A task structure-aware multi-region job resource configuration system and method, electronic equipment and storage medium

Abstract

The present application relates to the technical field of multi-region operation resource configuration and intelligent decision system, and provides a task structure-aware multi-region operation resource configuration system and method, an electronic device and a storage medium.The method acquires region topology, facility state, risk early warning, task target, historical feedback and resource constraint according to a period, constructs a region graph and node state;analyzes the task structure, distinguishes the direct task area and the support task area, generates the primary and secondary yield and the mask;the strategy model outputs the low-dimensional control parameter, dynamically fuses the primary and secondary yield and projects to the resource feasible region, forms the configuration result and the operation instruction, and then updates the next period state according to the feedback.The present application takes into account the direct and support yield, reduces the strategy output dimension, improves the constraint satisfaction and the executability.

Description

Technical Field

[0001] This invention relates to the field of multi-regional operation resource allocation and intelligent decision-making system technology, and in particular to a task structure-aware multi-regional operation resource allocation system and method, electronic device and storage medium, and a method and system for generating multi-regional operation resource allocation results based on primary return rate, secondary return rate and fusion weight. Background Technology

[0002] In multi-regional operation scenarios such as emergency repair, inspection and maintenance, facility support, service restoration and material delivery, the operation objects are distributed across multiple regional nodes. The system needs to generate resource configuration schemes according to the regional topology, facility status, risk warning, task objectives, historical feedback and operation resource constraints on a configuration cycle.

[0003] In this type of scenario, there are road connections, communication connections, spatial adjacencies, or security dependencies between regional nodes. The node status changes with execution feedback, and the total amount of operational resources, resource type capacity, node reachability, and minimum task guarantee requirements are all limited. Operational resources include at least one of the following: operational personnel, operational vehicles, operational equipment, support materials, drones, or sensors.

[0004] Existing solutions typically determine resource deployment order based on a single priority field or human experience, making it difficult to simultaneously express the key processing needs of the direct task area and the supporting role of the supporting area in the chain of road accessibility, communication connectivity, resource arrival, or service restoration. Solutions that directly output the resource allocation of each regional node also suffer from the problem that the strategy output dimension increases with the number of regional nodes, and the configuration results need to additionally handle multiple types of operational resource constraints.

[0005] Therefore, a technical solution is needed that integrates with a multi-regional operation resource allocation system to uniformly process complex task objectives, regional map structures, main task benefits, supporting task benefits, low-dimensional control parameters, and operation resource constraints, and output multi-regional operation resource allocation results that can be converted into operation instructions within the configuration cycle. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a task structure-aware multi-region job resource allocation system and method, electronic device, and storage medium. It resolves the issues of difficulty in simultaneously identifying direct task regions and supporting task regions, difficulty in uniformly expressing primary and supporting benefits, the increasing dimension of policy output with the number of region nodes, and the difficulty in converting configuration results into job resource constraints and job instructions.

[0007] The technical solution of this invention is as follows:

[0008] In a first aspect, the present invention provides a multi-regional job resource allocation method with task structure awareness, comprising the following steps:

[0009] S1. Call various interfaces and databases according to the configuration cycle to collect topology, facilities, risks, tasks, feedback and resource constraints data of multi-region operations, build regional graph structure and generate regional node status;

[0010] S2. Divide the direct task region and the supporting task region according to the complex task objective and the region map structure, and generate the main return rate, the secondary return rate and the supporting region mask.

[0011] S3. With the goal of maximizing the overall benefit of the current configuration cycle operation and taking into account the cumulative benefit of multiple configuration cycles, encode the regional graph structure, regional node status and historical feedback, and output low-dimensional control parameters.

[0012] S5. Perform dynamic primary and secondary return rate fusion decision based on the low-dimensional control parameters, primary return rate, secondary return rate and support area mask to generate a multi-region resource allocation scheme for the current allocation period.

[0013] S6. Perform job resource constraint verification and constraint projection on the multi-region resource configuration scheme, convert the constraint-projected scheme into job instructions and send them to the job terminal, job management platform or equipment control terminal, and update the regional node status for the next configuration cycle based on the execution feedback.

[0014] Further optimization is made in step S1, where the various interfaces and databases include IoT interfaces, GIS database interfaces, business system interfaces, or resource ledger databases.

[0015] The multi-source data includes regional topology, facility status, risk warning, complex task objectives, historical feedback, and operational resource constraint data; wherein the operational resource constraint data includes total operational resources, resource type capacity, regional configuration upper limit, node reachability constraints, and minimum task guarantee requirements.

[0016] The expression for the region map structure is:

[0017]

[0018] in For a set of region nodes, It is at least one of the following: road connection, communication connection, spatial adjacency, or security dependency. This is the region connectivity strength matrix;

[0019] The status of the regional nodes includes regional attributes, facility status, target value, resource accessibility, risk or demand status, resource allocation in the previous cycle, and execution feedback.

