Intelligent inspection vehicle dynamic cooperation method based on multi-mode perception fusion and related equipment

Abstract

The invention discloses an intelligent inspection vehicle dynamic collaboration method based on multi-modal sensing fusion and related equipment, which can effectively improve the inspection efficiency and realize efficient and dynamic collaborative operation. The method comprises the steps of receiving an inspection task of a target area; decomposing the inspection task into a plurality of sub-region tasks based on the electronic map of the target region; dynamically allocating a plurality of sub-region tasks to region edge nodes according to a dynamic cooperative control algorithm; receiving target data of each area edge node, wherein the target data is generated after the area edge node performs fusion analysis on multi-mode sensing data reported by the managed intelligent inspection vehicle cluster; performing global monitoring based on the target data to obtain a monitoring result; and when a monitoring result shows that a major abnormal event needing cross-regional cooperation occurs, issuing a cross-regional cooperation instruction to at least two adjacent region edge nodes, so that the intelligent inspection vehicle clusters managed by the at least two adjacent region edge nodes perform cooperative operation.

Description

Technical Field

[0001] This application relates to the field of intelligent inspection vehicle technology, and in particular to a dynamic collaborative method and related equipment for intelligent inspection vehicles based on multimodal perception fusion. Background Technology

[0002] Currently, mobile intelligent inspection vehicles have been applied in various fields, typically possessing conventional functions such as real-time control, map display, alarm statistics, environmental data collection, inspection point chart generation, inspection data statistics, and live video streaming. However, existing inspection vehicles suffer from a lack of collaborative operation capabilities; that is, the operating range of a single inspection vehicle is limited, and there is a lack of efficient task collaboration and data sharing mechanisms between multiple inspection vehicles, making it difficult to achieve three-dimensional inspection coverage over large areas.

[0003] Specifically, existing intelligent inspection systems generally suffer from a lack of cross-regional collaborative operation capabilities when handling major anomalies requiring cross-sub-regional coordination. Traditional solutions often rely on centralized scheduling, resulting in high response latency, limited communication bandwidth, and a lack of dynamic collaborative command issuance and execution mechanisms between regional edge nodes. This makes it difficult for intelligent inspection vehicle clusters to achieve real-time cross-regional linkage, thereby reducing the efficiency of handling emergencies and emergency response capabilities, and hindering efficient and dynamic collaborative operations. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides a dynamic collaborative method and related equipment for intelligent inspection vehicles based on multimodal perception fusion.

[0005] The technical solution provided in this application is described below: The first aspect of this application provides a dynamic collaborative method for intelligent inspection vehicles based on multimodal perception fusion, the method comprising: Receive inspection tasks for the target area; Based on the electronic map of the target area, the inspection task is broken down into multiple sub-area tasks; The multiple sub-region tasks are dynamically assigned to the regional edge nodes according to the dynamic collaborative control algorithm. The target region includes several regional edge nodes, and each regional edge node is used to manage the intelligent inspection vehicle cluster within a sub-region. Receive target data from each of the region edge nodes. The target data is generated by the region edge nodes after fusing and analyzing the multimodal perception data reported by the intelligent inspection vehicle cluster they manage. Perform global monitoring based on the target data to obtain monitoring results; When a major abnormal event requiring cross-regional collaboration occurs in the monitoring results, a cross-regional collaboration command is issued to at least two adjacent regional edge nodes so that the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes can perform collaborative operations.

[0006] Optionally, the inspection task can be decomposed into multiple sub-region tasks based on the electronic map of the target area, including: Obtain an electronic map of the target area; Analyze the electronic map of the target area to obtain the regional information of the target area; Based on the aforementioned regional information, inspection tasks and business constraints are extracted. The sub-region division rules are determined based on the inspection tasks and business constraints. The boundaries of the sub-regions are defined based on the sub-region division rules; The inspection task is decomposed into multiple assignable sub-region tasks based on the sub-region boundaries.

[0007] Optionally, the multiple sub-region tasks are dynamically assigned to region edge nodes according to a dynamic cooperative control algorithm, including: Collect real-time status data of edge nodes in each region to establish a dynamic capability profile of the nodes; Determine the required characteristics of the tasks in the sub-region; The dynamic capability profile of the node and the requirement characteristics of the sub-region task are input into the dynamic collaborative control algorithm to obtain the target fit. Based on the target adaptability, priority scheme, and load balancing scheme, a preliminary allocation scheme is generated; Verify the preliminary allocation scheme. Once it is confirmed that there are no priority conflicts and the load is balanced, the sub-region tasks are distributed to the corresponding region edge nodes.

[0008] Optionally, target data is received from each of the region edge nodes. This target data is generated by the region edge nodes through fusion analysis of the multimodal perception data reported by the managed intelligent inspection vehicle cluster, and includes: The receiving instruction is sent to each region edge node, and the receiving instruction includes the node number of each region edge node; When a send command is received from each distinct edge node, the identifier of the send command is compared with that of the receive command to obtain the comparison result; After confirming that the identification is correct, the target data of each edge node in the region is received. The target data is the multimodal perception data reported by the intelligent inspection vehicle cluster received by the edge node in the region. The multimodal perception data includes visual data, radar data and environmental data. The multimodal perception data is preprocessed to obtain preprocessed data. The preprocessed data is then processed by feature extraction, association matching and anomaly identification to generate the target data.

[0009] Optionally, when a major anomaly requiring cross-regional collaboration occurs in the monitoring results, a cross-regional collaboration command is issued to at least two adjacent regional edge nodes, enabling the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes to perform collaborative operations, including: Obtain monitoring results; The anomaly level is determined based on the monitoring results; When the anomaly level is determined to be high, identify major anomaly events that require cross-regional collaborative processing, and determine the regional scope of the event and the collaborative requirements. Based on the relationship between the regional range of the event and the distribution of the edge nodes of the region, at least two adjacent related edge nodes of the region are selected. Generate cross-regional collaboration instructions, the content of which includes collaboration objectives, scope of work, resource allocation rules, and data interaction requirements; The cross-regional collaboration instruction is sent to at least two adjacent edge nodes of the associated region, so that the at least two adjacent edge nodes control the intelligent inspection vehicle cluster to perform collaborative operations according to the cross-regional collaboration instruction.

[0010] Optionally, global monitoring can be performed based on the target data to obtain monitoring results, including: Obtain target data for all edge nodes in the region; Perform a global correlation analysis on the target data to obtain the correlation relationships between all sub-regions; A global inspection status view is constructed based on the aforementioned relationships; The abnormal information in the global inspection status view is evaluated and classified to obtain global monitoring results.

[0011] Optionally, after performing global monitoring based on the target data to obtain monitoring results, the method further includes: The abnormal events in the monitoring results are classified into different levels to distinguish between routine abnormal events and major abnormal events that require cross-regional coordination.

[0012] A second aspect of this application provides a dynamic collaborative device for intelligent inspection vehicles based on multimodal perception fusion, the device comprising: The first receiving unit is used to receive inspection tasks for the target area. The decomposition unit decomposes the inspection task into multiple sub-region tasks based on the electronic map of the target area. The allocation unit is used to dynamically allocate the multiple sub-region tasks to the regional edge nodes according to the dynamic collaborative control algorithm. The target region includes several regional edge nodes, and each regional edge node is used to manage the intelligent inspection vehicle cluster within a sub-region. The second receiving unit is used to receive target data from each of the regional edge nodes. The target data is generated by the regional edge nodes after fusing and analyzing the multimodal perception data reported by the intelligent inspection vehicle cluster they manage. The acquisition unit performs global monitoring based on the target data to obtain monitoring results. The issuing unit is used to issue cross-regional collaboration instructions to at least two adjacent regional edge nodes when a major abnormal event requiring cross-regional collaboration occurs in the monitoring results, so that the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes can carry out collaborative operations.

