Method and system for multi-agent multi-task planning and allocation in buried space

Abstract

The application discloses a kind of press bury space multi-agent multi-task planning and distribution method and system, it is related to emergency rescue task planning and distribution technical field, solve the technical problem that press bury space rescue cost and actual cost adaptation degree is lower and rescue scheme acquisition efficiency is lower, its technical scheme key point is by matrix verification function to the task-agent cost matrix Intelligent cutting, automatically delete the row and column that can not be executed, precluded a large number of invalid search space, substantially reduce the computational complexity, meet the real-time requirement of emergency response.Simultaneously in solving process, by dynamically deleting the row and column corresponding to the task and agent that has been allocated, it is not necessary to artificially adjust second time, can be directly used as execution basis.Simultaneously by the mechanism that the size of effective cost matrix scale dynamically selects solving path, the optimal balance of solving effect and efficiency is realized, overcome the defect that single algorithm is difficult to adapt to different emergency scene scale.

Description

Technical Field

[0001] This application relates to the field of emergency rescue mission planning and allocation technology, and in particular to a method and system for multi-subject, multi-mission planning and allocation of buried space. Background Technology

[0002] In emergency search and rescue operations in buried spaces (such as earthquake rubble, underground mines, and inside collapsed buildings), efficient and scientific task planning and resource (personnel and equipment) scheduling are key to improving the success rate of rescue and reducing secondary risks. These environments are typically characterized by complex structures, incomplete information, limited communication, and dynamically changing risks, posing a significant challenge to traditional command models that rely on human experience.

[0003] Existing technical solutions have the following main shortcomings when dealing with the specific scenario of emergency search and rescue in buried spaces: (1) The task cost model is overly simplified and out of touch with the actual rescue situation: Many existing task allocation methods typically use simple distance or time as cost measures, or treat task execution as a single phase. However, the rescue mission of buried space has a significant three-phase characteristic of "positioning-on-site operation-recovery". On-site operation costs are highly dependent on the scale of the operation area, environmental complexity, and the main operational capabilities, and cannot be accurately assessed by maneuver time alone. Existing technologies lack refined modeling of this multi-phase, multi-factor composite cost, which may lead to significant deviations in the actual implementation of the planning scheme.

[0004] (2) Lack of an efficient filtering mechanism for invalid task-subject combinations, resulting in low solution efficiency: When constructing the task-subject cost matrix, due to subject capabilities, resource constraints, or environmental limitations, there will be a large number of combinations where "subjects cannot perform a certain task". Existing methods usually set the cost of these invalid combinations to a maximum value and then directly feed them into the optimization algorithm for solution. This will result in the solution space containing a large number of invalid search branches. When the number of tasks and subjects is large, it will seriously reduce the convergence speed and solution efficiency of the optimization algorithm, and will not meet the real-time requirements of emergency response.

[0005] (3) Task allocation constraints fail to reflect the wave-based coordination characteristics of landslide rescue: Existing multi-agent task allocation schemes mostly focus on global optimization or minimizing disturbances, but do not fully consider the specific wave-based constraints in landslide rescue (within one wave, an agent can only perform one task, and a target area can only be operated by one team at the same time). If general algorithms do not embed such business rules, the generated schemes may not be directly executable and require secondary manual adjustments, which reduces the efficiency and reliability of automated decision-making.

[0006] (4) The algorithm strategy is too simplistic and difficult to adapt to different scales of rescue scenarios: The scale of emergency sites varies greatly, ranging from local rescues involving only a few targets to large-scale rescues involving dozens or hundreds of targets. Existing methods usually use a single algorithm (such as the Hungarian algorithm, auction algorithm, or ant colony algorithm). For small-scale scenarios, the initialization and iteration process of intelligent optimization algorithms may bring unnecessary overhead, while the simple exhaustive method can quickly find the global optimum; for large-scale scenarios, the exhaustive method is not feasible due to combinatorial explosion. Existing technologies lack an adaptive mechanism that can dynamically select the most suitable solution strategy according to the problem scale.

[0007] To address the shortcomings of existing technologies, this application proposes a method and system for multi-subject, multi-task planning and allocation of burial space based on multi-constraint cost matrix optimization. Summary of the Invention

[0008] This application provides a method and system for multi-subject, multi-task planning and allocation of buried space. Its technical purpose is to improve the accuracy of the cost assessment of buried space rescue and its adaptability to the actual rescue cost, improve the efficiency of rescue plan acquisition and the feasibility of the plan, take into account the optimality of the rescue plan, and enhance the robustness and adaptability of the system in response to emergencies.

[0009] The above-mentioned technical objective of this application is achieved through the following technical solution: A method for multi-subject, multi-task planning and allocation of buried space, comprising: Step S1: Acquire task data and subject data, and construct task set and subject set according to the task data and subject data respectively; wherein, the parameters of each job in the task set include job unique identifier, job location, job area range, job area risk, physical field index, business index and decision index, and the parameters of each subject in the subject set include subject unique identifier, subject real-time status information, subject capability attributes and subject currently available resources; Step S2: Filter the subject set to obtain a valid candidate subject set; Step S3: Obtain the priority level and category of each task in the task set according to the decision indicators; Based on the category of each task, resource type matching is performed on the task. Based on the result of resource type matching, subject matching is performed on each task in the valid candidate subject set. Based on the result of subject matching, the execution content of each task is determined. The execution content includes the task's placement path, operation path, and retrieval path. Step S4: Calculate the cost of each task that the subject can perform based on the content of the task, and construct a task-subject cost matrix based on the cost; Step S5: Trim the task-subject cost matrix to obtain the effective cost matrix; Step S6: Based on the effective cost matrix, solve for the final task allocation scheme with the minimum total cost, and allocate tasks according to the final task allocation scheme; Step S7: When executing a task according to the final task allocation scheme, monitor whether there are any abnormalities in the task execution status, main resource status, and environment status. If so, return to step S1 for replanning; otherwise, continue execution until the final task allocation scheme is completed.