[0020] Further optimization, step S2 includes the following steps:

[0021] S2.1, From the perspective of complex task objectives Get the set of direct target regions Direct target type set and task weight And from the state matrix of all region nodes Obtain the region identifier of each region node Node type Target Value Risk or level of disruption and demand pressure ;

[0022] S2.2 Construct the task matching function as follows: And simultaneously calculate the task matching degree vector for all regional nodes. The Each component :

[0023]

[0024] in, , , and Normalize to [0,1], and , , ;

[0025] S2.3, Based on the task matching degree vector and support connection strength matrix Calculate the structural support vector and determine the direct task area. Supported task areas and support region mask vector :

[0026]

[0027]

[0028] in, This is the threshold for the direct task region. To support task region thresholds, Indicates the first The region node pairs with the first The supporting connection strength of each regional node; This is the task matching degree vector. This is the structural support vector. Indicates the first Structural support of each regional node and For the set of region node indexes, To support region mask vectors.

[0029] S2.4 Calculate the main task contribution item according to the following formula. Sub-task contribution items :

[0030]

[0031] in, Indicates connection reliability. Indicates path reachability. Indicates the importance of the supporting object. and These are the non-negative weighting coefficients for the contribution items of the main task and the secondary task, respectively. ,and , , , ;

[0032] S2.5, Calculate the main return rate using the following formula. and yield :

[0033]

[0034] in, Main revenue marginal cost, For the marginal cost of secondary revenue, and This indicates the incremental resource amount required for the corresponding revenue branch. and Indicates arrival time. and Indicates the difficulty of execution. This is the cost-weighted coefficient. ,and , To prevent zero threshold.

[0035] Further optimization, step S3 includes the following steps:

[0036] S3.1 Add self-loops to the region connectivity strength matrix and perform row normalization to obtain the normalized connectivity matrix. Then, use the normalized connectivity matrix to perform neighborhood aggregation on the region node state matrix.

[0037] S3.2. By using summation pooling, average pooling, or attention pooling, the features of the aggregated neighborhood region graph and the total amount of job resources are analyzed. and historical feedback characteristics Synthesized into a global code independent of the number of regional nodes. :

[0038]

[0039] Where pool is one of summation pooling, average pooling or attention pooling, and ⊕ represents feature concatenation; This indicates the global code for the current configuration period. For the normalized connectivity matrix, This is the state matrix of the region nodes. Total amount of operational resources This is a historical feedback characteristic.

[0040] S3.3, Based on candidate resource allocation schemes Total job revenue generated during the current configuration period Cumulative returns over multiple allocation periods To achieve this goal, we construct quantitative evaluation criteria for the merits of resource allocation schemes, including:

[0041]

[0042] in, These are the candidate resource allocation schemes to be evaluated. This refers to the feasible domain for job resource configuration constructed based on job resource constraint data. To constrain deviations or impose penalties; and They represent the first The main task contribution items and secondary task contribution items of each regional node. For the current single-cycle operation's overall revenue, From Start and back The discounted cumulative return over each configuration period, The discount factor is... The cumulative revenue evaluation window length.

[0043] S3.4, Strategy Model According to global encoding Output raw control quantity And mapped to low-dimensional control parameters as follows:

[0044]

[0045] in, , , Focus on strength of main yield. The activation threshold for primary yield. To cover the temperature for secondary yield, Primary and secondary fusion coefficients; Maximize through historical configuration records or simulation samples It is obtained through training for the target, or determined by a pre-set rule table or linear rules. , , and The original control variables output by the strategy model correspond to the primary return focus intensity, primary return activation threshold, secondary return coverage temperature, and primary-secondary fusion coefficient, respectively. and This is a preset positive lower limit.

[0046] Further optimization is achieved in step S4, based on the main return activation threshold. and the main return focus intensity Generate main return hot spot weights :

[0047] ;

[0048] Based on the secondary yield coverage temperature and the supported region mask Generate supporting coverage weights :

[0049] The resource allocation vectors of all regional nodes are generated by merging them according to the following formula. :

[0050]

[0051] in, For the first One portion, Primary and secondary fusion coefficients This represents the total number of job resources for the current configuration period. This represents the total resource allocation for all nodes in the region. , , and They are vectors , , and The There are 1 component, of which , , For weighted components, For the first Resource allocation for each regional node This is the sum of the fusion weights of all regional nodes.

[0052] when Output an all-zero resource allocation vector when; , or Less than When the maximum main return region distribution, the uniform distribution of the reachable support region, or the uniform distribution of all reachable regions are used as the fallback distribution, non-negative pruning and total amount verification are performed on the fusion weights.

[0053] Further optimization, step S5 includes the following steps:

[0054] S5.1 Construct the feasible region for job resource configuration based on job resource constraint data. The feasible domain for configuring the job resources These include non-negativity constraints, single-cycle total amount constraints, resource type capacity constraints, regional configuration upper limit constraints, node reachability constraints, and minimum task guarantee constraints.

[0055] S5.2, The multi-regional resource allocation scheme is divided into... Project to the feasible domain of job resource configuration This yields the results of multi-regional operational resource allocation; among them, The constraint projection operator includes zeroing the unreachable region node mask, truncating the region configuration upper limit, normalizing the resource type capacity, supplementing the minimum guarantee for tasks, and redistributing the remaining job resources according to the region configuration weight.

[0056] S5.3 Convert the multi-region operation resource configuration results into operation instructions containing region node number, resource type, resource quantity or execution priority;

[0057] S5.4. Update the regional node status for the next configuration cycle based on the execution feedback, periodic operation revenue, or cumulative revenue changes returned by the operation terminal, operation management platform, or equipment control terminal.

