed information about entrances, windows, and shelters.

[0012] In a preferred embodiment, the priority dynamic evaluation submodule inputs the priority dynamic evaluation result into the task-path coupling analysis submodule in real time, and the task-path coupling analysis submodule adjusts the path weight according to the task point priority.

[0013] In a preferred embodiment, the task point dynamic perception submodule is internally equipped with a multispectral sensor, a thermal imager, and a real-time SLAM positioning system. The target area is scanned in full by an unmanned aerial vehicle platform, capturing the original data of terrain elevation, obstacle distribution, and potential task points in real time. The multimodal task point classification algorithm integrates a semantic segmentation network with a Transformer target detection model to extract terrain features, target contours, and temporary task point features from sensor data. The classification results are bound to geographic coordinates to generate a three-dimensional point cloud map with depth information, providing structured input for subsequent modules. The dynamic data calibration component fuses multi-source sensor time series data through Kalman filtering to ensure the accuracy and real-time performance of task point coordinates.

[0014] In a preferred embodiment, the dynamic priority evaluation submodule calculates the priority of task points in real time through a multi-objective optimization model. The process is as follows: First, the three objective functions of priority scoring are defined: time sensitivity Ts, path cost Cp, and strategic value Vs; second, an adaptive weight allocation strategy is adopted to dynamically adjust the weight coefficient wi(t) of each objective through an exponential decay function, so that the weight is automatically updated with the progress of task execution and sudden changes in the environment; finally, a nonlinear normalization method is introduced to map the multi-objective score into a single priority index P, and infeasible tasks are filtered out through constraints to output the final priority ranking;

[0015] The calculation formula of the dynamic weight coefficient wi(t) is:

[0016]

[0017] Where:

[0018] σ i Indicates the initial importance of the objective function;

[0019] λ i Represents the decay factor, which controls the rate of change of the weight over time t. A high λi value causes the weight to quickly shift towards short-term goals.

[0020] t represents the time the task has been executed or the time offset after the environment changes suddenly.

[0021] This formula achieves nonlinear dynamic allocation of weights through exponential decay, which not only retains the initial strategic intention but also strengthens the response capability to emergencies through λi, thereby supporting the adaptive adjustment of priority scores in the time dimension.

[0022] In a preferred embodiment, the task-path coupling analysis submodule deeply binds task point priorities to the unmanned vehicle path planning through a spatiotemporal constraint solver. First, a task point-path association matrix is ​​established, with priority score PP, path length LL, energy consumption EE, and conflict risk RR as constraint variables. Second, a mixed integer linear programming model is used to generate an initial path plan to ensure that the path weights of high-priority task points are maximized and spatiotemporal conflicts are minimized. Finally, a priority-driven path backtracking mechanism is introduced. When a high-priority task is detected to fail due to a sudden change in the environment, path backtracking is automatically triggered, the priority and resource allocation efficiency of the suboptimal task chain are re-evaluated, and an alternative path is generated through a heuristic search algorithm to achieve a global optimal solution for the parallel execution of multiple tasks.

[0023] The dynamic adjustment formula of path weight is:

[0024]

[0025] in:

[0026] W k represents the comprehensive weight of the k-th path;

[0027] P k Indicates the priority score of the task point associated with the path;

[0028] L k represents the path length;

[0029] E k Indicates estimated energy consumption;

[0030] R k represents the path conflict risk coefficient;

[0031] α, β, γ, and δ represent adjustable weight coefficients, which are preset by the command center according to the task type.

[0032] The calculation formula of the spatiotemporal conflict probability model is:

[0033]

[0034] Where:

[0035] C ij represents the spatiotemporal conflict probability of path i and path j;

[0036] t i ,t j represents the estimated execution time of paths i and j;

[0037] σ t represents the time conflict tolerance threshold;

[0038] A i ∩A j represents the spatial overlapping area of ​​paths i and j;

[0039] A i ∪A j represents the combined coverage area of ​​paths i and j.