[0013] A third aspect of this application provides a dynamic collaborative device for intelligent inspection vehicles based on multimodal perception fusion, the device comprising: Processor, memory, input / output units, and bus; The processor is connected to the memory, the input / output unit, and the bus; The memory stores a program, which the processor invokes to perform the method as described in the first aspect and any one of the first aspects.

[0014] A fourth aspect of this application provides a computer-readable storage medium on which a program is stored, which, when executed on a computer, performs the method as described in the first aspect and any one of the first aspects.

[0015] As can be seen from the above technical solutions, this application has the following beneficial effects: 1. This application decomposes the inspection task of the target area into multiple sub-region tasks and uses a dynamic collaborative control algorithm to dynamically allocate the sub-region tasks to the edge nodes of the region. This fully leverages the advantages of the intelligent inspection vehicle cluster in each region, enables parallel inspection, greatly shortens the inspection time, improves the inspection efficiency, and achieves efficient and dynamic collaborative operation.

[0016] 2. In this application, the regional edge node can fuse and process various types of perception data reported by the intelligent inspection vehicle cluster. Data of different modalities can complement each other, thereby providing more comprehensive and accurate information, thereby improving the detection and recognition accuracy of target objects or events and reducing misjudgments and omissions.

[0017] 3. In this application, the dynamic collaborative control algorithm can adjust the allocation of sub-region tasks in real time according to the actual situation, thereby flexibly responding to various changes. When the inspection task in a certain area becomes heavier or special circumstances occur, inspection vehicles can be promptly dispatched from other areas to provide support, ensuring the smooth progress of the inspection work.

[0018] 4. By receiving target data from edge nodes in various regions for global monitoring, the system can grasp the situation of the entire target area in real time, thereby obtaining comprehensive information, promptly identifying potential problems and anomalies, and making corresponding decisions, thus better serving the security of the target area.

[0019] 5. This technical solution realizes the automation and intelligent inspection of intelligent inspection vehicles, reducing the reliance on manual inspection and thus reducing labor costs. At the same time, intelligent inspection vehicles can replace manual inspection in some dangerous or harsh environments, such as high-voltage areas and toxic and harmful environments, effectively ensuring personnel safety and reducing safety risks. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of an embodiment of the intelligent inspection vehicle dynamic coordination method based on multimodal perception fusion according to this application; Figure 2 This is a schematic diagram of another embodiment of the intelligent inspection vehicle dynamic coordination method based on multimodal perception fusion in this application; Figure 3 This is a schematic diagram of another embodiment of the intelligent inspection vehicle dynamic coordination method based on multimodal perception fusion in this application; Figure 4 This is a schematic diagram of another embodiment of the intelligent inspection vehicle dynamic coordination method based on multimodal perception fusion in this application; Figure 5 This is a schematic diagram of another embodiment of the intelligent inspection vehicle dynamic coordination method based on multimodal perception fusion in this application; Figure 6 This is a schematic diagram of another embodiment of the intelligent inspection vehicle dynamic coordination method based on multimodal perception fusion in this application; Figure 7 This is a schematic diagram of an embodiment of the intelligent inspection vehicle dynamic collaborative device based on multimodal perception fusion according to this application; Figure 8 This is a schematic diagram of another embodiment of the intelligent inspection vehicle dynamic collaborative device based on multimodal perception fusion in this application. Detailed Implementation

[0022] Existing intelligent inspection systems generally suffer from a lack of cross-regional collaborative operation capabilities when handling major anomalies requiring cross-sub-regional coordination. Traditional solutions often rely on centralized scheduling, resulting in high response latency, limited communication bandwidth, and a lack of dynamic collaborative command issuance and execution mechanisms between regional edge nodes. This makes it difficult for intelligent inspection vehicle clusters to achieve real-time cross-regional linkage, thereby reducing the efficiency of handling emergencies and emergency response capabilities, and failing to achieve efficient and dynamic collaborative operations.

[0023] Based on this, this application provides a dynamic collaborative method and related equipment for intelligent inspection vehicles based on multimodal perception fusion. By decomposing the inspection task of the target area into multiple sub-region tasks and using a dynamic collaborative control algorithm to dynamically allocate the sub-region tasks to the edge nodes of the region, the advantages of the intelligent inspection vehicle cluster in each region can be fully utilized to achieve parallel inspection, greatly shorten the inspection time, improve the inspection efficiency, and realize efficient and dynamic collaborative operation.

[0024] It should be noted that the implementing entity of this application is the ground control center, and the following embodiments will use the ground control center as the implementing entity of this solution for explanation.

[0025] Please see Figure 1 The first method of dynamic coordination of intelligent inspection vehicles based on multimodal perception fusion in this application includes: 101. Receive inspection tasks for the target area; 102. Based on the electronic map of the target area, the inspection task is decomposed into multiple sub-area tasks; 103. The multiple sub-region tasks are dynamically allocated to the regional edge nodes according to the dynamic collaborative control algorithm. The target region includes several regional edge nodes, and each regional edge node is used to manage the intelligent inspection vehicle cluster within a sub-region. 104. Receive target data from each of the regional edge nodes, wherein the target data is generated by the regional edge nodes after fusing and analyzing the multimodal perception data reported by the intelligent inspection vehicle cluster they manage; 105. Perform global monitoring based on the target data to obtain monitoring results; 106. When a major abnormal event occurs in the monitoring results that requires cross-regional collaboration, a cross-regional collaboration instruction is issued to at least two adjacent regional edge nodes so that the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes can perform collaborative operations.

[0026] In this embodiment, the inspection task for the target area is first received. Then, based on the electronic map of the target area, the inspection task is decomposed into multiple sub-area tasks. Next, the multiple sub-area tasks are dynamically allocated to the regional edge nodes according to the dynamic collaborative control algorithm. The target area includes several regional edge nodes, and each regional edge node is used to manage the intelligent inspection vehicle cluster in a sub-area. Then, the target data of each regional edge node is received. The target data is generated by the regional edge node after fusing and analyzing the multimodal perception data reported by the intelligent inspection vehicle cluster it manages. Furthermore, global monitoring is performed based on the target data to obtain monitoring results. When the monitoring results show a major abnormal event that requires cross-regional collaboration, a cross-regional collaboration instruction is issued to at least two adjacent regional edge nodes so that the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes can perform collaborative operations.

[0027] Specifically, in step 101, an inspection task for the target area is received. This task is generated by the upper-level management platform based on a preset inspection cycle, the security level requirements of the target area, or an emergency inspection request. The task content must clearly include the geographical boundaries of the target area, the core objectives of the inspection task, the time requirements of the inspection task, and the accuracy standards of the inspection data. When the ground control center receives the inspection task, it will perform a completeness check on the task information to confirm whether it includes the aforementioned key parameters. If any information is missing, it will send a supplementary request to the task initiator until a complete and valid inspection task instruction is obtained.

[0028] In step 102, after the ground control center receives and verifies the target area inspection task, it decomposes the inspection task into multiple sub-area tasks based on the electronic map of the target area. Specifically, the electronic map used here is a high-precision geographic information map pre-built and updated in real time by the system. This map not only contains basic geographic data such as topography, road distribution, and building facilities of the target area, but also marks key elements related to the inspection, such as the location coordinates of the equipment to be inspected, the equipment type, the system to which the equipment belongs, and high-risk areas found in historical inspections.

[0029] During the task decomposition process, the ground control center first extracts the geographical features and inspection element distribution of the target area from the electronic map. It adopts the decomposition principles of geographical continuity and balanced task load to avoid splitting the inspection needs of the same equipment or functional area into different sub-regions. Specifically, the ground control center first identifies naturally formed geographical or functional boundaries within the target area as a reference for initial sub-region division. Then, combining parameters such as the number of equipment to be inspected in each potential sub-region, the required inspection time, and the amount of historical inspection data, it calculates the inspection task load for each potential sub-region. By iteratively adjusting the sub-region boundaries, it ensures that the tasks in the final divided sub-regions are balanced in terms of load, and that the geographical size of each sub-region is moderate. This facilitates centralized management of edge nodes while avoiding situations where excessively large sub-regions lead to reduced inspection efficiency or excessively small sub-regions increase coordination costs.

