An oil field emergency resource path deployment joint scheduling method

Abstract

The application discloses an oil field emergency resource path deployment joint scheduling method, and belongs to the technical field of oil field emergency resource scheduling and path planning. In order to solve the problem that the existing scheduling does not consider dynamic risks, road traffic, resource capacity and vehicle deployment conditions, a vehicle-mounted digital twin graph is constructed, a dynamic risk field is generated and predicted by fusing multi-source observation, the road edge traffic time and risk cost are obtained by using a graph neural network, a task-resource capacity matrix is established, the deployment hard constraint of a candidate parking point is verified, resource allocation, driving path and deployment posture are jointly solved, and according to the change of the dynamic risk field, rolling re-planning is carried out, the consistency of the scheduling result and the field risk and the deployment condition can be improved, and the method is suitable for the fields of oil field well site, oil field tank area and oil field pipeline area emergency resource scheduling and mobile emergency command.

Description

Technical Field

[0001] This invention belongs to the field of oilfield emergency resource scheduling and route planning technology, specifically involving a joint scheduling method for oilfield emergency resource route deployment. Background Technology

[0002] Oilfield well sites typically include roads, production facilities, restricted areas, and areas suitable for vehicle parking and equipment deployment. In emergency situations such as leaks or fires, it is necessary to dispatch emergency vehicles, drones, and firefighting resources to the site and complete vehicle parking, outrigger deployment, mast raising, radar detection, and communication connections. Traditional dispatching methods usually generate dispatch plans based on fixed emergency plans, current resource locations, or road distances, making it difficult to simultaneously reflect changes in on-site risks and the actual deployment conditions after vehicle arrival.

[0003] With the development of digital twin, multi-source sensing, and path planning technologies, existing emergency dispatch solutions can construct virtual scenarios using maps, facilities, and resource information, integrate observational data such as gas, infrared, meteorological, and video data for risk assessment, and determine resource allocation relationships and driving routes through predictive models or optimization algorithms. Some solutions can also predict the diffusion trends of pollutants or hazardous substances and dynamically adjust rescue routes based on road conditions; some emergency command vehicles are also equipped with outriggers, lifting masts, radar, and communication equipment to meet on-site command and detection needs.

[0004] However, existing solutions typically handle risk prediction, resource matching, route planning, and vehicle deployment separately. Risk prediction results fail to simultaneously influence both road toll costs and the feasibility of deploying candidate parking spots. Routes generated solely based on distance or road travel time may guide vehicles to locations where slope, parking orientation, outrigger deployment space, mast safety distance, radar coverage, or communication line-of-sight do not meet requirements, necessitating relocation or attitude adjustments upon arrival. Furthermore, conflicts may arise between different emergency resources regarding routes, tasks, or deployment locations, making it difficult to directly translate the generated scheduling results into executable instructions.

[0005] Therefore, it is necessary to establish the correlation between dynamic risk field, roadside cost, resource capacity and vehicle deployment conditions, determine resource allocation, driving path and deployment posture in the same solution process, and be able to update the scheduling results as the on-site risk changes. Summary of the Invention