[0040] In a preferred embodiment, the interaction between the task-path coupling analysis submodule and the priority dynamic evaluation submodule is realized through a dynamic data flow and a closed-loop feedback mechanism: the priority dynamic evaluation submodule outputs the task point priority score Pk in real time, and the task-path coupling analysis submodule calculates the path weight Wk based on Pk and evaluates the path conflict probability Cij; when it is detected that a high-priority task cannot be executed due to path conflict or insufficient resources, the coupling analysis module triggers a priority re-evaluation request and feeds back the conflict information to the dynamic evaluation module, which recalculates Pk by adjusting the weight attenuation factor λi or the strategic value Vs, forming a collaborative optimization cycle of "priority-driven path planning-path conflict feedback priority correction".

[0041] The priority-path feedback weight formula is:

[0042]

[0043] Where:

[0044] λ i Represents the attenuation factor of the objective function in the dynamic evaluation submodule;

[0045] λ0 represents the initial attenuation factor;

[0046] R k represents the path conflict risk coefficient;

[0047] Δt represents the task execution time deviation;

[0048] η,τ represent normalization constants.

[0049] In a preferred embodiment, the interactive task point correction submodule uses an augmented reality interface within the system, allowing ground commanders to manually adjust task point priorities, mark temporary tasks, or correct target coordinates using AR glasses or a touchscreen terminal. A two-way feedback mechanism feeds these manual correction instructions into the human-machine collaborative reinforcement learning model, dynamically optimizing the weight decay factor and strategic value coefficient within the priority evaluation parameters. Correction actions are also recorded in a decision log for subsequent model training and policy iteration.

[0050] In a preferred embodiment, a layered rendering pipeline is set up inside the multi-level visual identification sub-module. The first-level global view displays the task chain topology relationship and priority color coding, the second-level local view focuses on the task point details marked with dynamic icons, and the third layer overlays the path conflict warning heat map and resource allocation status.

[0051] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0052] 1. The present invention significantly improves the efficiency of multi-task planning in complex environments by integrating geographic information, dynamic perception of mission points, and dynamic priority assessment. Its layered design presents multi-dimensional information such as terrain, obstacles, mission objectives, and priorities in a unified manner, enabling unmanned vehicles or ground forces to quickly identify traversable areas, key mission points, and potential risks, avoiding decision-making delays caused by information fragmentation. At the same time, the task-path coupling analysis module associates task priorities with path planning in real time, ensuring that high-value tasks are executed first, allocating resources more accurately, and reducing ineffective path exploration and time loss. By layering and displaying various key elements such as different geography, traversable areas, service objectives, and service tasks, a more comprehensive understanding and analysis of specific areas can be achieved in multi-task and multi-path planning, achieving more efficient task decoupling.

[0053] 2. This invention enhances task adaptability and human-machine collaboration in dynamic environments. Through the interactive correction module, commanders can manually adjust task priorities based on battlefield changes. The system automatically and synchronously updates path planning and visualization identification, reducing the impact of sudden interference on the overall task chain. The multi-level visualization engine provides differentiated information display by role, allowing commanders to grasp global task dependencies and individual soldiers to focus on local action details. When the two work together, command transmission is clearer. This design ensures that when multiple tasks are executed in parallel, both strategic goals can be maintained and local environmental changes can be flexibly responded to. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a block diagram of the overall system of the present invention;

[0055] Figure 2This is a system block diagram of the task point and priority identification module in the present invention.

[0056] Markings in the figure: 1. Basic geographic information and service area service target identification module; 2. Building structure, resources and key target identification module; 3. Task point and priority identification module. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] Reference Figure 1-2 ,

[0059] Multi-layer semantic field map, which includes:

[0060] Basic geographic information and service area service target identification module;

[0061] Building structure, resource and key target identification module;

[0062] Task point and priority identification module;

[0063] The internal settings of the task point and priority identification module include: task point dynamic perception submodule, priority dynamic evaluation submodule, task point interactive correction submodule, multi-level visual identification submodule and task-path coupling analysis submodule.