[0030] For example, if the target area is an industrial park, the ground control center will break down the entire park inspection task into sub-area tasks such as workshop A, workshop B, and public facilities in the park, based on the distribution boundaries of different production workshops in the park and the number and complexity of the mechanical equipment to be inspected in each workshop. Each sub-area task has a clearly defined geographical scope, a list of equipment to be inspected, and inspection requirements.

[0031] In step 103, after the ground control center completes the sub-regional decomposition of the inspection task, it dynamically allocates multiple sub-regional tasks to the regional edge nodes according to the dynamic cooperative control algorithm. First, the ground control center obtains the real-time status information of all regional edge nodes within the target area. This information includes the current computing resource load of each regional edge node, such as CPU utilization, memory usage, communication link status, the number of managed intelligent inspection vehicles and their status, as well as the historical task execution efficiency of the edge node.

[0032] Specifically, to ensure the stability and real-time performance of data transmission between the intelligent inspection vehicle cluster and regional edge nodes, and between regional edge nodes and the ground control center, this application adopts a dual-mode communication method of 4G / 5G public network + LoRa private network: In routine inspection scenarios, LoRa private network communication is preferred. LoRa technology features low power consumption, wide coverage, and strong anti-interference capabilities, making it suitable for intelligent inspection vehicle clusters to periodically report multimodal sensing data to regional edge nodes. This can reduce public network traffic consumption and avoid the impact of public network signal fluctuations on data transmission.

[0033] In emergency coordination scenarios, the system will automatically switch to 4G / 5G public network communication. When a major abnormal event requiring cross-regional coordination occurs in the monitoring results, such as equipment fire or short circuit, the ground control center will trigger a communication mode switch. Through the high bandwidth characteristics of the 4G / 5G public network, the system can achieve real-time transmission of high-definition video and radar point cloud data from the intelligent inspection vehicle, as well as millisecond-level issuance of cross-regional coordination commands between regional edge nodes and the ground control center, ensuring emergency response efficiency.

[0034] The core objective of the dynamic collaborative control algorithm is to achieve optimal matching between sub-region tasks and regional edge node resources. Its specific calculation process includes: First, establishing a matching degree evaluation model between task requirements and node capabilities, quantifying and mapping the requirements of each sub-region task to the capabilities of each regional edge node, and generating an initial matching degree matrix; Second, introducing load balancing constraints to avoid allocating too many sub-region tasks to the same edge node, causing resource overload, while also preventing some edge nodes from being idle; Third, considering geographical proximity, prioritizing the allocation of sub-region tasks to edge nodes geographically closer to the sub-region to reduce communication latency between edge nodes and the intelligent inspection vehicle cluster, improving data transmission efficiency and task response speed; Fourth, solving for the optimal allocation scheme through iterative calculation to ensure that each sub-region task is allocated to the regional edge node with the highest matching degree, sufficient resources, and geographical proximity.

[0035] For example, if the task for a certain sub-region is to inspect the equipment in workshop A, which is closest to edge node C in the park, and edge node C currently has low computing resource load and the number of dispatchable intelligent inspection vehicles meets the inspection needs of workshop A, the dynamic collaborative control algorithm will assign the sub-region task of workshop A to edge node C, and at the same time assign the sub-region task of workshop B to edge node D, which is closer to workshop B and has suitable resources, to achieve reasonable allocation of tasks.

[0036] Once a regional edge node receives its assigned sub-region task, it dispatches the cluster of intelligent inspection vehicles it manages to perform the inspection operation. During the inspection, the intelligent inspection vehicles collect multimodal sensing data, including image data, audio data, and sensor data, and report this data to the corresponding regional edge node in real time. After receiving the multimodal sensing data, the regional edge node initiates a data fusion and analysis process. This process includes data preprocessing, data fusion, feature extraction, and anomaly identification, ultimately generating target data containing information such as a summary of the inspection equipment status within the sub-region, the location and type of anomalies, and inspection data statistics. This target data is then reported to the ground control center. Next, in step 104, the ground control center receives the target data reported by each area edge node and verifies the completeness and validity of each target data. If the target data reported by an edge node is found to be missing or abnormal, such as the absence of key equipment status information or incorrect data format, a data retransmission or supplementary analysis request is sent to that edge node to ensure that valid target data corresponding to all sub-area tasks is obtained.

[0037] In step 105, after the ground control center acquires the target data of all area edge nodes, it performs global monitoring based on the target data to obtain monitoring results. Specifically, the ground control center first integrates the target data of each sub-area globally to construct a global view of inspection data covering the entire target area. This view not only includes the real-time status of equipment in each sub-area but also presents the correlation between equipment in different sub-areas, such as the connection relationship between transformers and power poles in different sub-areas on the same power line.

[0038] The specific process of global monitoring includes: First, performing cross-sub-region correlation analysis on the data in the global view to identify potential problems that are not apparent in a single sub-region but exist from a global perspective. For example, if the target data of sub-region A in workshop A shows that the temperature of a certain motor is slightly high, while the target data of sub-region B in workshop B shows that the transmission equipment connected to the motor is vibrating abnormally, cross-sub-region correlation analysis can determine that there may be potential faults in the entire transmission link. Second, comparing and analyzing the current global inspection data with historical inspection data to identify the changing trend of equipment status. For example, if the temperature value of a certain piece of equipment gradually increases in three consecutive inspections, the system can determine that the equipment has a risk of performance degradation. Third, according to the preset monitoring rules and anomaly judgment criteria, performing anomaly detection on the globally integrated data and generating monitoring results. The monitoring results are divided into normal status, general abnormal status, and major abnormal status.

[0039] The "normal state" indicates that all inspection equipment within the entire target area is operating normally. The "generally abnormal state" indicates a minor, localized anomaly within a sub-area, such as a single device parameter deviating from the normal range but not affecting the overall system operation. The "major abnormal state" indicates an abnormal event affecting multiple sub-areas and requiring cross-regional collaborative handling, such as large-scale equipment failure or significant safety hazards. It should be noted that after obtaining the monitoring results, the abnormal events in the monitoring results will be classified into different levels to distinguish between routine abnormal events and major abnormal events requiring cross-regional collaboration.

[0040] For example, if the system detects through global monitoring that the power supply lines of workshops A and B in the park are simultaneously experiencing voltage anomalies, and this anomaly has caused multiple critical pieces of equipment in the two workshops to shut down, this situation is determined to be a major anomaly event requiring cross-regional coordination, and the corresponding monitoring result is a major anomaly status.

[0041] When the monitoring result obtained in step 105 indicates a major anomaly requiring cross-regional collaboration, step 106 is executed. At this point, the ground control center issues cross-regional collaboration instructions to at least two adjacent regional edge nodes, enabling the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes to perform collaborative operations. First, the ground control center needs to determine the scope of regional edge nodes participating in cross-regional collaboration. Based on the impact range and geographical distribution of the major anomaly, it selects the regional edge nodes that are most closely related to the anomaly, geographically adjacent, and possess collaborative capabilities.

[0042] For example, if a major anomaly is a power line failure between workshops A and B, the system will select edge node C, which manages workshop A, and edge node D, which manages workshop B, as collaborative nodes. Next, a cross-regional collaborative instruction is generated. This instruction must clearly define the collaborative operation's objectives, specific task allocations, timeframes, data collaboration requirements, and emergency response plans. Finally, the ground control center sends the cross-regional collaborative instruction to the selected adjacent edge nodes via a highly reliable communication link. Upon receiving the instruction, the edge nodes immediately adjust the operational plans of their managed intelligent inspection vehicle clusters, dispatching vehicles to execute collaborative tasks according to the instructions, and providing real-time progress data to the ground control center. The ground control center dynamically monitors the collaborative operation process. If resource shortages or task execution deviations are detected, the system promptly adjusts the collaborative instruction to ensure efficient completion of the cross-regional collaborative operation and ultimately resolve the major anomaly.