[0006] To address the problem in existing oilfield emergency dispatch technologies that fail to consider dynamic risks, road accessibility, resource capacity, and vehicle deployment conditions, resulting in a mismatch between dispatch results and on-site deployment conditions, this invention proposes the following solution: A joint scheduling method for emergency resource deployment in oilfields includes: S1. Obtain the roads, facilities, restricted areas, and deployable areas of the oilfield well site, as well as the status of emergency vehicles, drones, and fire-fighting resources; determine candidate parking points for emergency vehicles within the deployable areas; and construct an onboard digital twin map. S2 integrates radar, gas, infrared, meteorological, and video observations to generate a current dynamic risk field that includes wind direction, gas concentration, thermal anomalies, and radar confidence, and predicts the risk diffusion in future periods. S3. Input the road node features, roadside features and the dynamic risk field into the graph neural network to obtain the passage time and risk cost of each roadside in the future time period. S4. Establish a task-resource capability matrix based on emergency task requirements, resource capabilities, resource locations, and arrival time limits, and determine candidate matching relationships between resources and emergency tasks. S5. For the candidate parking spots, verify the deployment feasibility based on the vehicle parking orientation, support leg deployment area, mast safety distance, radar coverage and communication line of sight, and eliminate candidate parking spots that do not meet the hard constraints of deployment. S6. Based on the travel time and risk cost of each roadside, the candidate matching relationship, and the candidate parking points that have not been eliminated, jointly solve the resource allocation, driving path, and deployment posture, and issue the generated scheduling instructions and deployment instructions to the corresponding resources, and perform rolling replanning when the dynamic risk field changes and the replanning conditions are met.

[0007] Furthermore, in the vehicle-mounted digital twin graph described in S1, the graph nodes include road intersections, facilities, restricted area boundaries, deployable points, and resources. The deployable points include candidate parking spots for emergency vehicles. The graph edges represent road traffic relationships or spatial adjacency relationships. Each resource status record includes a timestamp, source node, version number, validity flag, and verification field.

[0008] Furthermore, the prediction of risk diffusion in future time periods as described in S2 includes: obtaining a predicted risk value based on the temporal changes in wind direction, gas concentration, thermal anomaly, and radar confidence, and mapping the predicted risk value to the roadside and candidate parking spots.

[0009] Furthermore, the graph neural network described in S3 concatenates the road node features at both ends of the roadside with the corresponding roadside features to output the passage time and risk cost of the roadside.

[0010] Furthermore, the task-resource capability matrix described in S4 records the capabilities required for each emergency task, the capabilities possessed by each resource, the current location of each resource, and the arrival time limit, in order to determine the candidate matching relationship between resources and emergency tasks.

[0011] Furthermore, the deployment feasibility verification described in S5 also includes verifying the slope of the candidate parking point, and forming Boolean constraints based on the slope of the candidate parking point, the vehicle parking orientation, the support leg deployment area, the mast safety distance, radar coverage, and communication line of sight, and using each Boolean constraint as a hard constraint for joint solution in S6.

[0012] Furthermore, the joint costs used in the joint solution described in S6 include travel time, parking time, deployment time, risk exposure cost, communication penalty, resource conflict penalty, and radar blind zone penalty.

[0013] Furthermore, the rolling replanning described in S6 also includes: when there is a sensor state conflict, inconsistent model output, communication interruption, or the critical state does not meet the preset safety requirements, stopping the use of the current solution result, performing at least one of re-acquisition, recalculation, rollback of the previous stable result, or manual verification, and saving the triggering reason, input version, and output verification information.

[0014] Furthermore, the joint solution described in S6 includes: under the conditions of satisfying resource capacity, arrival time limit and deployment hard constraints, calculating the joint cost for the combination of resource allocation, travel path and deployment attitude, and taking the combination with the minimum joint cost as the current solution result.

[0015] Based on the same inventive concept, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described method.

[0016] Compared with the prior art, the present invention has the following beneficial effects: By constructing an onboard digital twin map that includes roads, facilities, restricted areas, deployable areas, candidate parking spots, and resource status, and mapping the current dynamic risk field and predicted risk values ​​formed by multi-source observations to the roadside and candidate parking spots, the risk diffusion results can be directly used to calculate road traffic costs and determine parking locations, thereby improving the consistency between scheduling information and on-site risk status.

[0017] By establishing a task-resource capability matrix and using resource capability, resource location, and arrival time limit together to determine candidate matching relationships, resources that cannot execute corresponding emergency tasks or cannot arrive on time can be excluded, so that subsequent path and deployment attitude solutions are based on resources that have the conditions to execute.