[0058] Secondly, the present invention provides a task structure-aware multi-region job resource allocation system for performing any step of a task structure-aware multi-region job resource allocation method, including:

[0059] The data acquisition module is used to acquire regional operation status data and operation resource constraint data;

[0060] The region map construction module is used to construct the region map structure;

[0061] The task structure parsing module is used to determine the direct task region, the supporting task region, and the supporting region mask;

[0062] The primary and secondary rate of return calculation module is used to generate primary and secondary rates of return based on task matching degree, structural support degree, primary task contribution, secondary task contribution, and marginal cost.

[0063] The strategy model module is used to output the original control quantity based on the global encoding and map it to low-dimensional control parameters that are independent of the number of regional nodes.

[0064] The primary and secondary fusion allocation module is used to generate a regional resource allocation weight vector based on low-dimensional control parameters, primary rate of return, secondary rate of return, and support region mask.

[0065] The feasible region constraint module is used to construct the feasible region for job resource configuration;

[0066] The constraint projection output module is used to generate multi-region operation resource configuration results that satisfy the feasible domain of operation resource configuration and output operation instructions;

[0067] The feedback update module is used to update the regional node status for the next configuration period based on execution feedback, periodic job revenue, or cumulative revenue.

[0068] Thirdly, the present invention provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of a task structure-aware multi-region job resource allocation method.

[0069] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements any step of a task structure-aware multi-region job resource allocation method.

[0070] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0071] 1. By using complex task objectives and regional map structures to jointly identify direct task areas and supporting task areas, the resource allocation results can simultaneously reflect key disposal needs and structural support needs.

[0072] 2. The direct task revenue and the supporting task revenue are expressed by the primary rate of return and the secondary rate of return, respectively. The regional resource allocation weight is generated by the primary-secondary fusion coefficient to improve the adaptability of resource allocation to complex task structures.

[0073] 3. The policy model outputs low-dimensional control parameters that are independent of the number of regional nodes, reducing the dimensionality of the policy output and the computational load of inference.

[0074] 4. Utilize the feasible domain and constraint projection of job resource configuration to process non-negative constraints, total constraints, resource type capacity constraints, regional configuration upper limit constraints, node reachability constraints, and task minimum guarantee constraints, so that the configuration results can be transformed into job instructions;

[0075] 5. Update the regional node status for the next configuration cycle based on the execution feedback from the work terminal, work management platform, or equipment control terminal to form a dynamic closed loop. Attached Figure Description

[0076] Figure 1 This is a flowchart of the multi-regional job resource allocation method for task structure awareness according to the present invention.

[0077] Figure 2 This is a structural block diagram of the multi-regional operation resource allocation system with task structure awareness according to the present invention. Detailed Implementation

[0078] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0079] A multi-regional operational resource allocation method with task structure awareness, in this embodiment, operational resources include operational personnel, emergency repair vehicles, maintenance equipment, pumping equipment, support materials, drones, and sensors, and the method includes the following steps.

[0080] S1. In multi-regional operation scenarios, according to the configuration cycle, obtain regional topology, facility status, risk warning, complex task objectives, historical feedback and operation resource constraint data through IoT interface, GIS database interface, business system interface or resource ledger, construct regional map structure and generate regional node status.

[0081] The structure of the region map is represented as follows:

[0082]

[0083] In the formula, V is the set of regional nodes, E is at least one of road connection, communication connection, spatial adjacency or guarantee dependency relationship, and A is the regional connection strength matrix.

[0084] The status of regional nodes includes regional attributes, facility status, target value, resource accessibility, risk or demand status, resource allocation in the previous cycle, and execution feedback; the data on operational resource constraints includes the total amount of operational resources, resource type capacity, regional configuration upper limit, node accessibility constraints, and minimum guarantee requirements for tasks.

[0085] S2. Based on the complex task objectives and the region graph structure, divide the region into direct task regions and supporting task regions, and generate primary return rate, secondary return rate, and supporting region mask. The task structure perception includes identifying direct task matching relationships based on the complex task objectives and identifying structural support relationships that support the direct task regions based on the region graph structure.

[0086] S2.1 The system synchronously calculates the task matching degree vector d and the structural support degree vector s for all regional nodes in V; the following formula is the expression for the i-th component:

[0087]

[0088] in, This refers to the status of the regional node. For complex mission objectives, Let f represent the support connection strength between the i-th region node and the j-th region node, and let f be the matching function between the task objective and the state of the region nodes. This represents the support connection strength between the i-th region node and the j-th region node. The task matching degree is used to represent the direct matching relationship between the region node and the complex task objective, and the structural support degree is used to represent the structural support effect of the i-th region node on the direct task region. This is the task matching degree vector. This is the structural support vector. Indicates the first Structural support of each regional node and For the set of region node indexes, To support region mask vectors.

[0089] Where d=[ ,...,d_N]ᵀ,s=[ [,...,s_N]ᵀ, where N is the number of nodes in the region; when represented in matrix form, the structural support vector satisfies:

[0090] .

[0091] S2.2 Constructing the task matching function At that time, the system starts from complex task objectives Read the direct target region set Direct target type set And task weight β, and from the region node state The system reads the region identifier, node type, target value, risk or disruption level, and demand pressure. The matching value between the region node and the task target is determined by the following formula:

[0092]

[0093] Will Target Value Risk or level of disruption and demand pressure After normalizing to [0,1], the task matching function f outputs the task matching degree as follows:

[0094]

[0095] in, ≥0 and =1, clip(y,0,1)=min(1,max(0,y)), ∈[0,1]. The direct task matching relationship is jointly determined by the regional target matching field, the target value field, the risk or interruption field, and the demand pressure field.