[0010] Preferably, in step S1, the physical field indicators include traffic accessibility, environmental hazard, life status and sustainability, communication possibility, and the size and distribution of search and rescue targets; The business metrics are obtained based on the physical field metrics and include: the complexity of task information acquisition, task value, and the workload and difficulty of task implementation, which are expressed as follows: ; ; ; in, This indicates the complexity of obtaining task information. Indicates the value of the task. This indicates the workload and difficulty of the task. , , , , , , All are weighting coefficients. Indicates accessibility by transportation. Indicates environmental hazard. Indicates the possibility of communication. Indicates life status and sustainability. This indicates the scale and distribution of the search and rescue mission; The decision indicators are obtained based on the business indicators, including task grading and task classification. The task grading levels include high priority, medium priority and low priority, and the task classification categories include rapid reconnaissance, standard operation and heavy engineering. The task classification level is determined based on a priority score, which is expressed as: ; in, Indicates priority score, based on The value range of is used to divide tasks into high priority, medium priority, and low priority; , , All are weighting coefficients; The task categories include: when and In such cases, the mission is classified as a rapid reconnaissance mission; when and In this case, the task is classified into standard operation categories; when and At that time, the task will be classified as a heavy engineering project; in, , , , All of these are preset classification thresholds.

[0011] Preferably, step S2 includes: Step S21: Determine whether the current working status of the subject is dispatchable. If yes, proceed to step S22; otherwise, proceed to step S24. Step S22: Determine whether the baseline cost of the subject to the task set is less than the resource reservation value minus the cost tolerance value of the subject. If yes, proceed to step S23; otherwise, proceed to step S24. Step S23: Determine whether the subject's ability attributes match the category of a certain task. If yes, proceed to step S25; otherwise, proceed to step S24. Step S24: The subject is invalid; exclude the subject. Step S25: The subject is a valid candidate subject; therefore, the subject is included in the set of valid candidate subjects. The benchmark cost is expressed as: ,in, This represents the weighting coefficient determined based on the target's classification level. This indicates the estimated distance from the current location of the subject to the task set. Indicates the first One entity; This represents the cost calculation function, which is based on the estimated distance. and main body The subject's own ability attributes Estimate the baseline cost of performing the task.

[0012] Preferably, step S4 includes: Step S41: Traverse the task set and the set of valid candidate subjects, and calculate the cost of each task that each subject can execute, expressed as: ; in, Representing the subject Execute the task The cost; This represents the target priority correction factor, which is determined based on the priority score. Representing the subject Resource reservation value, This indicates the preset cost tolerance value; ; ; ; ; This indicates the cost during the placement phase. , , All represent weighting coefficients; This indicates the mission location, taking environmental constraints into account. With the main position The shortest path distance between them; Representing the subject Normal moving speed; This represents the cumulative risk value of the path; Representing the subject Ability attributes; Indicates the cost of the activity phase. , , All represent weighting coefficients; Indicates the size of the work area; Indicates the operational capability of the main unit; Indicates the environmental complexity factor; Indicates the risk level of the work area; Indicates the cost during the recovery phase. , , All represent weighting coefficients; This indicates the mission location, taking environmental constraints into account. Location of recycling points The shortest path distance between them; This represents the cumulative risk value of the recycling path; Step S42: Arrange all the calculated costs into an n×m initial task-subject cost matrix; where n represents the number of tasks in the task set and m represents the number of valid candidate subjects.

[0013] Preferably, step S5 includes: Step S51: For any column vector in the task-subject cost matrix If all its elements are infinite, then the column vector is... Corresponding subject If the cost of all executable tasks is infinite, then according to the matrix verification function, the corresponding column vector is removed from the task-body cost matrix. This yields the first task-main cost matrix; Step S52: For any row vector in the first task-main cost matrix If all its elements are infinite, it means that all entities execute this row vector. Corresponding task If the costs are all infinite, then according to the matrix verification function, delete that row vector from the task-subject cost matrix. This yields the effective cost matrix; The matrix verification function is used to verify each row and each column of the first task-subject cost matrix and the effective cost matrix to ensure that each row and each column of the matrix contains at least one finite cost value, and finally obtains the effective cost matrix. The matrix verification function is expressed as follows: ; in, Represents a matrix verification function; Represents the effective cost matrix. This represents the task-entity cost matrix.

[0014] Preferably, step S6 includes: Step S61: Obtain the dimensions of the effective cost matrix, i.e.: ;in, This represents the number of tasks in the effective cost matrix, and ; This represents the number of entities in the effective cost matrix, and ; Indicates dimension; Step S62: For dimensions Determine whether it is not greater than a preset scale threshold. If yes, proceed to step S63; otherwise, proceed to step S64. Step S63: Confirm all feasible solutions for executing all tasks based on the effective cost matrix, calculate the total cost of each feasible solution, and sort them in ascending order of total cost to form a candidate set. Perform feasibility verification on the feasible solutions in the candidate set in turn. If the current feasible solution fails the verification, continue to verify the next feasible solution with the second smallest total cost until the first feasible solution that passes the verification is found. This feasible solution is then used as the final task allocation scheme and output. Step S64: Set the starting loop index and create the current processing matrix. Simultaneously, an empty scheme queue is initialized; where the current processing matrix... The initial value is equal to the initial value of the effective cost matrix; The current processing matrix is ​​determined based on priority scores. Tasks are sorted in the order of high priority, medium priority, and low priority. When tasks have the same priority score, priority is determined by task value. Sort the tasks in descending order, as follows: The current processing matrix is ​​processed according to the order of the tasks. Sort the row vectors corresponding to the tasks to obtain the current processing matrix. ; Step S65: From the current processing matrix Get the first The row vector corresponding to the row The row vector is selected by the minimum value selection function. Compare all cost values ​​to find the minimum cost value. , is represented as: ; Among them, row vector That is, the number after priority sorting. Task Total cost for all optional entities; Indicates the first One entity; When multiple identical minimum cost values ​​exist At that time, select the minimum cost value with the highest priority according to the task's priority score. ; Determine the minimum cost value The corresponding paired units are added to the scheme queue, and then processed from the current processing matrix. Delete the row and number Columns, to obtain the current processing matrix ; Step S66: Process the current matrix Check if the dimension is greater than 1; if so, update the index. Return to step S65 to continue the next iteration until the final solution queue is obtained; otherwise, change the current processing matrix. The remaining unique element is added to the solution queue to obtain the final solution queue; Step S67: Perform a feasibility check on the final solution queue. If the check passes, output the final solution queue as the final task allocation solution. If the check fails, pair the corresponding costs in the effective cost matrix. Apply a penalty, then return to step S64 to restart the iteration; where, Representing the subject Execute the task The cost.