[0064] During the use of the basic geographic information and service area service target identification module, the drone uses sensors such as cameras, airborne synthetic aperture radar interferometry (InSAR) data, and multispectral data to build a high-precision first-layer physical map of the environment, showing terrain features, obstacle distribution, and navigable paths. Combined with algorithms such as ORB-SLAM or DSO, the drone generates a real-time updated map with depth and coordinate information, marking navigable and non-navigable areas, and constructing a detailed basic GIS map. This geographic information provides ground forces with the basic structure of the environment, while helping them quickly identify and avoid non-navigable areas in complex areas.

[0065] The Building Structure, Resource, and Key Target Recognition module uses a multispectral camera and thermal imager to automatically identify key resources, service targets, and mission execution points within an area of ​​interest, marking their location and status on a map. Using multi-angle photography and the Structure from Motion algorithm, the drone generates a second-layer model, displaying detailed information about entrances, windows, and shelters.

[0066] The priority dynamic evaluation submodule inputs the priority dynamic evaluation results into the task-path coupling analysis submodule in real time, and the task-path coupling analysis submodule adjusts the path weight according to the task point priority.

[0067] The mission point dynamic perception submodule is equipped with a multispectral sensor, thermal imager, and real-time SLAM positioning system. It uses a drone platform to perform a full-area scan of the target area, capturing raw data on terrain elevation, obstacle distribution, and potential mission points in real time. The multimodal mission point classification algorithm integrates a semantic segmentation network with a Transformer object detection model to extract terrain features, target outlines, and temporary mission point characteristics from sensor data. The classification results are then bound to geographic coordinates to generate a three-dimensional point cloud map with depth information, providing structured input for subsequent modules. The dynamic data calibration component uses Kalman filtering to fuse multi-source sensor time series data to ensure the accuracy and real-time performance of mission point coordinates.

[0068] The dynamic priority assessment submodule calculates task priorities in real time using a multi-objective optimization model. Its core algorithm constructs a dynamic weight adjustment mechanism based on multi-source data input. The specific process is as follows: First, the three objective functions for priority scoring are defined: time sensitivity Ts, path cost Cp, and strategic value Vs. Second, an adaptive weight allocation strategy is adopted to dynamically adjust the weight coefficient wi(t) of each objective through an exponential decay function, so that the weight is automatically updated with the progress of task execution and sudden changes in the environment. Finally, a nonlinear normalization method is introduced to map the multi-objective scores into a single priority indicator P. Infeasible tasks are filtered out using constraints, and the final priority ranking is output.

[0069] The calculation formula of the dynamic weight coefficient wi(t) is:

[0070]

[0071] Where:

[0072] σ i Indicates the initial importance of the objective function;

[0073] λ i Represents the decay factor, which controls the rate of change of the weight over time t. A high λi value causes the weight to quickly shift towards short-term goals.

[0074] t represents the time the task has been executed or the time offset after the environment changes suddenly.

[0075] This formula achieves nonlinear dynamic allocation of weights through exponential decay, which not only retains the initial strategic intention but also strengthens the response capability to emergencies through λi, thereby supporting the adaptive adjustment of priority scores in the time dimension.

[0076] The task-path coupling analysis submodule deeply binds task point priorities with the unmanned vehicle path planning through a spatiotemporal constraint solver. Its core algorithm constructs a multi-objective optimization model based on dynamic priority scores and real-time environmental data. The specific process is as follows: First, a task point-path association matrix is ​​established, with priority score PP, path length LL, energy consumption EE, and conflict risk RR as constraint variables; second, a mixed integer linear programming model is used to generate an initial path plan to ensure that the path weights of high-priority task points are maximized and spatiotemporal conflicts are minimized; finally, a priority-driven path backtracking mechanism is introduced. When a high-priority task is detected to fail due to a sudden change in the environment, path backtracking is automatically triggered, and the priority and resource allocation efficiency of the suboptimal task chain are re-evaluated. An alternative path is generated through a heuristic search algorithm to achieve a global optimal solution for the parallel execution of multiple tasks.