[0043] For example, after receiving a cross-regional collaborative instruction, edge nodes C and D respectively dispatch their corresponding intelligent inspection vehicle clusters to check for power line faults from both ends of workshops A and B toward the middle. At the same time, they share inspection data in real time and finally find a short circuit point in the line between the two workshops. They then collaboratively complete the fault point marking and on-site situation reporting, providing support for subsequent maintenance personnel to quickly handle the fault.

[0044] In the above technical solution, by decomposing the inspection task of the target area into multiple sub-area tasks and using a dynamic collaborative control algorithm to dynamically allocate the sub-area tasks to the edge nodes of the area, the advantages of the intelligent inspection vehicle cluster in each area can be fully utilized to achieve parallel inspection, greatly shorten the inspection time, improve the inspection efficiency, and realize efficient and dynamic collaborative operation.

[0045] Please refer to Figure 2 According to some embodiments of the present invention, the step 102 of decomposing the inspection task into multiple sub-region tasks based on the electronic map of the target area may specifically include, but is not limited to, the following: 201. Obtain an electronic map of the target area; 202. Parse the electronic map of the target area to obtain the regional information of the target area; 203. Extract inspection tasks and business constraints based on the aforementioned area information; 204. Determine the sub-region division rules based on the aforementioned inspection tasks and business constraints; 205. Determine the boundaries of the sub-regions based on the aforementioned sub-region division rules; 206. Decompose the inspection task into multiple assignable sub-region tasks according to the sub-region boundaries.

[0046] In this embodiment, when the ground control center executes the inspection task decomposition process, it first acquires an electronic map of the target area. This electronic map is a professional map, and its data sources include geographic information collected from previous on-site surveys, equipment asset ledger data of the target area's industry, and dynamic information updated during historical inspections. During the acquisition process, the ground control center establishes a secure and encrypted communication link with the map management server, prioritizing the retrieval of the latest version of the electronic map. If historical map data for the target area exists in the local cache, it will automatically compare the version with the server-side data to ensure that the acquired electronic map contains the latest geographic features, equipment distribution, and environmental parameters.

[0047] For example, for an inspection task in an industrial park, the electronic map obtained by the system needs to clearly mark the specific location and boundaries of each production workshop in the park, the installation coordinates of each piece of machinery in the workshop, the fire lanes and restricted areas in the park, the direction of power lines and pipelines, and information such as recently added temporary storage areas, so as to provide an accurate data foundation for subsequent regional information analysis and task division.

[0048] After acquiring the electronic map of the target area, the ground control center parses the map to obtain regional information. Then, it extracts and correlates the structured and unstructured data stored in the electronic map. The regional information mainly includes three core categories: First, geospatial information, covering the overall geographical scope of the target area, topographic features, and natural and man-made boundaries; second, equipment asset information, including the basic attributes of all equipment to be inspected within the area, their functional relationships, and inspection priorities; and third, environmental and constraint information, including safety control requirements, access restrictions, and historical inspection records.

[0049] For example, when parsing the electronic map of an industrial park, it is necessary to clearly extract the boundary coordinates of workshop A as (X1,Y1) to (X2,Y2). The workshop contains 3 production lines with a total of 28 mechanical devices. Among them, the 5 core motors of production line 1 are high-priority inspection devices. Workshop A and workshop B are separated by a main road with a width of 8 meters. This main road allows inspection vehicles to pass through all day. At the same time, the temporary storage area on the east side of workshop A is a recent addition. Its inspection frequency requirements need to be specially marked in the area information to ensure that the subsequent task division can fully adapt to these area characteristics.

[0050] After obtaining complete regional information, the ground control center extracts inspection tasks and operational constraints based on this information. The extraction of inspection tasks must be based on the equipment asset attributes and functional requirements within the regional information, breaking down the overall inspection objective into specific equipment-level or function-level inspection tasks. For example, based on equipment information parsed from the industrial park's electronic map, specific inspection tasks can be extracted such as "inspecting the operating temperature and vibration frequency of the five core motors on production line 1 in workshop A," "checking for loose wiring terminals in the power control cabinet of workshop B," and "investigating leaks in the park's water supply network." Each task clearly defines the corresponding inspection object, inspection parameters, and judgment criteria.

[0051] The extraction of business constraints needs to be combined with environmental constraints, safety control requirements, and equipment relationships in the regional information. These mainly include time constraints, resource constraints, collaboration constraints, and priority constraints. For example, for inspection tasks in an industrial park, the extracted business constraints include: "During the production period from 8:00 to 18:00 daily, inspection vehicles are not allowed to enter the production line operation area of ​​Workshop A, and can only conduct equipment appearance inspections in the outer passage of the workshop"; "Only explosion-proof inspection vehicles are allowed to enter the hazardous chemical storage area (Area C) in the park, and the inspection data must be uploaded to the ground control center in real time"; "Power line No. 1, which spans Workshops A and B, must be inspected in one go and cannot be split into different sub-area tasks." These constraints will directly guide the determination of subsequent sub-area division rules.

[0052] Based on the extracted inspection tasks and business constraints, the rules for dividing sub-regions are further determined. These rules can transform the inspection task requirements and constraints into quantifiable and executable division criteria, specifically including the following aspects: First, the physical boundary priority rule, that is, to prioritize natural or artificial boundaries as the basis for dividing sub-areas, and avoid splitting inspection tasks within the same physical separation unit into different sub-areas. For example, in an industrial park, workshop walls and main roads are used as boundaries to ensure that each sub-area corresponds to a complete physical unit. Second, the task association rules stipulate that for inspection tasks with functional associations, such as power line inspections across workshops or inspections of linked equipment on the production line, they should be included in the same sub-area or the sub-areas should be able to support the collaborative execution of such tasks. For example, the line segment corresponding to the inspection task of power line No. 1 across workshops A and B should be assigned to the sub-area of ​​workshop A and workshop B respectively, and the rules should clearly state that the two sub-areas should work together to complete the inspection of the line. Third, load balancing rules ensure that the workload of each sub-region is relatively balanced based on the inspection workload and inspection difficulty in each potential sub-region, so as to avoid inspection delays caused by excessive workload in a certain sub-region or waste of resources caused by excessively light workload in a certain sub-region. Fourth, constraint adaptation rules. For business constraints, corresponding division adaptation rules are formulated, such as "sub-regions containing hazardous chemical storage areas need to be divided independently, and only explosion-proof inspection vehicle resources are allocated to these sub-regions" and "areas with restricted access during production periods need to be divided separately from areas that can be freely inspected, so as to facilitate the formulation of separate inspection period plans." These rules together constitute the basis for sub-region division.

[0053] After determining the rules for dividing the sub-regions, the ground control center delineates the boundaries of the sub-regions based on these rules. During the conversion process, the visualization interface of the electronic map and spatial analysis algorithms are combined to translate the division rules into specific sub-region boundary coordinates. Specifically, the ground control center marks physical dividing lines that can serve as sub-region boundaries on the electronic map according to the "physical boundary priority rule," such as the coordinate lines of workshop walls or the center lines of main roads, and uses these as a basis to initially outline the approximate extent of the sub-regions.