[0015] Preferably, the feasibility verification includes: Assignment integrity verification, meaning that each subject is assigned only one task; Coverage verification, i.e., the scope of operations covered by the solution. Not less than the preset task size threshold.

[0016] Preferably, in step S67, if the verification fails, the corresponding costs in the effective cost matrix are paired. Imposing punishments, including: When the allocation integrity check fails, i.e., when a subject is assigned two or more tasks, a penalty is imposed on the subsequent cost pairings that cause the subject to be repeatedly assigned in the effective cost matrix. When the coverage check fails, a penalty is imposed on the cost pair with the smallest job size in the effective cost matrix.

[0017] Preferably, when the coverage check fails, a penalty is imposed on at least two cost pairs in the effective cost matrix that have the smallest and second smallest job sizes in the scheme.

[0018] A multi-entity, multi-task planning and allocation system for buried space, the system being used to implement the aforementioned multi-entity, multi-task planning and allocation method for buried space, the system comprising: The data acquisition module acquires task data and subject data, constructs a task set and a subject set based on the task data and the subject data respectively, and filters the subject set to obtain a valid candidate subject set. The cost modeling and matrix construction module calculates the priority level and category of each task in the task set according to the decision indicators; performs resource type matching for each task according to the category of each task; performs subject matching for each task in the effective candidate subject set according to the result of resource type matching; and determines the execution content of each task according to the result of subject matching. The cost of each task that the subject can perform is calculated based on the content of the task, and a task-subject cost matrix is ​​constructed based on the cost. The task allocation and optimization module trims the task-subject cost matrix to obtain an effective cost matrix; based on the effective cost matrix, it solves and optimizes the final task allocation scheme with the minimum total cost, and allocates tasks according to the final task allocation scheme. The monitoring and replanning module monitors the task execution status, main resource status, and environment status for any abnormalities when executing tasks according to the final task allocation scheme. If any abnormalities are found, the module returns to the data acquisition module for replanning; otherwise, execution continues until the final task allocation scheme is completed.

[0019] The above technical solution can achieve at least some of the following technical effects: (1) The task execution is decomposed into three stages: “positioning-on-site operation-recovery”. The scale of operation, environmental complexity and risk level are introduced as evaluation factors for on-site operation costs. The simple measurement of “movement time” is transformed into the quantification of the complete operation process of “arrival, operation and withdrawal”. This makes the final task planning scheme more reflective of the real resource consumption and risks, effectively reduces the deviation between planning and execution, and significantly improves the accuracy and practical relevance of cost assessment.

[0020] (2) The task-subject cost matrix is ​​intelligently pruned by a matrix verification function, automatically deleting rows and columns that cannot be executed, thus eliminating a large amount of invalid search space in advance, significantly reducing computational complexity and meeting the real-time requirements of emergency response. At the same time, during the solution process, the business logic of "one subject can only execute one task within one wave" is strictly guaranteed by dynamically deleting the rows and columns corresponding to the assigned tasks and subjects, without the need for manual secondary adjustment, and can be directly used as the basis for execution.

[0021] (3) By dynamically selecting the solution path based on the size of the effective cost matrix, the optimal balance between solution performance and efficiency is achieved, overcoming the limitation of a single algorithm being unable to adapt to different emergency scenarios. For small-scale problems, the global optimal solution can be obtained with lower overhead, avoiding unnecessary iterations of the intelligent algorithm; for large-scale problems, a high-quality feasible solution can be obtained within an acceptable time through heuristic search, thus balancing the optimality of the solution and the real-time generation in various scenarios.

[0022] (4) During execution, the system continuously monitors the task status, main resources, and environmental changes, and pre-sets a multi-dimensional trigger rule base, including task failure, main resource shortage, sudden environmental risk changes, and obstructed recovery paths. Once triggered, the system immediately and automatically re-drives the entire process from data update and cost calculation to solution finding based on the latest global status data, generating a new solution and smoothly switching. This greatly enhances the system's robustness and adaptability in response to emergencies. This mechanism transforms the system from a static planning tool into a dynamic response system, effectively handling the uncertainties and dynamic changes at disaster sites, ensuring that rescue operations are carried out continuously and safely under control, and improving the overall success rate and continuity of rescue efforts. Attached Figure Description

[0023] Figure 1 This is a flowchart of the multi-subject, multi-task planning and allocation method for buried space described in the embodiments of this application; Figure 2 This is a schematic diagram illustrating the division of task execution content in an embodiment of this application. Detailed Implementation

[0024] The technical solution of this application will be described in detail below with reference to the accompanying drawings.

[0025] like Figure 1 As shown, the multi-subject, multi-task planning and allocation method for buried space described in this application includes: Step S1: Acquire task data and subject data, and construct task set and subject set according to the task data and subject data respectively; wherein, the parameters of each job in the task set include job unique identifier, job location, job area range, job area risk, physical field index, business index and decision index, and the parameters of each subject in the subject set include subject unique identifier, subject real-time status information, subject capability attributes and subject currently available resources.

[0026] In this embodiment, the construction of the task set includes: receiving target information from the sensing system or command input, combining real-time on-site sensing data with a multi-level indicator system, comprehensively evaluating each target to be rescued, and forming a task set... The task set consists of individual targets awaiting rescue. Each task A structured data object is represented as: ; ; in, This represents a unique identifier for the task; Indicates the work location; Indicates the scope of the work area; This indicates the risk in the work area, that is, the identified risk, indicated by risk labels. Attribute set and risk level The overall risk of a work area is derived from the aggregation of all identified risks. Indicates physical field index, Indicates business metrics, Indicates decision-making indicators.