[0096] S2.3. Determine the direct task region, supporting task region, and supporting region mask using the following formula:

[0097]

[0098] in, and These are the direct task area threshold and the supporting task area threshold, respectively.

[0099] S2.4, for those belonging to For regional nodes, the contribution of the main task is calculated based on task matching degree, target value, risk or interruption level, and demand pressure. ; for those belonging to For regional nodes, the contribution of the sub-task is calculated based on structural support, connection reliability, path reachability, and importance of supporting objects. .

[0100] The contributions from the primary task and secondary tasks are calculated using a normalized weighted average. Let... Indicates connection reliability. Indicates path reachability. To indicate the importance of the supporting object, we have: and These are the non-negative weighting coefficients for the contribution items of the main task and the secondary task, respectively. .

[0101]

[0102]

[0103]

[0104] Where η and μ are preset or trained non-negative weights, and the main task contribution and the secondary task contribution are generated by weighting the above fields.

[0105] From this, we obtain , , , Both M and N are vectors of length N, corresponding to the main task contribution, secondary task contribution, main return rate, secondary return rate, and support region mask of all region nodes, respectively.

[0106] S2.5. Generate the primary return and secondary return using the following formula:

[0107]

[0108] in, and These represent the marginal cost of primary revenue and the marginal cost of secondary revenue, respectively, with ε being the zero-prevention threshold. Marginal costs can be determined by weighting the incremental resource amount, arrival time, and execution difficulty.

[0109]

[0110] in, and This indicates the incremental resource amount required for the corresponding revenue branch. and Indicates arrival time. and Indicates the difficulty of execution. ≥0, To prevent zero threshold.

[0111] S3. With the goal of maximizing the comprehensive return of the current configuration cycle operation and taking into account the cumulative return of multiple configuration cycles, the regional graph structure, regional node status and historical feedback are encoded, and low-dimensional control parameters are output. The low-dimensional control parameters include the main return focus intensity, the main return activation threshold, the secondary return coverage temperature and the main-secondary fusion coefficient.

[0112] S3.1. Add self-loops to the region connectivity strength matrix and normalize it to obtain the normalized connectivity matrix:

[0113]

[0114] Here, normalization can be achieved using row normalization. Let D be a row and diagonal matrix of A+I, then we have:

[0115]

[0116] The system uses a normalized connectivity matrix to perform neighborhood aggregation on the state features of regional nodes.

[0117] S3.2. By using summation pooling, average pooling, or attention pooling, the region graph features after neighborhood aggregation, global resource constraint features, and historical reward feedback features are synthesized into a global encoding that is independent of the number of region nodes. :

[0118]

[0119] in, This is the state matrix of the region nodes. Total amount of operational resources For historical feedback features, `pool` can be one of summation pooling, average pooling, or attention pooling, and ⊕ indicates feature concatenation. `pool` converts features of different numbers of region nodes into a fixed-length global code.

[0120] S3.3, Constructing the comprehensive return U for a single configuration cycle t The cumulative return J after discounting over multiple allocation periods t By leveraging penalty items, the upper limit of allocation, minimum guarantee, and operational execution risk are transformed into quantifiable constraints, forming a quantitative evaluation criterion for the merits of resource allocation schemes; for candidate resource allocation schemes to be evaluated... The overall return of a single configuration cycle job can be expressed as follows: The above return function and penalty term are used together to compare candidate schemes and determine the subsequent configuration strategy.

[0121] in, This represents the constraint deviation or execution cost penalty of the resource allocation scheme relative to the feasible region of job resource allocation, which can be constructed based on the regional allocation upper limit, the minimum guarantee for the task, and the execution cost as follows:

[0122]

[0123] in, This represents the upper limit vector of the region configuration. This represents the minimum guarantee vector for the task. Indicates taking the positive part. This indicates the execution cost resulting from arrival time, execution difficulty, or operational risks. , and The weights are non-negative. The cumulative return across multiple allocation periods can be expressed as:

[0124]

[0125] Where δ is the discount factor and K is the cumulative evaluation period length. Constructed from job resource constraint data, Used to characterize the deviation of candidate solutions from executable constraints or the execution cost. Policy model to improve U t And take into account J t Update the low-dimensional control parameters for the target.

[0126] S3.4. Use the revenue indicators in step S3.3 as the optimization target of the strategy model to guide the model to iteratively update the low-dimensional control parameters, realize the coordinated optimization of configuration parameters for current revenue and long-term cumulative revenue, and provide numerical judgment basis for resource constraint verification and scheme reallocation.

[0127] Strategy Model According to the global encoding zt Output four raw control variables:

[0128] .

[0129] The original control input is mapped to low-dimensional control parameters using the following formula:

[0130]

[0131] Wherein, sigmoid and softplus are monotonic mapping functions, defined as follows:

[0132]

[0133] Where γmin>0 and Tmin>0, therefore γ>0, 0≤τ≤1, T>0, and 0≤λ≤1. Strategy Model The model can be one of a linear model, a tree model, or a neural network model; the model parameters ω are obtained through training using historical configuration records or simulation samples, and the training objective is to maximize the cumulative return.