[0018] By verifying the slope of the candidate parking point, the vehicle parking orientation, the outrigger deployment area, the mast safety distance, radar coverage, and the communication line of sight, and using the verification results as hard constraints for the joint solution, the number of situations where the vehicle cannot complete support, mast deployment, radar detection, or communication connection after arriving at the candidate parking point can be reduced.

[0019] By incorporating travel time, parking time, deployment time, risk exposure cost, communication penalty, resource conflict penalty, and radar blind spot penalty into the joint cost, resource allocation, travel path, and deployment attitude are jointly determined under the conditions of satisfying resource capacity, arrival time limit, and deployment hard constraints, so that the generated scheduling results can directly form scheduling instructions and deployment instructions.

[0020] By setting up rolling replanning and anomaly rollback handling, when the dynamic risk field changes or the critical state does not meet the safety requirements, the system can recalculate or roll back to the previous stable result, and save the triggering cause, input version and output verification information, thus maintaining the continuity of the scheduling process and the traceability of the results.

[0021] This invention integrates dynamic risk prediction, road cost calculation, resource capacity matching, and vehicle deployment constraints. It can jointly generate emergency resource allocation results, driving paths, and deployment postures, and continuously update the scheduling results according to changes in the dynamic risk field. It is applicable to emergency resource scheduling and mobile emergency command in oilfield well sites, oilfield tank farms, and oilfield pipeline areas. Attached Figure Description

[0022] Figure 1 This is a flowchart of the joint scheduling method for the deployment of emergency resources in oilfields, as described in the implementation method. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will now be clearly and completely described in conjunction with the accompanying drawings.

[0024] Implementation Method 1 like Figure 1 As shown in this embodiment, a joint scheduling method for the deployment of emergency resources in oilfields includes: S1. Obtain the roads, facilities, restricted areas, and deployable areas of the oilfield well site, as well as the status of emergency vehicles, drones, and fire-fighting resources; determine candidate parking points for emergency vehicles within the deployable areas; and construct an onboard digital twin map. S2 integrates radar, gas, infrared, meteorological, and video observations to generate a current dynamic risk field that includes wind direction, gas concentration, thermal anomalies, and radar confidence, and predicts the risk diffusion in future periods. S3. Input the road node features, roadside features and the dynamic risk field into the graph neural network to obtain the passage time and risk cost of each roadside in the future time period. S4. Establish a task-resource capability matrix based on emergency task requirements, resource capabilities, resource locations, and arrival time limits, and determine candidate matching relationships between resources and emergency tasks. S5. For the candidate parking spots, verify the deployment feasibility based on the vehicle parking orientation, support leg deployment area, mast safety distance, radar coverage and communication line of sight, and eliminate candidate parking spots that do not meet the hard constraints of deployment. S6. Based on the travel time and risk cost of each roadside, the candidate matching relationship, and the candidate parking points that have not been eliminated, jointly solve the resource allocation, driving path, and deployment posture, and issue the generated scheduling instructions and deployment instructions to the corresponding resources, and perform rolling replanning when the dynamic risk field changes and the replanning conditions are met.

[0025] The above process forms a continuous processing chain of risk prediction, road cost, resource matching, and deployment verification, so that the joint solution result simultaneously includes resource, path, and deployment information.

[0026] Preferably, this implementation sequentially processes the input data through digital twin graph construction, dynamic risk field prediction, road cost prediction, resource capacity matching, and deployment constraint joint optimization, and outputs scheduling instructions and deployment instructions; when replanning or rollback occurs, the triggering reason, input version, and output verification information are saved.

[0027] Furthermore, in the vehicle-mounted digital twin graph described in S1, the graph nodes include road intersections, facilities, restricted area boundaries, deployable points, and resources. The deployable points include candidate parking spots for emergency vehicles. The graph edges represent road traffic relationships or spatial adjacency relationships. Each resource status record includes a timestamp, source node, version number, validity flag, and verification field.