[0027] The physical field indicators (primary indicators) are basic shape indicators of the work point extracted from real-time sensing data, including: Traffic accessibility: Based on the damage to surrounding roads and traffic routes, assess the traffic flow at the work site location and determine the accessibility level of the accessible roads. Environmental Hazards: Based on comprehensive preliminary detection data, secondary disaster risks, meteorological information, etc., the environmental risk level of the work site is quantitatively assessed; Life status and sustainability: Assess the possible survival status and sustainability window of buried life targets; Communication Possibility: Based on comprehensive analysis of communication signal data, assess the possibility of maintaining communication with the command system during operations; Search and rescue target size and distribution: Information on the number, density and depth of search and rescue targets within the physical topology range of the operation site.

[0028] The business metrics (secondary metrics) are quantitatively evaluated based on physical field indicators through weighting, reflecting the complexity and value of business implementation, including the complexity of task information acquisition, task value, and workload and difficulty of task implementation, respectively expressed as: ; ; ; in, Indicates the complexity of task information retrieval, and comprehensively , , The assessment of equipment capabilities (such as detection capabilities and adaptability to the buried environment) reflects the means and the level of supporting business organization required to achieve mission discovery. The value of a search and rescue operation is determined by comprehensively evaluating the benefits of the operation from the perspectives of the survival status, sustainability, and number of trapped individuals. It indicates the workload and difficulty of the operation, and the resource input requirements are assessed by comprehensively considering the operation area, operation medium, operation depth and distribution. , , , , , , These are all weighting coefficients and can be dynamically adjusted according to equipment capabilities; Indicates accessibility by transportation. Indicates environmental hazard. Indicates the possibility of communication. Indicates life status and sustainability. This indicates the size and distribution of the search and rescue targets.

[0029] The decision-making indicators (level 3 indicators) are obtained based on the business indicators and include task grading and task classification. Task grading combines task value and operational difficulty to form a priority ranking, including high priority, medium priority, and low priority, which serves as the basis for resource scheduling for task allocation. Task classification divides tasks into types based on information complexity and workload, including rapid reconnaissance, standard operations, and heavy engineering, matching different task execution methods.

[0030] Specifically, the task classification level is determined based on a priority score, which is expressed as follows: ; in, Indicates priority score, based on The value range of is used to divide tasks into high priority, medium priority, and low priority; , , All are weighting coefficients.

[0031] The specific categories of the task classification include: when and In such cases, the mission is classified as a rapid reconnaissance mission; when and In this case, the task is classified into standard operation categories; when and At that time, the task will be classified as a heavy engineering project; in, , , , All of these are preset classification thresholds.

[0032] If the task classification is unclear (i.e., it cannot be categorized into any of the above categories), the indicators will be reassessed or supplemented with sensing data. If the task classification does not meet the preset constraints (e.g., the priority score is below the minimum threshold), a decision will be made based on resource availability to whether to postpone processing or downgrade the task allocation.

[0033] The aforementioned real-time sensing data comes from the comprehensive aggregation and processing of on-site environmental monitoring (temperature, humidity, air pressure, hazardous gases, deformation, vibration, etc.), communication data analysis (simulation data, network management data, equipment heartbeat logs), space exploration data (images, photoelectric detection, ground penetrating radar), and basic geographic and spatial pre-construction data (space body structure data, local environment analysis).

[0034] Step S2: Filter the subject set to obtain a valid candidate subject set.

[0035] Preferably, step S2 includes: Step S21: Determine whether the current working status of the subject is dispatchable. If yes, proceed to step S22; otherwise, proceed to step S24.

[0036] Step S22: Determine whether the baseline cost from the subject to the task set is less than the subject's resource reservation value minus the cost tolerance value. If yes, proceed to step S23; otherwise, proceed to step S24. This determination is expressed as: ; in, Indicates the baseline cost. This represents the weighting coefficient determined based on the target's classification level. This indicates the estimated distance from the current location of the subject to the task set. Indicates the first One entity; This represents the cost calculation function, which is based on the estimated distance. and main body The subject's own ability attributes (such as speed and energy consumption) Estimate the baseline cost of performing the task. Representing the subject Current resource availability or remaining working capacity (such as remaining power and available working time). This indicates a safety margin, reserving some buffer for resources to ensure that the main body still has enough resources to perform tasks upon arrival.

[0037] Step S23: Determine whether the subject's ability attributes match the category of a certain task. If yes, proceed to step S25; otherwise, proceed to step S24.

[0038] Step S24: The subject is invalid; exclude the subject.

[0039] Step S25: The subject is a valid candidate subject, so the subject is included in the set of valid candidate subjects.

[0040] Step S3: Obtain the priority level and category of each task in the task set according to the decision indicators.

[0041] As mentioned above, priority scores are obtained based on decision indicators. Determine the priority level of each task and obtain its complexity based on the task information. In addition to assessing the workload and difficulty of the operation, the resource input requirements are determined by comprehensively considering the operation area, operating medium, operation depth, and distribution. Get the task category.

[0042] Tasks are matched to resource types based on their category. Based on the resource type matching results, subject matching is performed for each task within the valid candidate subject set. The execution content of each task is determined based on the subject matching results. The execution content includes the task's placement path, job path, and retrieval path, such as... Figure 2 As shown.

[0043] Specifically, based on the category of the target classification, it is determined whether the subject's ability attributes match the task type. The matching rules are shown in Table 1: Table 1. Matching Rules between Task Type and Subject Capabilities

[0044] Step S4: Calculate the cost of each task that the subject can perform based on the task's execution content, and construct a task-subject cost matrix based on the cost.

[0045] Preferably, step S4 includes: Step S41: Traverse the task set and the set of valid candidate subjects, and calculate the cost of each task that each subject can execute, expressed as: ; in, Representing the subject Execute the task The cost; Representing the subject Resource reservation value, This indicates the preset cost tolerance value; This represents the target priority correction factor, which is determined based on the priority score.

