A heterogeneous cluster robot cooperative power equipment inspection and fault positioning system

Abstract

The present application relates to the technical field of power inspection, and particularly relates to a heterogeneous cluster robot cooperative power equipment inspection and fault positioning system. The system comprises a cluster initialization module, a task decomposition module, a candidate set generation module, a subtask exchange module, a fault task processing module and a result output module. The cluster initialization module receives an inspection task list and a heterogeneous robot queue, and initializes robot state information; the task decomposition module decomposes the inspection task into a multi-level task tree using a large language model, and generates task characteristic annotations; the candidate set generation module forms a candidate execution robot set according to capability matching degrees; the subtask exchange module optimizes task allocation in a temporary alliance using a Nash equilibrium negotiation mechanism; the fault task processing module expands the task tree based on fault information and performs local task allocation; and the result output module outputs the task list of each robot and the alliance structure, and records the inspection and fault positioning conditions.

Description

Technical Field

[0001] This invention relates to the field of power inspection technology, and in particular to a heterogeneous cluster robot collaborative power equipment inspection and fault location system. Background Technology

[0002] As power systems continue to expand, the inspection tasks for transmission lines, substations, and supporting equipment are becoming increasingly complex. Current power inspections primarily rely on single types of robots, such as drones for aerial inspections or ground robots for close-range checks. Because these inspection methods are relatively limited, they struggle to accommodate the needs of complex terrain, varying altitudes, and different angles, often resulting in blind spots or missing information in the inspection results. This hinders the timely detection and diagnosis of potential faults.

[0003] On the other hand, existing task planning methods are generally too static. They typically employ preset paths or fixed task allocation patterns, lacking a mechanism for flexible adjustment based on task characteristics and the robot's real-time status. When unexpected situations arise during inspections, such as insufficient robot power, environmental interference, or the discovery of new fault points, existing systems struggle to adjust task assignments in a timely manner, often resulting in task delays or repetitive executions, thus reducing overall inspection efficiency and resource utilization. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides a heterogeneous cluster robot collaborative power equipment inspection and fault location system, which aims to improve the problems of insufficient inspection coverage and inefficient task scheduling caused by the single inspection mode and rigid task planning in the existing technology.

[0005] This invention provides the following technical solution: a heterogeneous cluster robot collaborative power equipment inspection and fault location system, comprising:

[0006] The cluster initialization module is used to receive the inspection task list and the heterogeneous robot queue, and initialize the status information of the robots, which includes at least location, power level and capability tags.

[0007] The task decomposition module is used to decompose the inspection task into a multi-level task tree using a large language model. Each sub-task in the task tree is labeled with task characteristics.

[0008] The candidate set generation module is used for each robot to calculate the capability matching degree based on its own capabilities and the characteristics of the task. When the capability matching degree is greater than zero, the robot is added to the candidate execution set of the corresponding sub-task. Multiple robots are allowed to enter the candidate set for the same sub-task at the same time.

[0009] The subtask exchange module is used to exchange subtasks among multiple candidate robots for the same subtask based on the actual execution cost, so as to achieve optimal global subtask allocation.

[0010] The fault task processing module is used for each robot to execute assigned sub-tasks and report status information in real time. When a power equipment fault is detected, a fault location task tree is generated based on the fault information, and a local candidate set is generated and tasks are exchanged on the fault location task tree.

[0011] The results output module is used to output the task list and alliance structure of each robot, and record the equipment inspection and fault location.

[0012] Furthermore, the steps to break down the inspection task into a multi-level task tree include:

[0013] The large language model parses the input inspection task, identifies the target object, inspection area and constraints of the task, and decomposes it into executable sub-tasks based on the parsing results.

[0014] In the process of hierarchical decomposition, task characteristic labels are generated for each subtask.

[0015] Furthermore, the steps for generating the candidate execution set include:

[0016] Each robot is matched with the task characteristics of the sub-task based on its capability tags, and the capability matching degree is calculated.

[0017] When the capability matching degree is greater than zero, the corresponding robot is added to the candidate set for this sub-task;

[0018] In the candidate set, multiple robots are allowed to perform capability matching simultaneously for the same subtask to form a candidate robot set for that subtask.

[0019] Furthermore, the calculation steps for the capability matching degree are as follows:

[0020] The required capabilities of the subtask are compared with the robot's capability tags one by one. If the robot has the required capability, a positive matching value is assigned; otherwise, a zero value is assigned and the corresponding robot is excluded.

[0021] Based on the comprehensive comparison results, a capability matching score is generated.

[0022] Furthermore, the steps for exchanging subtasks include:

[0023] The set of candidate robots for the same subtask is analyzed, and their performance is compared based on the benefit function of each robot.

[0024] Based on the comparison results, subtasks are swapped among candidate robots to optimize the overall subtask completion performance.