[0054] Next, based on the task relevance rules, the initial scope is adjusted. For example, if the length of the No. 1 power line spanning workshops A and B is 500 meters in workshop A and 300 meters in workshop B, the ground control center will ensure that the boundary of the workshop A sub-region includes the complete section of the line within workshop A, and the boundary of the workshop B sub-region includes the complete section of the line within workshop B, to avoid the line being cut off by the boundary. Then, according to the load balancing rules, the inspection workload of each initial sub-region is calculated. If the number of inspection devices in a certain initial sub-region reaches 45, far exceeding the threshold of 30, the ground control center will further analyze the equipment distribution and physical separation within the workshop. If it is found that there is an internal passage within workshop A that can divide the workshop into two relatively independent equipment clusters on the east and west sides, and the workload of the 22 devices on the east side and the 23 devices on the west side both meet the load requirements, then workshop A will be divided into two sub-regions, A1 and A2, with this internal passage as the boundary.

[0055] Finally, based on the constraint adaptation rules, the boundaries of special areas are confirmed. For example, the boundary of the hazardous chemical area C must be strictly defined according to the safety isolation zone marked on the electronic map, ensuring that the sub-area boundary is completely consistent with the safety isolation zone boundary, and avoiding the inclusion of non-hazardous areas into the sub-area or the omission of hazardous areas. For example, the final sub-area boundaries of the industrial park include sub-area A1, sub-area A2, workshop B, hazardous chemical C, and public facilities sub-area. The boundaries of each sub-area are based on the physical separation in the electronic map and meet the requirements of task association, load balancing, and constraint adaptation.

[0056] After defining the sub-region boundaries, the ground control center breaks down the overall inspection task into multiple assignable sub-region tasks based on these boundaries. Specifically, the ground control center establishes an "equipment-sub-region" association table. Based on the matching relationship between the coordinates of each piece of equipment to be inspected and the coordinates of the sub-region boundary, it determines the sub-region to which the equipment belongs. For example, the 12 motors and 8 pumps within sub-region A1 are all included in the inspection object list for sub-region A1. Then, combined with the corresponding business constraints of the sub-region, execution parameters are added to the sub-region tasks. For instance, the inspection time for sub-region A1 is set to 18:30-20:30 daily (non-production hours), the inspection equipment must use a standard intelligent inspection vehicle, and the inspection data must be... Upload every 30 minutes; then clarify the goals and acceptance criteria of the sub-region tasks. For example, the A1 sub-region task needs to complete the temperature, vibration, and appearance checks of all 20 devices to ensure that the device parameters are within the normal range and the inspection data accuracy rate is not less than 98%; finally, generate a standardized sub-region task document, which includes information such as sub-region name, boundary coordinates, inspection object list, execution time period, resource requirements, acceptance criteria, and collaboration requirements, to ensure that each sub-region task is independent and assignable and can be directly issued to the corresponding regional edge node for execution.

[0057] For example, the overall inspection task of the industrial park is ultimately broken down into the inspection tasks of sub-area A1 (including 20 pieces of equipment, to be executed from 18:30 to 20:30), sub-area A2 (including 23 pieces of equipment, to be executed from 18:30 to 20:30), workshop B (including 28 pieces of equipment, to be executed from 19:00 to 21:00), hazardous chemicals C (including 5 storage devices, to be executed from 20:00 to 21:00, requiring an explosion-proof inspection vehicle), and public facilities in the park (including 15 fire-fighting devices and 3 pipelines, to be executed from 21:00 to 22:30). Each sub-area task can be independently assigned to the corresponding area edge node, and can meet the overall inspection objectives and business constraints.

[0058] Please refer to Figure 3According to some embodiments of the present invention, the dynamic allocation of the multiple sub-region tasks to the region edge nodes in step 103 according to the dynamic cooperative control algorithm may specifically include, but is not limited to, the following: 301. Collect real-time status data of edge nodes in each region and establish a dynamic capability profile of the nodes; 302. Determine the requirements and characteristics of the sub-regional tasks; 303. Input the dynamic capability profile of the node and the requirement characteristics of the sub-region task into the dynamic collaborative control algorithm to obtain the target fit. 304. Based on the target adaptability, priority scheme, and load balancing scheme, generate a preliminary allocation scheme; 305. Verify the preliminary allocation scheme. Once it is confirmed that there are no priority conflicts and the load is balanced, distribute the sub-region tasks to the corresponding region edge nodes.

[0059] In this embodiment, when dynamically allocating tasks from multiple sub-regions to edge nodes according to the dynamic collaborative control algorithm, real-time status data of each edge node is first collected to establish a dynamic capability profile of the node. The real-time status data collection of the edge nodes needs to cover the core dimensions of node operation to ensure data comprehensiveness and timeliness. The collection process is implemented through a node status monitoring module built into the ground control center. This module uses a 1-minute collection cycle and acquires multi-dimensional data through the sensors and communication interfaces built into the edge nodes.

[0060] After completing the collection of multi-dimensional real-time status data, the ground control center preprocesses the data, including removing outliers, standardizing the data, and assigning weights, ultimately integrating the data to form a dynamic capability profile of the nodes. This profile is presented in structured data format. For example, the capability profile of edge node 1 in workshop A can be labeled as "computing resource score 82, communication capability score 78, equipment scheduling score 90 (can schedule 8 inspection vehicles, 3 of which are equipped with infrared thermal imagers), task execution score 85, and comprehensive capability score 84". The profile is updated every minute in sync with the real-time status data to ensure that it accurately reflects the node's current actual capabilities.

[0061] After establishing the dynamic capability profile of the nodes, the demand characteristics of sub-region tasks are determined. This process requires extracting key characteristic indicators based on the decomposed sub-region task content, starting from the core requirements of task execution, to ensure accurate matching with the node capability profile. First, the task type characteristics are clarified. Sub-region tasks are categorized according to differences in inspection objectives, such as power line inspection tasks, equipment fault diagnosis tasks, and environmental parameter monitoring tasks. Different types of tasks correspond to different capability requirements. Next, task load characteristics are extracted, including the number of devices to be inspected in the sub-region, the amount of multimodal data required for the inspection of a single device, and the real-time requirements for data processing. This quantifies the intensity of the task's demand for node computing and communication resources.

[0062] Next, the geographical scope characteristics are further determined, clarifying the geographical boundaries corresponding to the sub-region tasks, the terrain complexity within the region, and the straight-line distance between the sub-region and the edge nodes of each region. Geographical scope characteristics directly affect the dispatch efficiency and communication latency of the inspection vehicle. Finally, task priority characteristics are labeled, and priority levels are set according to the urgency and impact of the tasks. For example: Level 1 is an urgent task, requiring completion within 1 hour; Level 2 is a routine task, requiring completion within 4 hours; and Level 3 is a low-priority task, requiring completion within 8 hours. Through the above feature extraction, a structured sub-region task requirement feature table is formed.

[0063] After preparing the dynamic capability profiles of nodes and the task requirement characteristics of sub-regions, the node capability profiles and task requirement characteristics are input into the dynamic collaborative control algorithm to obtain the target fit. Specifically, the core of this dynamic collaborative control algorithm is to construct a multi-dimensional fit calculation model to achieve quantitative matching between node capabilities and task requirements.

[0064] First, establish a feature mapping relationship, associating each dimension of the task requirement feature with the corresponding dimension of the node capability profile. For example, the task requirement of "data real-time requirement of 10 seconds" corresponds to the node capability of "data processing response delay", the task requirement of "mountainous terrain" corresponds to the node capability of "inspection vehicle off-road capability score", and the task requirement of "first priority" corresponds to the node capability of "emergency task processing efficiency".

[0065] Next, the adaptation scoring rules are set. For each related dimension, the adaptation score of that dimension is obtained by difference calculation and comparison. For example, if the task requires data real-time performance within 10 seconds, and the data processing response delay of a certain edge node is 8 seconds, the adaptation score of that dimension is 90 points; if the response delay of another edge node is 15 seconds, the adaptation score of that dimension is 40 points; if the task requires mountainous terrain, and the off-road capability score of a certain edge node inspection vehicle is 85 points, the adaptation score of that dimension is 85 points; if the node inspection vehicle only has the ability to drive on flat ground with a score of 30 points, the adaptation score of that dimension is 30 points.