[0134] ;

[0135] Historical configuration records or simulation samples include the region node state matrix, region connectivity strength matrix, total job resources, historical feedback characteristics, actual configuration schemes, periodic job revenue, and execution feedback, represented as:

[0136]

[0137] in, This includes on-site status, task completion status, changes in facility status, service restoration results, or reasons for incomplete completion. The system uses the data from the sample... and subsequent periodic feedback calculation Used for training or validating policy models .

[0138] When training samples are lacking, The mapping is implemented using a preset rule table or linear rules, and outputs γ, τ, T and λ according to the above mapping formula.

[0139] S4. Perform dynamic primary and secondary return rate fusion decision based on the low-dimensional control parameters, primary return rate, secondary return rate and support area mask to generate a multi-region resource allocation scheme for the current allocation period.

[0140] Main return hotspot weights are generated based on the main return activation threshold τ and the main return focus intensity γ:

[0141] Generate support coverage weights based on the secondary yield coverage temperature T and the support region mask Mᵢ:

[0142]

[0143] The resource configuration vector 'a' for all region nodes is generated by merging according to the following formula, where The resource allocation for the i-th region node:

[0144]

[0145] Where λ is the primary and secondary fusion coefficient, and B t This represents the total number of job resources for the current configuration period. p, c, w, and a are all vectors of length N, where a = [a1, ..., a_N]ᵀ is the resource configuration vector for all region nodes. When B t When = 0, the system directly outputs an all-zero resource allocation vector; when , or When the value is less than ε, the system adopts the maximum main return region distribution, the uniform distribution of the reachable support region, or the uniform distribution of all reachable regions as the backoff distribution, respectively; non-negative pruning is performed on negative values ​​before and after fusion, and total amount verification is performed. , , and They are vectors , , and The There are 1 component, of which , , For weighted components, For the first Resource allocation for each regional node This is the sum of the fusion weights of all regional nodes.

[0146] S5. Perform job resource constraint verification and constraint projection on the multi-region resource configuration scheme, convert the constraint-projected scheme into job instructions and send them to the job terminal, job management platform or equipment control terminal, and update the regional node status for the next configuration cycle based on the execution feedback.

[0147] S5.1 Construct the feasible region for job resource configuration based on job resource constraint data. The feasible domain for configuring the job resources These include non-negativity constraints, single-cycle total amount constraints, resource type capacity constraints, regional configuration upper limit constraints, node reachability constraints, and minimum task guarantee constraints.

[0148] S5.2 The system projects multi-regional resource allocation schemes onto the feasible domain of job resource allocation. The result of multi-region operation resource allocation is obtained; the constraint projection includes zeroing the unreachable region node mask, truncating the region configuration upper limit, normalizing the resource type capacity, supplementing the minimum guarantee of tasks, and redistributing the remaining operation resources according to the region configuration weight.

[0149] S5.3 The multi-region operation resource configuration results after projection are converted into operation instructions containing region node number, resource type, resource quantity or execution priority.

[0150] S5.4 The operation instructions include emergency personnel dispatch instructions, maintenance equipment allocation instructions, material delivery instructions, or service restoration priority instructions. The system updates the regional node status for the next configuration cycle based on execution feedback, changes in periodic operation revenue, or cumulative revenue returned by the operation terminal, operation management platform, or equipment control terminal.

[0151] Example 2

[0152] A task structure-aware multi-region job resource allocation system is used to execute the above method, such as... Figure 2 As shown, it includes a data acquisition module, a region map construction module, a task structure parsing module, a primary and secondary rate of return calculation module, a strategy model module, a primary and secondary fusion allocation module, a feasible region constraint module, a constraint projection output module, and a feedback update module.

[0153] The data acquisition module acquires regional operation status data and operation resource constraint data; the regional map construction module constructs the regional map structure G=(V,E,A); the task structure parsing module determines the direct task region, supporting task region, and supporting region mask; the primary and secondary return rate calculation module generates primary and secondary return rates based on task matching degree, structural support degree, primary task contribution, secondary task contribution, and marginal cost; the strategy model module outputs the original control quantity based on global encoding and maps it to low-dimensional control parameters independent of the number of regional nodes; the primary and secondary fusion allocation module generates a regional resource allocation weight vector based on low-dimensional control parameters, primary return rate, secondary return rate, and supporting region mask; and the feasible region constraint module constructs the feasible region for operation resource allocation. The constraint projection output module is used to generate a feasible region that satisfies the job resource configuration. The system configures multi-regional operation resources and outputs operation instructions; the feedback update module is used to update the regional node status for the next configuration period based on execution feedback, periodic operation benefits, or cumulative benefits.

[0154] This embodiment takes the scenario of resource allocation for emergency repair and service restoration operations in multiple regions after a disaster as an example. The multi-regional operation resource allocation system is deployed on the emergency support management platform, and the set of regions to be configured includes regional nodes such as disaster-stricken roads, communication base stations, energy supply facilities, drainage facilities, temporary support points, and public service facilities.

[0155] The data acquisition module is used to acquire regional operation status data and operation resource constraint data. Within the current configuration period, it acquires water level, equipment temperature, power status, communication link status, and road flooding status through the IoT interface; it acquires regional boundaries, road network, facility locations, and regional adjacency relationships through the GIS database interface; it acquires complex task objectives, pending work orders, service recovery objectives, and historical feedback through the business system interface; and it acquires the quantity and availability status of operation personnel, operation vehicles, operation equipment, support materials, drones, and sensors through the resource ledger database.