[0077] The dynamic adjustment formula of path weight is:

[0078]

[0079] in:

[0080] W k represents the comprehensive weight of the k-th path;

[0081] P k Indicates the priority score of the task point associated with the path;

[0082] L k represents the path length;

[0083] E k Indicates estimated energy consumption;

[0084] R k represents the path conflict risk coefficient;

[0085] α, β, γ, and δ represent adjustable weight coefficients, which are preset by the command center according to the task type.

[0086] The calculation formula of the spatiotemporal conflict probability model is:

[0087]

[0088] Where:

[0089] C ij represents the spatiotemporal conflict probability of path i and path j;

[0090] t i ,t j represents the estimated execution time of paths i and j;

[0091] σ t represents the time conflict tolerance threshold;

[0092] A i ∩A j represents the spatial overlapping area of ​​paths i and j;

[0093] A i ∪A j represents the combined coverage area of ​​paths i and j.

[0094] The interaction between the task-path coupling analysis submodule and the priority dynamic evaluation submodule is realized through dynamic data flow and closed-loop feedback mechanism: the priority dynamic evaluation submodule outputs the task point priority score Pk in real time, and the task-path coupling analysis submodule calculates the path weight Wk based on Pk and evaluates the path conflict probability Cij; when it is detected that a high-priority task cannot be executed due to path conflict or insufficient resources, the coupling analysis module triggers a priority reassessment request and feeds back the conflict information to the dynamic evaluation module, which recalculates Pk by adjusting the weight attenuation factor λi or the strategic value Vs, forming a collaborative optimization cycle of "priority-driven path planning-path conflict feedback priority correction".

[0095] The priority-path feedback weight formula is:

[0096]

[0097] Where:

[0098] λ i Represents the attenuation factor of the objective function in the dynamic evaluation submodule;

[0099] λ0 represents the initial attenuation factor;

[0100] R k represents the path conflict risk coefficient;

[0101] Δt represents the task execution time deviation;

[0102] η,τ represent normalization constants.

[0103] The interactive task point correction submodule uses an augmented reality interface within the system. Ground commanders use AR glasses or touchscreen terminals to manually adjust task point priorities, mark temporary tasks, or correct target coordinates. A two-way feedback mechanism feeds manual correction instructions into the human-machine collaborative reinforcement learning model, dynamically optimizing the weight decay factors and strategic value coefficients within the priority evaluation parameters. Correction actions are recorded in the decision log for subsequent model training and policy iteration. By comparing historical correction data with task execution results, the closed-loop learning system continuously improves the robustness of task planning in complex scenarios, ensuring the progressive collaborative optimization of human-machine decision-making.

[0104] The multi-level visual identification submodule features a layered rendering pipeline. The first-level global view displays task chain topology and priority color coding. The second-level local view focuses on task point details, annotated with dynamic icons. The third-level overlays a path conflict warning heat map and resource allocation status. The situational awareness component switches display modes based on user role. The commander's view highlights strategic task dependencies and global resource scheduling trends, while the individual soldier's view enhances real-time path navigation and task point semantic labeling. Dynamic icons reflect task urgency through flashing frequency and rotation angle, while semantic labels indicate specific action instructions, such as the need for coordinated attack on high-value targets or the remaining resources at supply points. This enables immersive perception of battlefield information and efficient decision-making.