[0066] Next, weighting coefficients are introduced for comprehensive calculation. Based on the degree of dependence of the task type on each requirement dimension, weights are assigned to the adaptation score of each dimension. For example, in the power line inspection task, the weight of the data real-time dimension is 0.3, the weight of the inspection vehicle equipment configuration dimension is 0.25, the weight of the communication stability dimension is 0.2, the weight of the geographical distance dimension is 0.15, and the weight of the emergency handling capability dimension is 0.1. Then, the comprehensive target adaptation degree of a single node and a single task is calculated by the weighted summation formula: Target Adaptability = Σ (Adaptability score of a certain dimension × weight of that dimension). For example, the adaptation process for a power line inspection task in a certain sub-region with edge node 1 is as follows: data real-time performance score: 90 × 0.3 = 27; inspection vehicle equipment configuration score: 85 × 0.25 = 21.25; communication stability score: 80 × 0.2 = 16; geographical distance score: 85 × 0.15 = 12.75; emergency handling capability score: 90 × 0.1 = 9; overall target adaptation score = 27 + 21.25 + 16 + 12.75 + 9 = 86 points. The calculation result for the same task with edge node 2 is: data real-time performance score: 40 × 0.3 = 12; inspection vehicle equipment configuration score: 85 × 0.25 = 21.25; communication stability score: 75 × 0.2 = 15; geographical distance score: 60 × 0.15 = 9; emergency handling capability score: 80 × 0.1 = 8; overall target adaptation score = 12 + 21.25 + 15 + 9 + 8 = 65.25 points.

[0067] The algorithm clearly demonstrates that edge node 1 has a higher fit with the target of this task. Simultaneously, the algorithm calculates the fit between all region edge nodes and the currently assigned sub-region task, generating a fit ranking table.

[0068] After obtaining the target fit between each node and the task, a preliminary allocation plan is generated based on the target fit, priority scheme, and load balancing scheme. Specifically, the application logic of the priority scheme is executed first. The ground control center sorts the allocation order according to the priority level of the sub-region tasks (Level 1 > Level 2 > Level 3), prioritizing high-priority tasks to ensure that urgent tasks can be quickly matched with suitable nodes. In the allocation of tasks of the same priority, the task is first assigned to the edge node with the highest fit according to the target fit sorting table.

[0069] Secondly, a load balancing scheme is introduced for allocation adjustment to avoid resource overload caused by a single node taking on too many tasks. Furthermore, for sub-region tasks with interrelationships, the ground control center, based on load balancing and adaptability, prioritizes their allocation to the same or adjacent edge nodes to reduce subsequent coordination costs. Through the synergistic effect of the above priority ranking, adaptability matching, and load balancing adjustment, the ground control center determines the initial allocation targets for all sub-region tasks to be assigned, forming a preliminary allocation scheme that includes "task number - assigned edge node - task load - estimated completion time".

[0070] After generating the initial allocation scheme, the initial allocation scheme is verified. Once it is confirmed that there are no priority conflicts and the load is balanced, the sub-region tasks are distributed to the corresponding region edge nodes.

[0071] Please refer to Figure 4 According to some embodiments of the present invention, in step 104, target data is received from each of the regional edge nodes. The target data is generated by the regional edge nodes through fusion analysis of the multimodal perception data reported by the intelligent inspection vehicle cluster they manage. Specifically, it may include, but is not limited to, the following: 401. Send the receiving instruction to each region edge node, wherein the receiving instruction includes the node number of each region edge node; 402. When a transmission command is received from each distinct edge node, the identifier of the transmission command is compared with the identifier of the reception command to obtain the comparison result; 403. After confirming that the comparison result is correct, the target data of each edge node of the region is received. The target data is the multimodal perception data reported by the intelligent inspection vehicle cluster received by the edge node of the region. The multimodal perception data includes visual data, radar data and environmental data. The multimodal perception data is preprocessed to obtain preprocessed data. The preprocessed data is then processed by feature extraction, association matching and anomaly identification to generate the target data.

[0072] In this embodiment, a receiving instruction is sent to each regional edge node, and the receiving instruction must include the node number of each regional edge node. Before sending the receiving instruction to the regional edge nodes, the ground control center first verifies the identity of all regional edge nodes assigned sub-regional tasks within the target area. Each regional edge node has a unique node number, which is bound to the edge node's hardware device information, network address information, and managed sub-regional range information to ensure that each node number corresponds to a unique regional edge node. When generating the receiving instruction, the system identifies the regional edge node corresponding to each receiving instruction based on the previous task allocation results, embeds the node's unique number into the header identifier field of the receiving instruction, and also includes the data reception time window, target data format requirements, and data transmission encryption protocol in the receiving instruction.

[0073] After generating the receive command, the ground control center sends the corresponding receive command to each area edge node via a pre-set dedicated communication link. After issuing the receive command, the ground control center enters a data reception state. When it receives a send command from each area edge node, it compares the identifier of the send command with that of the receive command to obtain the comparison result. Upon receiving the receive command from the system, the area edge node first parses the node number in the command to confirm whether the receive command is for itself. If the node number matches, it prepares to send the generated target data to the ground control center according to the data format and transmission requirements in the receive command. Before sending the target data, it first sends a send command to the ground control center. This send command is automatically generated by the area edge node and contains core identification information that corresponds to the identification information in the receive command, specifically including the area edge node number corresponding to the send command, the data transmission request sequence number, and the checksum of the receive command.

[0074] After receiving a transmission command from an edge node in a certain area, the ground control center will immediately initiate an identification comparison process. After confirming that the identification is correct, it will receive the target data from each edge node in the area. The generation of this target data requires a series of processing steps by the edge nodes in the area to obtain the multimodal perception data reported by the intelligent inspection vehicle cluster. After obtaining the target data, preprocessing is performed.

[0075] After preprocessing, the regional edge nodes perform feature extraction, correlation matching, and anomaly identification on the preprocessed data. Finally, the regional edge nodes integrate the equipment inspection features, correlation results, anomaly identification results, and key indicators of the original preprocessed data in all sub-regions to generate structured target data. The target data is then sent to the ground control center, which successfully receives the target data after confirming that the identification is correct, providing complete and effective data support for subsequent global monitoring.

[0076] Please refer to Figure 5 According to some embodiments of the present invention, when a major abnormal event requiring cross-regional coordination occurs in the monitoring results in step 106, a cross-regional coordination instruction is issued to at least two adjacent regional edge nodes so that the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes can perform coordinated operations. Specifically, this may include, but is not limited to, the following: 501. Obtain monitoring results; 502. Determine the anomaly level based on the monitoring results; 503. When the anomaly level is determined to be high, identify major anomaly events that require cross-regional collaborative processing, and determine the regional scope of the event and the collaborative requirements. 504. Based on the relationship between the regional range of the event and the distribution of the edge nodes of the region, filter out at least two adjacent related edge nodes of the region; 505. Generate cross-regional collaboration instructions, the content of which includes collaboration objectives, scope of work, resource allocation rules, and data interaction requirements; 506. Send the cross-regional collaboration instruction to at least two adjacent associated region edge nodes, so that the at least two adjacent region edge nodes control the intelligent inspection vehicle cluster to perform collaborative operations according to the cross-regional collaboration instruction.

[0077] In this embodiment, when a major anomaly requiring cross-regional coordination is detected by the monitoring results, a cross-regional coordination command is issued to at least two adjacent regional edge nodes, enabling the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes to perform coordinated operations. Specifically, the monitoring results are first obtained. These monitoring results originate from the global integration and analysis of target data reported by each regional edge node by the ground control center. This result is not a local status feedback from a single sub-region, but rather a comprehensive inspection status presentation covering the entire target region.