[0028] The graph structure described above uniformly represents road traffic relationships, spatial adjacency relationships, and resource status, enabling subsequent risk mapping and path calculation to use interconnected data.

[0029] Preferably, the input data undergoes range verification and outlier filtering before entering the corresponding processing procedure; the records related to the model output also include model hash, threshold configuration, and hardware execution node.

[0030] Furthermore, the prediction of risk diffusion in future time periods as described in S2 includes: obtaining a predicted risk value based on the temporal changes in wind direction, gas concentration, thermal anomaly, and radar confidence, and mapping the predicted risk value to the roadside and candidate parking spots.

[0031] The aforementioned risk mapping enables the same predicted risk value to participate in both roadside cost calculation and candidate parking point verification, maintaining consistency between the risk status on which path planning and deployment location judgment are based.

[0032] Furthermore, the graph neural network described in S3 concatenates the road node features at both ends of the roadside with the corresponding roadside features to output the passage time and risk cost of the roadside.

[0033] The above feature splicing method makes the prediction results of the roadside affected by both the state of the adjacent road nodes and the state of the roadside itself.

[0034] Furthermore, the task-resource capability matrix described in S4 records the capabilities required for each emergency task, the capabilities possessed by each resource, the current location of each resource, and the arrival time limit, in order to determine the candidate matching relationship between resources and emergency tasks.

[0035] The matrix above links task conditions with resource status, enabling the joint solution to select execution resources from those that meet the task capabilities and reach requirements.

[0036] Furthermore, the deployment feasibility verification described in S5 also includes verifying the slope of the candidate parking point, and forming Boolean constraints based on the slope of the candidate parking point, the vehicle parking orientation, the support leg deployment area, the mast safety distance, radar coverage, and communication line of sight, and using each Boolean constraint as a hard constraint for joint solution in S6.

[0037] The Boolean constraints described above can eliminate undeployable locations before path solving, preventing undeployable candidate parking points from entering the current solution results.

[0038] Furthermore, the joint costs used in the joint solution described in S6 include travel time, parking time, deployment time, risk exposure cost, communication penalty, resource conflict penalty, and radar blind zone penalty.

[0039] The aforementioned joint costs evaluate candidate combinations from multiple aspects, including driving, parking, deployment, risk, communication, resource conflicts, and radar coverage, providing a unified basis for comparison among different candidate combinations.

[0040] Furthermore, the rolling replanning described in S6 also includes: when there is a sensor state conflict, inconsistent model output, communication interruption, or the critical state does not meet the preset safety requirements, stopping the use of the current solution result, performing at least one of re-acquisition, recalculation, rollback of the previous stable result, or manual verification, and saving the triggering reason, input version, and output verification information.

[0041] The above processing can prevent abnormal data or unverified calculation results from directly generating execution instructions, and retain the state records required for replanning or rollback processes.

[0042] Furthermore, the joint solution described in S6 includes: under the conditions of satisfying resource capacity, arrival time limit and deployment hard constraints, calculating the joint cost for the combination of resource allocation, travel path and deployment attitude, and taking the combination with the minimum joint cost as the current solution result.

[0043] The above solution method first ensures that the candidate combination meets the execution conditions, and then determines the current solution result based on the joint cost, so that the output result has resource, time limit and deployment feasibility at the same time.

[0044] The method described in this embodiment can be executed by a processor calling a computer program, which can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the above-mentioned joint scheduling method for deploying emergency resources in oilfields can be implemented.

[0045] Implementation Method 2 This implementation method uses the joint dispatching of emergency vehicles, drones, and firefighting resources in an oilfield well site, along with its numerical comparison process, as a specific example. For instance... Figure 1 As shown, the joint scheduling is executed in the following order: construction of vehicle-mounted digital twin graph, generation and prediction of dynamic risk field, prediction of roadside cost, matching of task and resource capabilities, verification of candidate parking points and joint solution.