[0105] From the above we can know:

[0106] In this invention, by integrating geographic information, dynamic perception of mission points, and dynamic evaluation of priorities, the efficiency of multi-task planning in complex environments is significantly improved. Its hierarchical design presents multi-dimensional information such as terrain, obstacles, mission objectives, and priorities in a unified manner, allowing unmanned vehicles or ground forces to quickly identify traversable areas, key mission points, and potential risks, avoiding decision-making delays caused by information fragmentation. At the same time, the task-path coupling analysis module associates task priorities with path planning in real time, ensuring that high-value tasks are executed first, allocating resources more accurately, reducing ineffective path exploration and time loss, and by displaying different key elements such as geography, traversable areas, service objectives, and service tasks in a hierarchical manner, a more comprehensive understanding and analysis of specific areas can be achieved in multi-task and multi-path planning, and more efficient task decoupling can be achieved.

[0107] This invention enhances task adaptability and human-machine collaboration in dynamic environments. Through the interactive correction module, commanders can manually adjust task priorities based on battlefield changes, and the system automatically and synchronously updates path planning and visualization identification, reducing the impact of sudden interference on the overall task chain. The multi-level visualization engine provides differentiated information display by role, allowing commanders to grasp global task dependencies and individual soldiers to focus on local action details. When the two work together, command transmission is clearer. This design ensures that when multiple tasks are executed in parallel, both the consistency of strategic goals can be maintained and the flexibility to respond to sudden changes in the local environment can be achieved.

[0108] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements that are inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0109] The above description is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. Multi-layer semantic field map, characterized by: The multi-layer semantic field map includes: Basic geographic information and service area service target identification module (1); Building structure, resource and key target identification module (2); Task point and priority identification module (3); The task point and priority identification module (3) is internally provided with: a task point dynamic perception submodule, a priority dynamic evaluation submodule, a task point interactive correction submodule, a multi-level visual identification submodule and a task-path coupling analysis submodule.

2. The multi-layer semantic field map according to claim 1, characterized in that: During use, the basic geographic information and service area service target identification module (1) uses a camera, airborne synthetic aperture radar interferometry data, and a multispectral data sensor to construct a high-precision first-layer physical map of the environment, displaying terrain features, obstacle distribution, and navigable paths.

3. The multi-layer semantic field map according to claim 1, wherein: During use, the building structure, resource and key target identification module (2) uses a multispectral camera and a thermal imager to automatically identify key resources and service targets in the area of ​​interest, and marks the resource location and status on a map; through multi-angle shooting and the Structure from Motion algorithm, the drone generates a second-layer model to display detailed entrance, window and shelter information.

4. The multi-layer semantic field map according to claim 1, wherein: The priority dynamic evaluation submodule inputs the priority dynamic evaluation result into the task-path coupling analysis submodule in real time, and the task-path coupling analysis submodule adjusts the path weight according to the task point priority.

5. The multi-layer semantic field map according to claim 1, wherein: The task point dynamic perception submodule is internally equipped with a multispectral sensor, a thermal imager and a real-time SLAM positioning system. It uses an unmanned aerial vehicle (UAV) platform to perform a full-area scan of the target area, capturing the original data of terrain elevation, obstacle distribution and potential task points in real time. The multimodal task point classification algorithm integrates a semantic segmentation network and a Transformer target detection model to extract terrain features, target contours and temporary task point features from the sensor data. The classification results are bound to geographic coordinates to generate a three-dimensional point cloud map with depth information, providing structured input for subsequent modules.

6. The multi-layer semantic field map according to claim 1, wherein: The priority dynamic evaluation submodule calculates the priority of the task point in real time through a multi-objective optimization model. The process is as follows: First, the three objective functions of priority scoring are defined: time sensitivity Ts, path cost C p 、Strategic Value V s ; Secondly, an adaptive weight allocation strategy is adopted to dynamically adjust the weight coefficient w of each target through an exponential decay function i (t), so that the weights are automatically updated with the progress of task execution and sudden changes in the environment; finally, a nonlinear normalization method is introduced to map the multi-objective scores into a single priority index P, and infeasible tasks are filtered out through constraints to output the final priority ranking; The calculation formula of the dynamic weight coefficient wi(t) is: Where: σ i Indicates the initial importance of the objective function; λ i Represents the decay factor, which controls the rate of change of the weight over time t. A high λi value causes the weight to quickly shift towards short-term goals. t represents the time the task has been executed or the time offset after the environment suddenly changes; This formula achieves nonlinear dynamic allocation of weights through exponential decay, which not only retains the initial strategic intention but also strengthens the response capability to emergencies through λi, thereby supporting the adaptive adjustment of priority scores in the time dimension.