[0156] The regional graph construction module is used to build the regional graph structure. It uses regional nodes as graph nodes and road connections, communication connections, spatial adjacencies, and security dependencies as graph edges to construct the regional graph structure and generate a regional node state matrix. The regional node state includes regional attributes, facility status, target value, resource accessibility, risk or demand status, resource allocation in the previous cycle, and execution feedback.

[0157] The task structure parsing module identifies direct task matching relationships based on complex task objectives, used to determine direct task areas, supporting task areas, and supporting area masks. When the complex task objective is to restore critical communication links and ensure the operation of drainage facilities, communication link interruption areas and high-risk drainage facility areas are identified as direct task areas; adjacent areas that serve as road access, equipment transportation, communication relay, or material delivery areas are identified as supporting task areas.

[0158] The primary and secondary return rate calculation module calculates the primary and secondary return rates based on task matching degree, structural support degree, primary and secondary task contribution items and marginal cost, and combines regional nodes and regional connection strength matrix to uniformly generate corresponding multi-class vectors and task region sets.

[0159] In this embodiment, there are N region nodes in the current configuration period, where N is an integer greater than 1, and in practical applications, it can be tens to hundreds. The system uses an N×N region connection strength matrix A and a state matrix X composed of the states of the N region nodes. t Total operational resources B t As input, a resource allocation vector 'a' of length N is generated in one go. In actual computation, no subset of region nodes is extracted. To facilitate illustrating the computation process, the following uses a complete small-scale example with N=5, where the five region nodes are A, B, C, D, and E, and the total operational resources... =20, the direct target regions are A and D, and the region connectivity strength matrix is:

[0160]

[0161]

[0162]

[0163]

[0164]

[0165] The task matching degree is calculated based on the task objective and the status of regional nodes:

[0166]

[0167] according to The structural support was calculated as follows:

[0168]

[0169] when =0.80、 When the value is 0.70, A and D are determined as direct task regions, and B, C, and E are determined as supporting task regions. The supporting region mask is:

[0170]

[0171] The primary and secondary rates of return are calculated based on the contributions of the primary task, the contributions of the secondary task, and the marginal cost. The primary task contribution and the marginal cost of the primary revenue are taken as follows:

[0172]

[0173] The contribution item and marginal cost of the secondary task and its benefits are as follows:

[0174]

[0175] Substituting into the formulas for primary and secondary returns, we get:

[0176] .

[0177] The strategy model module is used to output the original control quantity based on the global encoding and map it to low-dimensional control parameters that are independent of the number of regional nodes. It adds self-loops to the regional connectivity strength matrix and normalizes it, performs neighborhood aggregation, and outputs the primary return focus intensity γ, primary return activation threshold τ, secondary return coverage temperature T, and primary-secondary fusion coefficient λ based on regional graph characteristics, global resource constraint characteristics, and historical feedback characteristics.

[0178] In this embodiment, take =0.5、 =0.1; If the original control quantity output by the strategy model is:

[0179]

[0180] The mapping formula yields γ≈2, τ≈0.80, T≈1, and λ≈0.35. When using a rule-based approach, the rule table maps the state of "significant differences in direct task area returns and moderate support area coverage requirements" to the aforementioned low-dimensional control parameters.

[0181] The primary and secondary fusion allocation module generates a regional resource allocation weight vector based on low-dimensional control parameters, primary return rate, secondary return rate, and support region mask. It generates primary return rate hotspot weights based on the primary return rate, support coverage weights based on the secondary return rate and support region mask, and finally generates regional resource allocation weights using the primary-secondary fusion coefficient λ. When λ is small, resource allocation favors the direct task region; when λ is large, resource allocation favors support region coverage.

[0182] When τ=0.80, γ=2, T=1, and λ=0.35, we can obtain:

[0183]

[0184]

[0185]

[0186] Therefore, the resource allocation for multi-regional operations is as follows:

[0187]

[0188] When the job resources are discrete resources such as personnel, vehicles, or equipment, the system uses continuous allocation quantities as allocation weights and generates integer job instructions through rounding and residual allocation. First, the allocation quantities for each region are rounded down, and then the remaining resources are allocated to the region node with the largest decimal part.

[0189]

[0190] In this example, floor(a) = [7, 2, 1, 5, 2], the remaining resource r = 3, the three region nodes with the largest fractional part are C, A, and B, and the integer resource allocation result is:

[0191]

[0192] Therefore, the system prioritizes the direct task areas A and D, while reserving the resources needed for road accessibility, communication connectivity, or service recovery chains to support areas B, C, and E. This example includes a connectivity matrix, task matching degree, structural support degree, primary and secondary task contributions, marginal cost, primary and secondary return rates, fusion weights, continuous resource allocation amounts, and integer resource allocation results.

[0193] The feasible region constraint module configures the feasible region based on the resources used to build the job. The current configuration cycle's total job resources, resource type capacity, regional configuration upper limit, node reachability constraints, and minimum task guarantee requirements construct a feasible region for job resource configuration. The constraint projection output module projects the multi-region resource configuration scheme onto this feasible region to obtain the multi-region job resource configuration result that satisfies the constraints. The projection process can be represented as:

[0194]

[0195] in, This indicates that the resource allocation scheme will be projected onto the feasible domain of job resource allocation. The projection operator, This represents the result of multi-region operation resource allocation after constraint projection.