[0046] like =High, then This reduces the cost of high-priority tasks, making them more likely to be selected during allocation; if =in, then ;like =low, then This increases the cost of low-priority tasks, causing them to be delayed in allocation when resources are scarce. 、 、 These are configurable parameters. value range The default value can be dynamically adjusted according to the rescue strategy.

[0047] ; ; ; ; in, This refers to the cost of the positioning phase, which is the cost of the subject moving from its current location to the task area, taking into account time cost, distance cost, and path risk cost. , , All of these represent weighting coefficients, which can be dynamically adjusted according to the rescue strategy. This indicates the mission location, taking into account environmental constraints (such as obstacles, collapse areas, and risk areas). With the main position The shortest path distance between them. Representing the subject Normal moving speed; This represents the cumulative risk value of the path, calculated by weighting the risk level of the risk areas the path traverses with its length. Representing the subject Ability attributes.

[0048] This refers to the cost of the operation phase, which is the cost incurred by the entity in performing operations within the task area. It takes into account factors such as the scale of the operation, environmental complexity, the entity's operational capabilities, and the risks of the operation area. , , All of these represent weighting coefficients, which can be dynamically adjusted according to the task type and rescue priority. Indicates the size of the work area (such as area, volume, or the amount of earthwork to be processed); Indicates the operational capacity of the main unit (such as the operational area or earthwork volume that can be processed per unit time). Environmental complexity factors (such as the density of debris accumulation, the narrowness of the working space, visibility, etc.) are derived from a comprehensive evaluation of on-site perception data. This indicates the risk level of the work area (such as structural instability risk, concentration of harmful gases, probability of secondary collapse, etc.).

[0049] This represents the cost of the recovery phase, which is the cost incurred by the entity moving from the task area to the recovery point after completing the task. , , All of these represent weighting coefficients, which can be dynamically adjusted based on the urgency of recycling and safety requirements. This indicates the mission location, taking environmental constraints into account. Location of recycling points The shortest path distance between them; This represents the cumulative risk value of the recycling path.

[0050] Step S42: Arrange all calculated costs into an n×m initial task-subject cost matrix; where n represents the number of tasks in the task set, and m represents the number of valid candidate subjects, expressed as: .

[0051] Step S5: Trim the task-subject cost matrix to obtain the effective cost matrix.

[0052] Preferably, step S5 includes: Step S51: For any column vector in the task-subject cost matrix If all its elements are infinite, then the column vector is... Corresponding subject If the cost of all executable tasks is infinite, then according to the matrix verification function, the corresponding column vector is removed from the task-body cost matrix. This yields the first task-main cost matrix.

[0053] Step S52: For any row vector in the first task-main cost matrix If all its elements are infinite, it means that all entities execute this row vector. Corresponding task If the costs are all infinite, then according to the matrix verification function, delete that row vector from the task-subject cost matrix. This yields the effective cost matrix.

[0054] The matrix verification function is used to verify each row and each column of the first task-subject cost matrix and the effective cost matrix to ensure that each row and each column of the matrix contains at least one finite cost value, thus obtaining the effective cost matrix.

[0055] The matrix verification function is expressed as follows: ; in, This represents the matrix verification function, which takes the initial task-subject cost matrix as input and outputs a trimmed effective cost matrix. Represents the effective cost matrix. This represents the task-entity cost matrix.

[0056] Step S6: Based on the effective cost matrix, solve for the final task allocation scheme with the minimum total cost, and allocate tasks according to the final task allocation scheme.

[0057] Preferably, step S6 includes: Step S61: Obtain the dimensions of the effective cost matrix, i.e.: ;in, This represents the number of tasks in the effective cost matrix, and ; This represents the number of entities in the effective cost matrix, and ; Indicates dimension.

[0058] Step S62: For dimensions Determine whether the value is not greater than a preset threshold. If yes, proceed to step S63; otherwise, proceed to step S64.

[0059] Specifically, a configurable scale threshold can be preset. (The specific value range can be dynamically adjusted according to the system's computing power and real-time requirements.) and Based on the comparison results, an adaptive solution strategy is selected.

[0060] Step S63: Confirm all feasible solutions for executing all tasks based on the effective cost matrix, calculate the total cost of each feasible solution, and sort them in ascending order of total cost to form a candidate set. Perform feasibility verification on the feasible solutions in the candidate set in turn. If the current feasible solution fails the verification, continue to verify the next feasible solution with the second smallest total cost until the first feasible solution that passes the verification is found. This feasible solution is then used as the final task allocation solution and output.

[0061] Preferably, the feasibility verification includes: Assignment integrity verification, meaning that each subject is assigned only one task; Coverage verification, i.e., the scope of operations covered by the solution. Not less than the preset task size threshold.

[0062] A feasible solution passes the verification if and only if all feasible solutions pass the allocation integrity verification and coverage verification; otherwise, the feasible solution fails the verification.

[0063] Step S64: Set the starting loop index and create the current processing matrix. Simultaneously, an empty scheme queue is initialized; where the current processing matrix... The initial value is equal to the initial value of the effective cost matrix; The current processing matrix is ​​determined based on priority scores. Tasks are sorted in the order of high priority, medium priority, and low priority. When tasks have the same priority score, priority is determined by task value. Sort the tasks in descending order, as follows: The current processing matrix is ​​processed according to the order of the tasks. Sort the row vectors corresponding to the tasks in the matrix (i.e., the row vector corresponding to the task with the first sorted position is placed in the first row of the matrix, and so on), to obtain the current processing matrix. .

[0064] Step S65: From the current processing matrix Get the first The row vector corresponding to the row The row vector is selected by the minimum value selection function. Compare all cost values ​​to find the minimum cost value. , is represented as: ; Among them, row vector That is, the number after priority sorting. Task Total cost for all optional entities; Indicates the first One entity; When multiple identical minimum cost values ​​exist At that time, select the minimum cost value with the highest priority according to the task's priority score. ; Determine the minimum cost value The corresponding paired units are added to the scheme queue, and then processed from the current processing matrix. Delete the row and number Columns, to obtain the current processing matrix .