7. The multi-layer semantic field map according to claim 1, wherein: The task-path coupling analysis submodule deeply binds the task point priority with the unmanned vehicle path planning through a spatiotemporal constraint solver. First, a task point-path association matrix is ​​established, with priority score PP, path length LL, energy consumption EE, and conflict risk RR as constraint variables. Secondly, a mixed-integer linear programming model is used to generate an initial path plan, ensuring that the path weights of high-priority task points are maximized and spatiotemporal conflicts are minimized. Finally, a priority-driven path backtracking mechanism is introduced. When a high-priority task fails due to a sudden environmental change, path backtracking is automatically triggered, re-evaluating the priority and resource allocation efficiency of the suboptimal task chain. An alternative path is generated through a heuristic search algorithm to achieve a global optimal solution for the parallel execution of multiple tasks. The dynamic adjustment formula of path weight is: in: W k represents the comprehensive weight of the k-th path; P k Indicates the priority score of the task point associated with the path; L k represents the path length; E k Indicates estimated energy consumption; R k represents the path conflict risk coefficient; α, β, γ, and δ represent adjustable weight coefficients, which are preset by the command center according to the task type; The calculation formula of the spatiotemporal conflict probability model is: Where: C ij represents the spatiotemporal conflict probability of path i and path j; t i ,t j represents the estimated execution time of paths i and j; σ t represents the time conflict tolerance threshold; A i ∩A j represents the spatial overlapping area of ​​paths i and j; A i ∪A j represents the combined coverage area of ​​paths i and j.

8. The multi-layer semantic field map according to claim 1, wherein: The interaction between the task-path coupling analysis submodule and the priority dynamic evaluation submodule is achieved through a dynamic data flow and closed-loop feedback mechanism: the priority dynamic evaluation submodule outputs the task point priority score Pk in real time, and the task-path coupling analysis submodule calculates the path weight Wk based on Pk and evaluates the path conflict probability Cij; when it is detected that a high-priority task cannot be executed due to path conflict or insufficient resources, the coupling analysis module triggers a priority reassessment request and feeds back the conflict information to the dynamic evaluation module, which recalculates Pk by adjusting the weight attenuation factor λi or the strategic value Vs, forming a collaborative optimization cycle of priority-driven path planning and path conflict feedback priority correction; The priority-path feedback weight formula is: Where: λ i Represents the attenuation factor of the objective function in the dynamic evaluation submodule; λ0 represents the initial attenuation factor; R k represents the path conflict risk coefficient; Δt represents the task execution time deviation; η,τ represent normalization constants.

9. The multi-layer semantic field map according to claim 1, wherein: In the internal augmented reality operation interface of the task point interactive correction submodule, ground commanders manually adjust the priority of task points, mark temporary tasks, or correct target coordinates through AR glasses or touch terminals; the two-way feedback mechanism inputs manual correction instructions into the human-computer collaborative reinforcement learning model, dynamically optimizes the weight attenuation factor and strategic value coefficient in the priority evaluation parameters, and records the correction behavior in the decision log library.

10. The multi-layer semantic field map according to claim 1, wherein: The multi-level visual identification sub-module is internally provided with a layered rendering pipeline. The first-level global view displays the task chain topology relationship and priority color coding, the second-level local view focuses on the details of the task points marked with dynamic icons, and the third-level overlays the path conflict warning heat map and resource allocation status.

Images