[0046] First, the roads, facilities, restricted areas, and deployable areas of the oilfield well site are acquired, along with the status of emergency vehicles, drones, and firefighting resources. Within the deployable areas, candidate parking spots for emergency vehicles are determined. Road intersections, facilities, restricted area boundaries, deployable points, and resources are set as graph nodes, and road traffic relationships or spatial adjacency relationships are set as graph edges, constructing an onboard digital twin graph. Each resource status record includes a timestamp, source node, version number, validity flag, and verification field. Before input data enters the corresponding processing procedure, the data range and abnormal states are verified; records related to model output also include model hashes, threshold configurations, and hardware execution nodes.

[0047] This implementation inputs radar, gas, infrared, meteorological, and video observations into the dynamic risk field prediction process. It correlates these observations according to timestamps to generate a current dynamic risk field including wind direction, gas concentration, thermal anomalies, and radar confidence levels. Based on the temporal changes in wind direction, gas concentration, thermal anomalies, and radar confidence levels, it obtains predicted risk values ​​for future periods and maps these predicted risk values ​​to roadside and candidate parking spots in the vehicle-mounted digital twin map.

[0048] For each roadside, the road node features at both ends and the corresponding roadside features are read. The road node features at both ends of the roadside are concatenated with the corresponding roadside features, and the concatenation result, along with the dynamic risk field, is input into a graph neural network to obtain the passage time and risk cost of each roadside in future time periods. The training data of the graph neural network is isolated according to different well sites, and the training configuration, random seed, and model hash are recorded. After training, the graph neural network is exported and deployed on GPU or NPU execution nodes.

[0049] A task-resource capability matrix is ​​established based on emergency mission requirements, resource capabilities, resource locations, and arrival time limits. This matrix records the capabilities required for each emergency mission and the capabilities possessed by each resource. Combined with the current location and arrival time limit of each resource, candidate matching relationships between resources and emergency missions are determined. Resources that do not meet the required capabilities or arrival time limits are not included in the candidate matching relationships for their corresponding emergency missions.

[0050] For emergency vehicles in the candidate matching relationships, the slope of the candidate parking point, the vehicle's parking orientation, the outrigger deployment area, the mast safety distance, radar coverage, and the communication line of sight are verified. Boolean constraints are formed based on the slope of the candidate parking point, the vehicle's parking orientation, the outrigger deployment area, the mast safety distance, radar coverage, and the communication line of sight. If any Boolean constraint is not satisfied, the corresponding candidate parking point is removed from the current joint solution scope.

[0051] For each candidate combination consisting of resource allocation, travel path, and deployment attitude, the joint cost is calculated under the conditions of satisfying resource capacity, arrival time limit, and deployment hard constraints. The joint cost is:

[0052] in, Indicates the cost of joint action. Indicates travel time. Indicates parking time. Indicates the deployment time. Indicates the cost of risk exposure. Indicates a communication penalty. Indicates resource conflict penalties. Indicates radar blind spot penalty, The weights representing the costs of risk exposure The weight of the communication penalty is indicated. This indicates the weight of the resource conflict penalty. This indicates the weight of the radar blind zone penalty.

[0053] Candidate combinations that meet the resource capacity, arrival time limit, and deployment hard constraints are compared according to their joint cost, and the candidate combination with the minimum joint cost is taken as the current solution result. Scheduling instructions and deployment instructions are generated based on the current solution result and then issued to the corresponding emergency vehicles, drones, or fire-fighting resources.

[0054] During command execution, radar, gas, infrared, meteorological, and video observations are continuously acquired and the current dynamic risk field is updated. When changes in the dynamic risk field satisfy the replanning conditions, the travel time and risk cost at the roadside are recalculated, and resource allocation, travel paths, and deployment attitudes are resolved based on the updated candidate matching relationships and candidate parking points.