[0065] Step S66: Process the current matrix Check if the dimension is greater than 1; if so, update the index. Return to step S65 to continue the next iteration until the final solution queue is obtained; otherwise, change the current processing matrix. The remaining unique element is added to the solution queue to obtain the final solution queue.

[0066] Step S67: Perform a feasibility check on the final solution queue. If the check passes, output the final solution queue as the final task allocation solution. If the check fails, pair the corresponding costs in the effective cost matrix. Apply a penalty, then return to step S64 to restart the iteration; where, Representing the subject Execute the task The cost.

[0067] Preferably, in step S67, if the verification fails, the corresponding costs in the effective cost matrix are paired. Imposing punishments, including: (1) When the allocation integrity check fails, i.e., when a subject is assigned two or more tasks, a penalty is imposed in the effective cost matrix on subsequent cost pairs that lead to the subject being repeatedly assigned. For example, subject Simultaneously assigned to , and During the iteration process, if the order of cost pairing is selected as follows: , , Then, in the effective cost matrix, cost pairing is performed. and Apply penalties (e.g., assign a weight greater than 1 to the cost pair or make the cost pair infinite) to increase the cost of the two subsequently assigned cost pairs, making them less likely to be selected in subsequent iterations.

[0068] (2) When the coverage check fails, a penalty is imposed on the cost pair with the smallest job size in the effective cost matrix for that scheme. For example, the task with the smallest job size in the scheme is... Then for All corresponding cost pairs are penalized to increase their costs.

[0069] Preferably, when the coverage check fails, a penalty is imposed on at least two cost pairs in the effective cost matrix that have the smallest and second smallest task size in the scheme. For example, the task with the smallest task size in the scheme is... The second smallest task is Then for and The corresponding cost pairs are all penalized to increase their costs.

[0070] In this embodiment, after obtaining the final task allocation scheme, a structured task planning scheme is generated based on the final task allocation scheme. Since the final task allocation scheme defines the mapping relationship between the task set and the effective candidate subject set in the current task wave, a structured task planning scheme is generated based on this mapping relationship. The structured task planning scheme is a composite data object, including at least the following elements: (1) Task Assignment List: Clearly recorded as each task One or more assigned entities (forming task teams) form a pairing list. .

[0071] (2) Estimated total cost per wave: Based on the final task allocation scheme, from the effective cost matrix Extract all cost pairs and sum them to obtain the estimated total cost of this mission. .

[0072] (3) Task Chains for Each Participating Participant: For each participating entity in the task This clarifies the specific tasks assigned to them in this wave.

[0073] For each subject in the structured task planning scheme This generates specific path planning for the user, from the current location to the mission area, and after completing the mission, to the recovery point. Specifically, this includes: (1) In-situ path planning: based on the main body Starting from the current position and ending at the position of the assigned task, a time- or distance-cost-optimized path is planned, taking into account environmental constraints (such as obstacles and risk areas). This path planning can be implemented using known algorithms, such as A* algorithm, Dijkstra's shortest path algorithm, RRT (Rapidly-exploring Random Tree) algorithm, and Artificial Potential Field method, and path optimization is performed in conjunction with environmental risk areas, obstacle distribution, and dynamic constraints.

[0074] (2) Pre-planning of recovery path: Starting from the assigned task location and ending at the predetermined recovery point or the next standby point, and taking into account environmental constraints, a recovery path is pre-planned. This recovery path will serve as the default withdrawal route after the main body completes the operation.

[0075] Finally, the structured task planning scheme and the specific path plans of each entity are integrated and encapsulated to form a final executable planning instruction set for the command system and the task entities. The specific process is as follows: (1) Instruction encapsulation: for each task body Generate its unique task instruction package, which includes at least: details of the assigned task objective, the placement path, the pre-planned retrieval path, and the constraints and expected parameters for task execution.

[0076] (2) Generate a global task planning summary, summarizing the task allocation relationships, total cost, status of each entity, and overall timeline expectations for this wave.

[0077] (3) Output and distribution: The task instruction packages are distributed to the corresponding task subject (or its control system) through the preset standardized business interface.

[0078] Step S7: When executing a task according to the final task allocation scheme, monitor the task execution status, main resource status, and environment status. If the task status, main resource status, or environment status is abnormal, return to step S1 for replanning.

[0079] In this embodiment of the application, the following elements are collected and aggregated in real time or periodically: (1) Task execution status: Monitor the current location, work progress, and task completion status of each task subject.

[0080] (2) Main resource status: Monitor the remaining resource reserves of each task subject. Health status and communication status.

[0081] (3) Environmental status: By accessing real-time sensing data (such as point cloud, vibration, gas monitoring), monitor the structural deformation of the work space, the emergence of new risk points, or newly discovered targets to be rescued.

[0082] Based on the monitored status data, automatic judgment is made according to the preset replanning trigger rules, including: (1) Task failure or serious obstruction: The subject is unable to reach the task area or confirms that the assigned task cannot be completed.

[0083] (2) Abnormal status of the main body: The main body's resources are about to be exhausted. (Below the safety threshold), malfunction, or loss of connection.

[0084] (3) Significant environmental changes or newly discovered high-priority targets: The sensing system detects new structural risks in the original planned route or work area, or newly discovered targets to be rescued, and the priority determination of the target meets one of the following conditions: (3.1) The new target =High (high priority); (3.2) New Targets A higher priority score than any currently executing task.

[0085] For newly discovered high-priority targets, the system adds them to the task set and automatically proceeds to step S1 for replanning, ensuring that high-priority targets can receive resource allocation in a timely manner.

[0086] (4) Task recovery phase anomaly: During the task recovery phase, the subject's current state Unable to meet the pre-set cost constraints of the recycling path They need to independently plan new recycling solutions.

[0087] (5) Command instruction intervention: Receive manual replanning instructions from the commander, especially when the commander judges that the task priority allocation needs to be adjusted based on the situation on the ground.

[0088] When any of the aforementioned triggering rules are met, a replanning trigger signal is generated, and all the latest task data, subject status, and environmental data are collected as input for replanning. In particular, for replanning triggered by the discovery of a new high-priority target, the priority and classification information of that target must also be included so that it can be processed first during the replanning process.