[0196] The constraint projection output module is used to generate a feasible region that satisfies the job resource configuration. The system configures multi-regional operation resources and outputs operation instructions.

[0197] The feedback update module is used to update the status of regional nodes for the next configuration period based on execution feedback, periodic job revenue, or cumulative revenue.

[0198] For various heterogeneous operational resources such as emergency repair vehicles, pumping equipment, and communication support equipment, the system performs the aforementioned weight calculation, feasible domain construction, and constraint projection processes according to the independent capacity of each resource type. After independently completing the configuration calculation for each resource type, the configuration quantity of nodes in unreachable areas is set to zero, resources are supplemented for nodes in critical areas that do not meet the minimum guarantee requirements, and the allocation of nodes that exceed the regional configuration limit is truncated. The surplus resources generated by the truncation are then redistributed to the remaining reachable nodes. Finally, the system outputs the number of vehicles, equipment, support materials, and operational priority for each regional node.

[0199] Example 3

[0200] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of a task structure-aware multi-region job resource allocation method as described in Embodiment 1.

[0201] Furthermore, the task structure-aware multi-region job resource allocation method described in Embodiment 1 can be implemented as a computer software program. For example, this embodiment includes a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the method. In such an embodiment, the computer program can be downloaded and installed from a network, and / or installed from a removable medium. When the computer program is executed by a processor, it performs the functions defined in the method of this application.

[0202] Example 4

[0203] A computer-readable storage medium storing a computer program that, when executed by a processor, implements any step of the multi-region job resource allocation method for task structure awareness as described in Embodiment 1.

[0204] The computer-readable medium described in this application may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0205] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0206] In summary, this invention effectively overcomes the shortcomings of the prior art and has high industrial applicability. The above embodiments are intended to illustrate the substantive content of this invention, but are not intended to limit the scope of protection of this invention. Those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of this invention without departing from the essence and scope of protection of this invention.

[0207] The above embodiments are specific implementations of the present invention, but the implementation of the present invention is not limited to the above embodiments. Any other combination, change, modification, substitution, or simplification that does not exceed the design concept of the present invention shall fall within the protection scope of the present invention.

Claims

1. A method for allocating multi-regional job resources with task structure awareness, characterized in that, The application to a multi-regional operational resource allocation system includes the following steps: S1. Call various interfaces and databases according to the configuration cycle to collect topology, facilities, risks, tasks, feedback and resource constraints data of multi-region operations, build regional graph structure and generate regional node status; S2. Divide the direct task region and the supporting task region according to the complex task objective and the region map structure, and generate the main return rate, the secondary return rate and the supporting region mask. S3. With the goal of maximizing the overall benefit of the current configuration cycle operation and taking into account the cumulative benefit of multiple configuration cycles, encode the regional graph structure, regional node status and historical feedback, and output low-dimensional control parameters. S5. Perform dynamic primary and secondary return rate fusion decision based on the low-dimensional control parameters, primary return rate, secondary return rate and support area mask to generate a multi-region resource allocation scheme for the current allocation period. S6. Perform job resource constraint verification and constraint projection on the multi-region resource configuration scheme, convert the constraint-projected scheme into job instructions and send them to the job terminal, job management platform or equipment control terminal, and update the regional node status for the next configuration cycle based on the execution feedback.

2. The method as described in claim 1, characterized in that, In step S1, the various interfaces and databases include IoT interfaces, GIS database interfaces, business system interfaces, or resource ledger databases; The multi-source data includes regional topology, facility status, risk warning, complex task objectives, historical feedback, and operational resource constraint data; The job resource constraint data mentioned above includes the total amount of job resources, resource type capacity, regional configuration limit, node reachability constraints, and minimum task guarantee requirements; The expression for the region map structure is: in For a set of region nodes, It is at least one of the following: road connection, communication connection, spatial adjacency, or security dependency. This is the region connectivity strength matrix; The status of the regional nodes includes regional attributes, facility status, target value, resource accessibility, risk or demand status, resource allocation in the previous cycle, and execution feedback.

3. The method as described in claim 1, characterized in that, Step S2 includes the following steps: S2.1, From the perspective of complex task objectives Get the set of direct target regions Direct target type set and task weight And from the state matrix of all region nodes Obtain the region identifier of each region node Node type Target Value Risk or level of disruption and demand pressure ; S2.2 Construct the task matching function as follows: And simultaneously calculate the task matching degree vector for all regional nodes. The Each component : in, , , and Normalize to [0,1] and , , ; S2.3, Based on the task matching degree vector and support connection strength matrix Calculate the structural support vector and determine the direct task area. Supported task areas and support region mask vector : in, This is the threshold for the direct task region. To support task region thresholds, This represents the support connection strength between the i-th region node and the j-th region node; This is the task matching degree vector. This is the structural support vector. Indicates the first Structural support of each regional node and For the set of region node indexes, To support region mask vectors. S2.4 Calculate the main task contribution item according to the following formula. Sub-task contribution items : in, Indicates connection reliability. Indicates path reachability. Indicates the importance of the supporting object. and These are the non-negative weighting coefficients for the contribution items of the main task and the secondary task, respectively. ,and , , , ; S2.5, Calculate the main return rate using the following formula. and yield : in, Main revenue marginal cost, For the marginal cost of secondary revenue, and This indicates the amount of incremental resources required for the corresponding revenue branch. and Indicates arrival time. and Indicates the difficulty of execution. This is the cost-weighted coefficient. ,and , To prevent zero threshold.