[0055] When sensor state conflicts, inconsistent model outputs, communication interruptions, or critical states failing to meet preset safety requirements occur, the current solution result is stopped, and at least one of the following is executed: re-acquisition, recalculation, rollback from the previous stable result, or manual verification. The triggering reason, input version, and output verification information are saved during the process.

[0056] To verify the impact of different scheduling techniques on scheduling results, five ablation groups, A0 to A4, were set up. Among them, A0 uses static shortest path, A1 adds real-time road status to A0, A2 adds dynamic risk field to A1, A3 adds deployment constraints and communication conditions to A2, and A4 adopts a complete joint optimization scheme that includes resource matching, dynamic risk, road cost, deployment hard constraints, and joint cost.

[0057] Numerical experiments utilize ARM / NPU edge nodes and GPU or NPU acceleration nodes, configured with vehicle-mounted switches, uninterruptible power supplies, temperature recording devices, and network status control equipment. The runtime environment includes Python, PyTorch, ONNX Runtime, and the runtime environment corresponding to the execution nodes. Input data includes oilfield well site radar point clouds, infrared images, visible light video, meteorological data, location data, equipment status data, and emergency task sequences. Training, validation, and test data are isolated according to well site or date.

[0058] Each ablation group used the same input, confidence threshold, power consumption mode, number of warm-ups, and environmental conditions. Centered on the design mean of groups A0 to A4, five independent replicates were generated for each group. Each ablation group used the same input, confidence threshold, power consumption mode, number of warm-ups, and environmental conditions, and five independent numerical calculations were performed using a fixed random seed. The values ​​in the table below are expressed as the mean ± standard deviation of the five numerical calculations.

[0059] The numerical results for each ablation group are as follows:

[0060] In the above numerical results, as real-time road conditions, dynamic risk fields, deployment constraints, and communication conditions are added sequentially, the average resource arrival time, the number of resource conflicts, and the risk exposure index decrease sequentially. The average resource arrival time for A4 relative to A0 decreases from 18.600 min to 14.100 min, corresponding to a change of 24.2%; the number of resource conflicts decreases from 14.000 to 5.000; and the risk exposure index decreases from 100.000 to 61.800. The average resource arrival time for A0 and A4 is calculated using... Test results , .in, Indicates the significance probability. This represents the standardized effect size.

[0061] By further altering the update time of the dynamic risk field, and recording the average resource arrival time for both the conventional approach and the method described in this embodiment, the following values ​​were obtained:

[0062] When the dynamic risk field update time changes from 1s to 20s, the average resource arrival time of the method described in this embodiment changes from 13.90min to 16.70min; under the same update time, the values ​​of each group are lower than the average resource arrival time of the corresponding conventional scheme.

[0063] To record changes in risk exposure during continuous operation, the risk exposure indices of the conventional approach and the method described in this embodiment were recorded during continuous operation, yielding the following values:

[0064] During the continuous numerical process up to 30 minutes, the risk exposure index of the conventional scheme changed from 89.95 to 101.80, while the risk exposure index of the method described in this embodiment changed from 56.90 to 62.75.

[0065] Further, standard operating conditions, high-temperature operating conditions, weak network operating conditions, and nighttime or obstructed operating conditions were set up, and the average resource arrival time under different operating conditions was recorded, yielding the following values:

[0066] In the numerical results above, the method described in this embodiment has a lower average resource arrival time than the corresponding conventional scheme under standard operating conditions, high temperature operating conditions, weak network operating conditions, and nighttime or shading operating conditions. The complete joint optimization scheme can uniformly adjust resource allocation, travel path, and deployment posture based on dynamic risk field, roadside cost, resource candidate matching relationship, and deployment hard constraints of candidate parking points.

[0067] The embodiments of the present invention have been described above. These embodiments are used to help understand the technical solutions of the present invention and should not be construed as limiting the scope of the technical solutions of the present invention. Equivalent substitutions or conventional modifications made by those skilled in the art to the embodiments without departing from the technical concept of the present invention should all fall within the scope of the technical solutions of the present invention.