[0089] Based on the above-mentioned multi-subject, multi-task planning and allocation method for burial space, this application provides a multi-subject, multi-task planning and allocation system for burial space. The system includes a data acquisition module, a cost modeling and matrix construction module, a task allocation and optimization module, and a monitoring and replanning module.

[0090] The data acquisition module is used to acquire task data and subject data, and construct task sets and subject sets based on the task data and subject data respectively; the subject sets are filtered to obtain valid candidate subject sets.

[0091] The cost modeling and matrix construction module is used to calculate the priority level and category of each task in the task set according to the decision indicators; to perform resource type matching for each task according to the category of each task; to perform subject matching for each task in the effective candidate subject set according to the result of resource type matching; and to determine the execution content of each task according to the result of subject matching. The cost of each task that the subject can perform is calculated based on the content of the task, and a task-subject cost matrix is ​​constructed based on the cost.

[0092] The task allocation and optimization module is used to trim the task-subject cost matrix to obtain an effective cost matrix; based on the effective cost matrix, it solves and optimizes the final task allocation scheme with the minimum total cost, and allocates tasks according to the final task allocation scheme.

[0093] The monitoring and replanning module is used to monitor the task execution status, main resource status, and environment status for any abnormalities when executing tasks according to the assigned final task allocation scheme. If any abnormalities are found, the module returns to the data acquisition module for replanning; otherwise, execution continues until the final task allocation scheme is completed.

[0094] The above are exemplary embodiments of this application, and the scope of protection of this application is defined by the claims and their equivalents.

Claims

1. A method for multi-subject, multi-task planning and allocation of buried space, characterized in that, include: Step S1: Acquire task data and subject data, and construct task set and subject set according to the task data and subject data respectively; wherein, the parameters of each job in the task set include job unique identifier, job location, job area range, job area risk, physical field index, business index and decision index, and the parameters of each subject in the subject set include subject unique identifier, subject real-time status information, subject capability attributes and subject currently available resources; Step S2: Filter the subject set to obtain a valid candidate subject set; Step S3: Obtain the priority level and category of each task in the task set according to the decision indicators; Based on the category of each task, resource type matching is performed on the task. Based on the result of resource type matching, subject matching is performed on each task in the valid candidate subject set. Based on the result of subject matching, the execution content of each task is determined. The execution content includes the task's placement path, operation path, and retrieval path. Step S4: Calculate the cost of each task that the subject can perform based on the content of the task, and construct a task-subject cost matrix based on the cost; Step S5: Trim the task-subject cost matrix to obtain the effective cost matrix; Step S6: Based on the effective cost matrix, solve for the final task allocation scheme with the minimum total cost, and allocate tasks according to the final task allocation scheme; Step S7: When executing a task according to the final task allocation scheme, monitor whether there are any abnormalities in the task execution status, main resource status, and environment status. If so, return to step S1 for replanning; otherwise, continue execution until the final task allocation scheme is completed.

2. The multi-subject, multi-task planning and allocation method for buried space as described in claim 1, characterized in that, In step S1, the physical field indicators include traffic accessibility, environmental hazard, life status and sustainability, communication possibility, and the size and distribution of search and rescue targets. The business metrics are obtained based on the physical field metrics and include: the complexity of task information acquisition, task value, and the workload and difficulty of task implementation, which are expressed as follows: ； ； ； in, This indicates the complexity of obtaining task information. Indicates the value of the task. This indicates the workload and difficulty of the task. , , , , , , All are weighting coefficients. Indicates accessibility by transportation. Indicates environmental hazard. Indicates the possibility of communication. Indicates life status and sustainability. This indicates the scale and distribution of the search and rescue mission; The decision indicators are obtained based on the business indicators, including task grading and task classification. The task grading levels include high priority, medium priority and low priority, and the task classification categories include rapid reconnaissance, standard operation and heavy engineering. The task classification level is determined based on a priority score, which is expressed as: ； in, Indicates priority score, based on The range of values ​​is used to divide tasks into high priority, medium priority, and low priority. , , All are weighting coefficients; The task classification categories include: when and In such cases, the mission is classified as a rapid reconnaissance mission; when and In this case, the task is classified into standard operation categories; when and At that time, the task will be classified as a heavy engineering project; in, , , , All of these are preset classification thresholds.

3. The multi-subject, multi-task planning and allocation method for buried space as described in claim 2, characterized in that, Step S2 includes: Step S21: Determine whether the current working status of the subject is dispatchable. If yes, proceed to step S22; otherwise, proceed to step S24. Step S22: Determine whether the baseline cost of the subject to the task set is less than the resource reservation value minus the cost tolerance value of the subject. If yes, proceed to step S23; otherwise, proceed to step S24. Step S23: Determine whether the subject's ability attributes match the category of a certain task. If yes, proceed to step S25; otherwise, proceed to step S24. Step S24: The subject is invalid; exclude the subject. Step S25: The subject is a valid candidate subject; therefore, the subject is included in the set of valid candidate subjects. The benchmark cost is expressed as: ,in, This represents the weighting coefficient determined based on the target's classification level. This indicates the estimated distance from the current location of the subject to the task set. Indicates the first One entity; This represents the cost calculation function, which is based on the estimated distance. and main body The subject's own ability attributes Estimate the baseline cost of performing the task.

4. The multi-subject, multi-task planning and allocation method for buried space as described in claim 3, characterized in that, Step S4 includes: Step S41: Traverse the task set and the set of valid candidate subjects, and calculate the cost of each task that each subject can execute, expressed as: ； in, Representing the subject Execute the task The cost; This represents the target priority correction factor, which is determined based on the priority score. Representing the subject Resource reservation value, This indicates the preset cost tolerance value; ； ； ； ； This indicates the cost during the placement phase. , , All represent weighting coefficients; This indicates the mission location, taking environmental constraints into account. With the main position The shortest path distance between them; Representing the subject Normal moving speed; This represents the cumulative risk value of the path; Representing the subject Ability attributes; Indicates the cost of the activity phase. , , All represent weighting coefficients; Indicates the size of the work area; Indicates the operational capability of the main unit; This represents the environmental complexity factor; Indicates the risk level of the work area; Indicates the cost during the recovery phase. , , All represent weighting coefficients; This indicates the mission location, taking environmental constraints into account. Location of recycling points The shortest path distance between them; This represents the cumulative risk value of the recovery path; Step S42: Arrange all the calculated costs into an n×m initial task-subject cost matrix; where n represents the number of tasks in the task set and m represents the number of valid candidate subjects.