4. The method as described in claim 1, characterized in that, Step S3 includes the following steps: S3.1 Add self-loops to the region connectivity strength matrix and perform row normalization to obtain the normalized connectivity matrix. Then, use the normalized connectivity matrix to perform neighborhood aggregation on the region node state matrix. S3.

2. By using summation pooling, average pooling, or attention pooling, the features of the aggregated neighborhood region graph and the total amount of job resources are analyzed. and historical feedback characteristics Synthesized into a global code independent of the number of regional nodes. : Where pool is one of summation pooling, average pooling or attention pooling, and ⊕ represents feature concatenation; This indicates the global code for the current configuration period. For the normalized connectivity matrix, This is the state matrix of the region nodes. Total amount of operational resources This is a historical feedback characteristic. S3.3, Based on candidate resource allocation schemes Total job revenue generated during the current configuration period Cumulative returns over multiple allocation periods To achieve this goal, we construct quantitative evaluation criteria for the merits of resource allocation schemes, including: in, These are the candidate resource allocation schemes to be evaluated. This refers to the feasible domain for job resource configuration constructed based on job resource constraint data. To constrain deviations or impose penalties; and They represent the first The main task contribution items and secondary task contribution items of each regional node. For the current single-cycle operation's overall revenue, From Start and back The discounted cumulative return over each configuration period, The discount factor is... This is the length of the cumulative revenue evaluation window. S3.4, Strategy Model According to global encoding Output raw control quantity And mapped to low-dimensional control parameters as follows: in, , , Focus on strength of main yield. The activation threshold for primary yield. To cover the temperature for secondary yield, Primary and secondary fusion coefficients; Maximize through historical configuration records or simulation samples It is obtained through training for the target, or determined by a pre-set rule table or linear rules. , , and The original control variables output by the strategy model correspond to the primary return focus intensity, primary return activation threshold, secondary return coverage temperature, and primary-secondary fusion coefficient, respectively. and This is a preset positive lower limit.

5. The method as described in claim 1, characterized in that, In step S4, based on the main return activation threshold and the main return focus intensity Generate main return hot spot weights : ； Based on the secondary yield coverage temperature and the supported region mask Generate supporting coverage weights : ； The resource allocation vectors of all region nodes are generated by merging them according to the following formula. : in, For the first One portion, Primary and secondary fusion coefficients Let p, c, w, and a be the total number of job resources in the current configuration period, where p, c, w, and a are all vectors of length N. Resource allocation vector for all region nodes; , , and They are vectors , , and The There are 1 component, of which , , For weighted components, For the first Resource allocation for each regional node This is the sum of the fusion weights of all regional nodes. when Output an all-zero resource allocation vector when; , or Less than When the maximum main return region distribution, the uniform distribution of the reachable support region, or the uniform distribution of all reachable regions are used as the fallback distribution, non-negative pruning and total amount verification are performed on the fusion weights.

6. The method as described in claim 1, characterized in that, Step S5 includes the following steps: S5.1 Construct the feasible region for job resource configuration based on job resource constraint data. The feasible domain for configuring the job resources These include non-negativity constraints, single-cycle total amount constraints, resource type capacity constraints, regional configuration upper limit constraints, node reachability constraints, and minimum task guarantee constraints. S5.2, The multi-regional resource allocation scheme is divided into... Project to the feasible domain of job resource configuration This yields the results of multi-regional operational resource allocation; among them, The constraint projection operator includes zeroing the unreachable region node mask, truncating the region configuration upper limit, normalizing the resource type capacity, supplementing the minimum guarantee for tasks, and redistributing the remaining job resources according to the region configuration weight. S5.3 Convert the multi-region operation resource configuration results into operation instructions containing region node number, resource type, resource quantity or execution priority; S5.

4. Update the regional node status for the next configuration cycle based on the execution feedback, periodic operation revenue, or cumulative revenue changes returned by the operation terminal, operation management platform, or equipment control terminal.

7. A task structure-aware multi-regional operation resource allocation system, characterized in that, For performing the method according to any one of claims 1-6, comprising: The data acquisition module is used to acquire regional operation status data and operation resource constraint data; The region map construction module is used to build the region map structure; The task structure parsing module is used to determine the direct task region, the supporting task region, and the supporting region mask; The primary and secondary rate of return calculation module is used to generate primary and secondary rates of return based on task matching degree, structural support degree, primary task contribution, secondary task contribution, and marginal cost. The strategy model module is used to output the original control quantity based on the global encoding and map it to low-dimensional control parameters that are independent of the number of regional nodes. The primary and secondary fusion allocation module is used to generate a regional resource allocation weight vector based on low-dimensional control parameters, primary rate of return, secondary rate of return, and support region mask. The feasible region constraint module is used to construct the feasible region for job resource configuration; The constraint projection output module is used to generate multi-region operation resource configuration results that satisfy the feasible domain of operation resource configuration and output operation instructions; The feedback update module is used to update the regional node status for the next configuration period based on execution feedback, periodic job revenue, or cumulative revenue.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the method of any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the method described in any one of claims 1-6.

Images