Claims

1. A joint scheduling method for emergency resource deployment in oilfields, characterized in that, include: S1. Obtain the roads, facilities, restricted areas, and deployable areas of the oilfield well site, as well as the status of emergency vehicles, drones, and fire-fighting resources; determine candidate parking points for emergency vehicles within the deployable areas; and construct an onboard digital twin map. S2 integrates radar, gas, infrared, meteorological, and video observations to generate a current dynamic risk field that includes wind direction, gas concentration, thermal anomalies, and radar confidence, and predicts the risk diffusion in future periods. S3. Input the road node features, roadside features and the dynamic risk field into the graph neural network to obtain the passage time and risk cost of each roadside in the future time period. S4. Establish a task-resource capability matrix based on emergency task requirements, resource capabilities, resource locations, and arrival time limits, and determine candidate matching relationships between resources and emergency tasks. S5. For the candidate parking spots, verify the deployment feasibility based on the vehicle parking orientation, support leg deployment area, mast safety distance, radar coverage and communication line of sight, and eliminate candidate parking spots that do not meet the hard constraints of deployment. S6. Based on the travel time and risk cost of each roadside, the candidate matching relationship, and the candidate parking points that have not been eliminated, jointly solve the resource allocation, driving path, and deployment posture, and issue the generated scheduling instructions and deployment instructions to the corresponding resources, and perform rolling replanning when the dynamic risk field changes and the replanning conditions are met.

2. The method according to claim 1, characterized in that, In the vehicle-mounted digital twin graph described in S1, the graph nodes include road intersections, facilities, restricted area boundaries, deployable points, and resources. The deployable points include candidate parking points for emergency vehicles, and the graph edges represent road traffic relationships or spatial adjacency relationships. Each resource status record includes a timestamp, source node, version number, validity flag, and verification field.

3. The method according to claim 1, characterized in that, S2's prediction of risk diffusion in future time periods includes: obtaining predicted risk values ​​based on the temporal changes in wind direction, gas concentration, thermal anomalies, and radar confidence, and mapping the predicted risk values ​​to roadside and candidate parking spots.

4. The method according to claim 1, characterized in that, The graph neural network described in S3 concatenates the road node features at both ends of the roadside with the corresponding roadside features, and outputs the passage time and risk cost of the roadside.

5. The method according to claim 1, characterized in that, The task-resource capability matrix described in S4 records the capabilities required for each emergency task, the capabilities possessed by each resource, the current location of each resource, and the arrival time limit, in order to determine the candidate matching relationship between resources and emergency tasks.

6. The method according to claim 1, characterized in that, The deployment feasibility verification described in S5 also includes verifying the slope of the candidate parking point, and forming Boolean constraints based on the slope of the candidate parking point, the vehicle parking orientation, the support leg deployment area, the mast safety distance, radar coverage, and communication line of sight, and using each Boolean constraint as a hard constraint for joint solution in S6.

7. The method according to claim 1, characterized in that, The joint costs used in the joint solution described in S6 include travel time, parking time, deployment time, risk exposure cost, communication penalty, resource conflict penalty, and radar blind zone penalty.

8. The method according to claim 1, characterized in that, The rolling replanning described in S6 further includes: when there is a sensor state conflict, inconsistent model output, communication interruption, or the critical state does not meet the preset safety requirements, stopping the use of the current solution result, performing at least one of re-acquisition, recalculation, rollback of the previous stable result, or manual verification, and saving the triggering reason, input version, and output verification information.

9. The method according to claim 7, characterized in that, The joint solution described in S6 includes: under the conditions of satisfying resource capacity, arrival time limit and deployment hard constraints, calculating the joint cost for the combination of resource allocation, travel path and deployment attitude, and taking the combination with the minimum joint cost as the current solution result.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1 to 9.

Images