5. The multi-subject, multi-task planning and allocation method for buried space as described in claim 4, characterized in that, Step S5 includes: Step S51: For any column vector in the task-subject cost matrix If all its elements are infinite, then the column vector is... Corresponding subject If the cost of all executable tasks is infinite, then according to the matrix verification function, the corresponding column vector is removed from the task-body cost matrix. This yields the first task-main cost matrix; Step S52: For any row vector in the first task-main cost matrix If all its elements are infinite, it means that all entities execute this row vector. Corresponding task If the costs are all infinite, then according to the matrix verification function, delete that row vector from the task-subject cost matrix. This yields the effective cost matrix; The matrix verification function is used to verify each row and each column of the first task-subject cost matrix and the effective cost matrix to ensure that each row and each column of the matrix contains at least one finite cost value, and finally obtains the effective cost matrix. The matrix verification function is expressed as follows: ； in, Represents a matrix verification function; Represents the effective cost matrix. This represents the task-entity cost matrix.

6. The multi-subject, multi-task planning and allocation method for buried space as described in claim 5, characterized in that, Step S6 includes: Step S61: Obtain the dimensions of the effective cost matrix, i.e.: ;in, This represents the number of tasks in the effective cost matrix, and ; This represents the number of entities in the effective cost matrix, and ; Indicates dimension; Step S62: For dimensions Determine whether it is not greater than a preset scale threshold. If yes, proceed to step S63; otherwise, proceed to step S64. Step S63: Confirm all feasible solutions for executing all tasks based on the effective cost matrix, calculate the total cost of each feasible solution, and sort them in ascending order of total cost to form a candidate set. Perform feasibility verification on the feasible solutions in the candidate set in turn. If the current feasible solution fails the verification, continue to verify the next feasible solution with the second smallest total cost until the first feasible solution that passes the verification is found. This feasible solution is then used as the final task allocation scheme and output. Step S64: Set the starting loop index and create the current processing matrix. Simultaneously, an empty scheme queue is initialized; where the current processing matrix... The initial value is equal to the initial value of the effective cost matrix; The current processing matrix is ​​determined based on priority scores. Tasks are sorted in the order of high priority, medium priority, and low priority. When tasks have the same priority score, priority is determined by task value. Sort the tasks in descending order, as follows: The current processing matrix is ​​processed according to the order of the tasks. Sort the row vectors corresponding to the tasks to obtain the current processing matrix. ; Step S65: From the current processing matrix Get the first The row vector corresponding to the row The row vector is selected by the minimum value selection function. Compare all cost values ​​to find the minimum cost value. , represented as: ； Among them, row vector That is, the number after priority sorting. Task Total cost for all optional entities; Indicates the first One entity; When multiple identical minimum cost values ​​exist At that time, select the minimum cost value with the highest priority according to the task's priority score. ; Determine the minimum cost value The corresponding paired units are added to the scheme queue, and then processed from the current processing matrix. Delete the row and number Columns, to obtain the current processing matrix ; Step S66: Process the current matrix Check if the dimension is greater than 1; if so, update the index. Return to step S65 to continue the next iteration until the final solution queue is obtained; otherwise, change the current processing matrix. The remaining unique element is added to the solution queue to obtain the final solution queue; Step S67: Perform a feasibility check on the final solution queue. If the check passes, output the final solution queue as the final task allocation solution. If the check fails, pair the corresponding costs in the effective cost matrix. Apply a penalty, then return to step S64 to restart the iteration; where, Representing the subject Execute the task The cost.

7. The multi-subject, multi-task planning and allocation method for buried space as described in claim 6, characterized in that, The feasibility verification includes: Assignment integrity verification, meaning that each subject is assigned only one task; Coverage verification, i.e., the scope of operations covered by the solution. Not less than the preset task size threshold.

8. The method for multi-subject, multi-task planning and allocation of buried space as described in claim 7, characterized in that, In step S67, if the verification fails, the corresponding costs in the effective cost matrix are paired. Imposing punishments, including: When the allocation integrity check fails, i.e., when a subject is assigned two or more tasks, a penalty is imposed on the subsequent cost pairings that cause the subject to be repeatedly assigned in the effective cost matrix. When the coverage check fails, a penalty is imposed on the cost pair with the smallest job size in the effective cost matrix.

9. The multi-subject, multi-task planning and allocation method for buried space as described in claim 8, characterized in that, When the coverage check fails, a penalty is imposed on at least two cost pairs in the effective cost matrix that have the smallest and second smallest job sizes in the scheme.

10. A multi-subject, multi-task planning and allocation system for buried space, the system being used to implement the multi-subject, multi-task planning and allocation method for buried space as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition module acquires task data and subject data, constructs a task set and a subject set based on the task data and the subject data respectively, and filters the subject set to obtain a valid candidate subject set. The cost modeling and matrix construction module calculates the priority level and category of each task in the task set according to the decision indicators; performs resource type matching for each task according to the category of each task; performs subject matching for each task in the effective candidate subject set according to the result of resource type matching; and determines the execution content of each task according to the result of subject matching. The cost of each task that the subject can perform is calculated based on the content of the task, and a task-subject cost matrix is ​​constructed based on the cost. The task allocation and optimization module trims the task-subject cost matrix to obtain an effective cost matrix; based on the effective cost matrix, it solves and optimizes the final task allocation scheme with the minimum total cost, and allocates tasks according to the final task allocation scheme. The monitoring and replanning module monitors the task execution status, main resource status, and environment status for any abnormalities when executing tasks according to the final task allocation scheme. If any abnormalities are found, the module returns to the data acquisition module for replanning; otherwise, execution continues until the final task allocation scheme is completed.

Images