[0078] Specifically, the monitoring results include not only equipment status data collected by the intelligent inspection vehicle cluster in each sub-region and fused and analyzed by edge nodes, but also potential risk warnings obtained by the ground control center through cross-sub-region correlation analysis and historical data comparison analysis, as well as statistical information on the progress of the inspection task in the entire target area.

[0079] After acquiring the monitoring results, the anomaly level is determined based on these results. To standardize and refine the anomaly level determination, when the anomaly level is determined to be high, major anomaly events requiring cross-regional collaborative processing are identified, and the regional scope and collaborative requirements of the events are determined. After identifying major anomaly events and determining collaborative requirements, at least two adjacent related regional edge nodes are selected based on the regional scope of the events and the distribution relationship of the regional edge nodes. The ground control center first retrieves the distribution location information of all regional edge nodes within the target area, performs spatial overlay analysis of the regional scope of the events and the distribution of edge nodes, and initially filters out edge nodes whose geographical locations intersect or are adjacent to the event's affected area, excluding edge nodes that have no geographical connection to the event area. Subsequently, the ground control center performs capability adaptability verification on the initially selected edge nodes. Verification indicators include: the current resource idleness of the edge node, the status of the dispatchable intelligent inspection vehicle, the stability of the communication link, and the edge node's experience in handling similar anomaly events in the past.

[0080] For edge nodes that fail the capability verification, the ground control center removes them from the candidate list. Finally, the ground control center combines the geographical proximity and mission relevance of the edge nodes to determine the final associated edge nodes, ensuring that at least two selected edge nodes are not only geographically adjacent but also complement each other in terms of mission.

[0081] After identifying the edge nodes of the associated regions, cross-regional collaboration instructions are generated. These instructions include collaboration objectives, work scope, resource allocation rules, and data interaction requirements. The collaboration objectives must align with the priority of handling major anomalies and be clearly defined and quantifiable, avoiding vague statements. For example, "Complete the precise location of power line faults in the core areas of workshops A and B within 25 minutes, and simultaneously complete the status monitoring of power supply equipment in the peripheral area of ​​workshop C to ensure no new anomalies," rather than the general statement "Handle power faults." Finally, the cross-regional collaboration instructions are sent to at least two adjacent edge nodes of the associated regions, enabling these nodes to control the intelligent inspection vehicle cluster to perform collaborative operations according to the instructions.

[0082] Please refer to Figure 6 According to some embodiments of the present invention, step 105, which involves performing global monitoring based on the target data to obtain monitoring results, may specifically include, but is not limited to, the following: 601. Obtain target data for all region edge nodes; 602. Perform a global correlation analysis on the target data to obtain the correlation relationships between all sub-regions; 603. Construct a global inspection status view based on the aforementioned relationships; 604. The abnormal information in the global inspection status view is evaluated and classified to obtain the global monitoring results.

[0083] In the process of globally monitoring the target data to obtain monitoring results, specifically, the target data of all regional edge nodes is first acquired. This target data is structured data generated by each regional edge node after fusing and analyzing the multimodal perception data reported by the intelligent inspection vehicle clusters in the sub-regions it manages. During the acquisition process, the ground control center does not simply receive the data; instead, it first establishes dedicated communication links with each regional edge node. These links employ encrypted transmission protocols to ensure data security, and a dynamic bandwidth adjustment mechanism ensures that transmission congestion is avoided when multiple edge nodes report data simultaneously.

[0084] The ground control center performs legality and integrity checks on the target data reported by each area edge node. Legality checks primarily verify whether the data source is a registered area edge node and whether the data format conforms to the system's preset standards. Integrity checks inspect whether the target data covers the status information of all devices to be inspected within the sub-area and whether there are any missing data or empty fields. If the data reported by an area edge node has legality issues, the ground control center will immediately interrupt the data transmission and issue a safety warning. If there are integrity issues, the ground control center will send a data retransmission command to the edge node, requiring it to re-perform secondary sensing analysis on the devices corresponding to the missing data and report it again, until complete and legal target data from all area edge nodes is obtained, providing a reliable data foundation for subsequent global correlation analysis.

[0085] After acquiring and verifying the target data of all regional edge nodes, a global correlation analysis is performed on the target data to obtain the correlation between all sub-regions. Since the sub-regions are geographically adjacent and may have device linkage or system coupling relationships in terms of function, the target data of a single sub-region cannot reflect this cross-regional correlation. Therefore, the hidden correlation information is mined through global correlation analysis.

[0086] After obtaining the relationships between all sub-regions, a global inspection status view is constructed based on these relationships. The global inspection status view is not simply an overlay of sub-region data, but rather an integrated, logically connected, global, and visualized data model that integrates scattered sub-region target data through these relationships. First, the ground control center determines the core components of the global inspection status view, including a basic geographic layer, an equipment status layer, a relationship layer, and a dynamic update layer: the basic geographic layer uses a high-precision electronic map of the target area as its base map, marking the boundaries of all sub-regions, the locations of regional edge nodes, and the real-time location of the intelligent inspection vehicle, providing a geographic reference for the global view. The device status layer presents the status data of all inspected devices in each sub-region using icons of different colors at the corresponding locations in the geographic layer. For example, devices operating normally use green icons, devices with minor anomalies use yellow icons, and devices with serious malfunctions use red icons. Hovering the mouse over a device icon displays detailed parameters. The relationship layer connects related devices or sub-regions with different styles of lines based on the acquired relationships. For example, geographic relationships are connected with solid black lines, functional relationships with dashed blue lines, and temporal relationships with dotted orange lines. The thickness of the lines corresponds to the strength of the relationship, visually demonstrating the linkage between sub-regions. The dynamic update layer sets up a data refresh mechanism to synchronize the latest target data reported by the edge nodes of each region according to a preset cycle, updating the content of the device status layer and the relationship layer in real time to ensure that the global inspection status view can reflect the latest inspection situation.

[0087] After constructing the various components, the ground control center integrates the four layers of content into a unified global inspection status view through visualization rendering. Users can zoom and pan the view to view the details of different areas, and can also focus on specific targets through the filtering function. At the same time, the system supports the view data export function, which can save the current global inspection status in report or image format to provide a basis for subsequent analysis and decision-making.

[0088] After constructing the global inspection status view, the abnormal information in the view is assessed and classified to obtain global monitoring results. First, the ground control center needs to extract all abnormal information from the global inspection status view. This abnormal information includes both local abnormalities of equipment within a single sub-region and cross-regional abnormalities caused by factors related to different regions. During the extraction process, each abnormal information is labeled with the corresponding equipment identifier, sub-region affiliation, abnormality occurrence time, list of associated equipment, and original abnormal data.

[0089] Subsequently, the ground control center employed a multi-dimensional indicator evaluation method to assess the level of the extracted anomaly information. The evaluation indicators included three core metrics: the scope of the anomaly's impact, its severity, and its development trend. Based on the assessment results of these three metrics, the ground control center categorized the anomaly information into four levels according to pre-defined grading rules: Level 1 (major anomaly), Level 2 (significant anomaly), Level 3 (general anomaly), and Level 4 (minor anomaly). After the grading assessment, the system further classified the anomaly information, including anomaly type, the system to which the anomaly belongs, and the type of anomaly association. This classification enabled structured management of the anomaly information.

[0090] Finally, the ground control center integrates the level assessment results and classification results of all anomalies, and combines them with the overall operation status of the global inspection status view to generate a global monitoring result containing four modules: "Global Operation Overview", "Anomaly Level Distribution", "Anomaly Classification Statistics" and "Key Anomaly Details", thereby providing a clear decision-making basis for subsequent anomaly handling.

[0091] Please see Figure 7 The second aspect of this application provides a dynamic collaborative device for intelligent inspection vehicles based on multimodal perception fusion, the device comprising: The first receiving unit 701 is used to receive the inspection task of the target area; Decomposition unit 702 decomposes the inspection task into multiple sub-region tasks based on the electronic map of the target area; The allocation unit 703 is used to dynamically allocate the multiple sub-region tasks to the regional edge nodes according to the dynamic collaborative control algorithm. The target region includes several regional edge nodes, and each regional edge node is used to manage the intelligent inspection vehicle cluster within a sub-region. The second receiving unit 704 is used to receive target data from each of the regional edge nodes. The target data is generated by the regional edge nodes after fusing and analyzing the multimodal perception data reported by the intelligent inspection vehicle cluster they manage. The acquisition unit 705 performs global monitoring based on the target data to obtain monitoring results. The issuing unit 706 is used to issue cross-regional collaboration instructions to at least two adjacent regional edge nodes when a major abnormal event requiring cross-regional collaboration occurs in the monitoring results, so that the intelligent inspection vehicle clusters managed by at least two adjacent regional edge nodes can carry out collaborative operations.

[0092] Please see Figure 8 This application also provides a dynamic collaborative device for intelligent inspection vehicles based on multimodal perception fusion, the device comprising: Processor 801, memory 802, input / output unit 803, bus 804; The processor 801 is connected to the memory 802, the input / output unit 803, and the bus 804; The memory 802 stores a program, and the processor 801 calls the program to execute any of the methods described above.

[0093] This application also relates to a computer-readable storage medium on which a program is stored, which, when run on a computer, causes the computer to perform any of the methods described above.

[0094] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0096] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0097] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0098] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A multi-modal perception fusion-based intelligent patrol vehicle dynamic coordination method, characterized in that, The method comprises: receiving a target area inspection task; decomposing the inspection task into a plurality of sub-area tasks based on an electronic map of the target area; dynamically assigning the plurality of sub-area tasks to regional edge nodes according to a dynamic coordination control algorithm, the target area comprising a plurality of regional edge nodes, each regional edge node being configured to manage a cluster of intelligent inspection vehicles in a sub-area; receiving target data of each regional edge node, the target data being generated by fusion analysis of multi-modal perception data reported by the cluster of intelligent inspection vehicles managed by the regional edge node; performing global monitoring based on the target data to obtain a monitoring result; when the monitoring result indicates a major abnormal event requiring cross-regional coordination, issuing a cross-regional coordination instruction to at least two adjacent regional edge nodes to enable the cluster of intelligent inspection vehicles managed by the at least two adjacent regional edge nodes to perform coordinated operations.

2. The intelligent patrol vehicle dynamic coordination method based on multi-modal perception fusion according to claim 1, characterized in that, Decomposing the inspection task into a plurality of sub-area tasks based on an electronic map of the target area comprises: obtaining an electronic map of the target area; parsing the electronic map of the target area to obtain regional information of the target area; extracting an inspection task and business constraint conditions according to the regional information; determining a sub-area division rule according to the inspection task and business constraint conditions; defining a sub-area boundary based on the sub-area division rule; decomposing the inspection task into a plurality of assignable sub-area tasks according to the sub-area boundary. 3.The intelligent inspection vehicle dynamic coordination method based on multi-modal perception fusion according to claim 1, characterized in that, Dynamically assigning the plurality of sub-area tasks to regional edge nodes according to a dynamic coordination control algorithm comprises: collecting real-time state data of each regional edge node to establish a node dynamic capability profile; determining a demand feature of the sub-area task; inputting the node dynamic capability profile and the demand feature of the sub-area task into a dynamic coordination control algorithm to obtain a target adaptation degree; generating a preliminary allocation scheme based on the target adaptation degree, a priority scheme, and a load balancing scheme; verifying the preliminary allocation scheme, and when it is determined that there is no conflict in priority and load balancing, assigning the sub-area task to the corresponding regional edge node.

4. The intelligent inspection vehicle dynamic coordination method based on multi-modal perception fusion according to claim 1, characterized in that, Receiving target data of each regional edge node, the target data being generated by fusion analysis of multi-modal perception data reported by the cluster of intelligent inspection vehicles managed by the regional edge node, comprises: sending a receiving instruction to each regional edge node, the receiving instruction comprising a node number of each regional edge node; when receiving a sending instruction sent by each regional edge node, comparing the sending instruction with an identifier of the receiving instruction to obtain a comparison result; after determining that the comparison result is correct, receiving target data of each regional edge node, the target data being multi-modal perception data reported by the cluster of intelligent inspection vehicles managed by the regional edge node, the multi-modal perception data comprising visual data, radar data, and environmental data, and the target data being generated by preprocessing, feature extraction, correlation matching, and abnormality identification processing of the multi-modal perception data.

5. The intelligent inspection vehicle dynamic coordination method based on multi-modal perception fusion according to claim 1, characterized in that, When the monitoring result shows a major abnormal event requiring cross-regional coordination, a cross-regional coordination instruction is issued to at least two adjacent regional edge nodes, so that the intelligent inspection vehicle clusters managed by the at least two adjacent regional edge nodes perform coordinated work, including: obtaining a monitoring result; determining an abnormal level based on the monitoring result; when it is determined that the abnormal level is high, identifying a major abnormal event requiring cross-regional coordination processing, and determining the regional range of the event and the coordination requirement; based on the regional range of the event and the distribution relationship of the regional edge nodes, screening out at least two adjacent associated regional edge nodes; generating a cross-regional coordination instruction, the content of the cross-regional coordination instruction including a coordination target, a work range, a resource allocation rule, and a data interaction requirement; sending the cross-regional coordination instruction to the screened at least two adjacent associated regional edge nodes, so that the at least two adjacent regional edge nodes control the intelligent inspection vehicle clusters to perform coordinated work according to the cross-regional coordination instruction.

6. The intelligent patrol vehicle dynamic coordination method based on multi-modal perception fusion according to claim 1, characterized in that, based on the target data, performing global monitoring to obtain a monitoring result, including: obtaining target data of all regional edge nodes; performing global correlation analysis on the target data to obtain the correlation relationship between all sub-regions; constructing a global inspection state view based on the correlation relationship; performing level evaluation and classification on the abnormal information of the global inspection state view to obtain a global monitoring result.

7. The intelligent patrol vehicle dynamic coordination method based on multi-modal perception fusion according to claim 1, characterized in that, After the global monitoring based on the target data to obtain the monitoring result, the method further includes: dividing the abnormal events in the monitoring result into levels to distinguish between regular abnormal events and major abnormal events requiring cross-regional coordination.

8. An intelligent inspection vehicle dynamic coordination device based on multi-modal perception fusion, characterized in that, The device includes: a first receiving unit configured to receive an inspection task of a target region; a decomposition unit configured to decompose the inspection task into a plurality of sub-region tasks based on an electronic map of the target region; an allocation unit configured to dynamically allocate the plurality of sub-region tasks to regional edge nodes according to a dynamic coordination control algorithm, the target region including a plurality of regional edge nodes, each regional edge node being configured to manage an intelligent inspection vehicle cluster in a sub-region; a second receiving unit configured to receive target data of each regional edge node, the target data being generated by fusion analysis of multi-modal perception data reported by the intelligent inspection vehicle cluster managed by the regional edge node; an obtaining unit configured to perform global monitoring based on the target data to obtain a monitoring result; an issuing unit configured to, when the monitoring result shows a major abnormal event requiring cross-regional coordination, issue a cross-regional coordination instruction to at least two adjacent regional edge nodes, so that the intelligent inspection vehicle clusters managed by the at least two adjacent regional edge nodes perform coordinated work.

9. An intelligent inspection vehicle dynamic coordination device based on multi-modal perception fusion, characterized in that, The device includes: a processor, a memory, an input / output unit, and a bus; the processor is connected with the memory, the input / output unit, and the bus; the memory stores a program, and the processor invokes the program to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium has a program saved thereon, and the program performs the method according to any one of claims 1 to 7 when executed on the computer.

Images