[0025] After the subtasks are swapped, update the task list and execution plan for each robot.

[0026] Furthermore, the subtask exchange takes place within a temporary alliance, the establishment of which includes the following steps:

[0027] Identify robots with overlapping functions from the set of candidate execution robots;

[0028] A temporary alliance is established among the robots, and their respective status information is shared within the alliance;

[0029] The alliance uses a Nash equilibrium negotiation mechanism to allocate subtasks, so as to form a stable allocation result among the robots.

[0030] Furthermore, the implementation steps of the Nash equilibrium negotiation mechanism include:

[0031] Each robot in the alliance is modeled as a game participant, and a payoff function is constructed for each robot. The payoff function is determined based on the difference between the value of the subtask and the execution cost.

[0032] An iterative update approach is adopted, which allows each robot to select sub-tasks that can improve its own benefits while keeping the task selections of other robots unchanged.

[0033] When no single robot within the alliance can gain a higher benefit by changing its task selection, a stable task allocation is determined within the alliance, and this allocation is used as the alliance's execution plan.

[0034] Furthermore, the steps for generating local candidate sets and exchanging tasks include:

[0035] Based on the fault information, identify the faulty equipment, and infer the potentially associated faulty areas or equipment according to the type of the faulty equipment, so as to expand the coverage of the fault location task tree.

[0036] Select a set of candidate robots from the heterogeneous robot queue that are located in or near the fault area and whose capability labels match.

[0037] Local candidate set generation and task exchange are performed within the candidate robot set to achieve rapid allocation of fault location tasks;

[0038] When the local task allocation cannot meet the task requirements, the fault location task tree is incorporated into the global candidate set generation and task exchange process.

[0039] Furthermore, the steps for reasoning about potentially related fault areas or devices include:

[0040] A causal graph is constructed based on the operating topology of power equipment, historical fault logs, and expert knowledge to describe the causal dependencies between equipment.

[0041] When a device fails, the system uses a causal graph to infer its potential associated fault links, generates fault candidate nodes, and adds them to the fault location task tree.

[0042] The present invention has the following beneficial effects:

[0043] 1. In this invention, by decomposing the power equipment inspection task into a multi-level task tree using a large language model, and combining candidate set generation, sub-task exchange mechanisms, and task allocation using a Nash equilibrium negotiation method within a temporary alliance, efficient collaboration among heterogeneous robot clusters is achieved. This approach fully considers the capability tags, execution costs, and real-time status of each robot, and can flexibly adjust the task allocation scheme, enabling the inspection task to be completed quickly and accurately in complex power environments, thereby improving inspection efficiency and resource utilization.

[0044] 2. In this invention, a causal graph is constructed based on the operating topology of power equipment, historical fault logs, and expert knowledge. Potentially related fault links are identified through causal reasoning, and potentially affected equipment or areas are incorporated into the fault location task tree, enabling dynamic expansion and rapid task allocation. This method can detect potential faults in advance, shorten fault investigation time, and, combined with the collaborative capabilities of heterogeneous robot clusters, improve the accuracy and response speed of fault handling. Attached Figure Description

[0045] Figure 1 This is a structural diagram of a heterogeneous cluster robot collaborative power equipment inspection and fault location system proposed in this invention;

[0046] Figure 2 This is a flowchart of a method for collaborative inspection and fault location of power equipment using heterogeneous cluster robots proposed in this invention. Detailed Implementation

[0047] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1

[0049] In the first embodiment of the present invention, a heterogeneous cluster robot collaborative power equipment inspection and fault location system is provided, such as... Figure 1 As shown, it includes:

[0050] The cluster initialization module is used to receive the inspection task list and the heterogeneous robot queue, and initialize the status information of the robots, which includes at least location, power level and capability tags.

[0051] Specifically, the system receives a list of inspection tasks to be inspected through the task management interface. Each task includes information such as task ID, task type, inspection area, and execution time window. The system also receives a queue of heterogeneous robots through the robot management interface. Each robot contains a unique identifier ID and capability tags, such as mobility, load capacity, sensor type, and battery status. A status record structure or database entry is created for each robot, including: location (current GPS coordinates); battery level (real-time battery percentage or estimated available energy); capability tags (including executable task type, load capacity, sensor type, and maximum speed); a task list (initialized as an empty list to store subsequently assigned inspection sub-tasks); and alliance information (initialized as an empty list for subsequent temporary alliance formation).

[0052] The task decomposition module is used to decompose the inspection task into a multi-level task tree using a large language model. Each sub-task in the task tree is labeled with task characteristics.

[0053] Furthermore, the steps to break down the inspection task into a multi-level task tree include:

[0054] The large language model parses the input inspection task, identifies the target object, inspection area and constraints of the task, and decomposes it into executable sub-tasks based on the parsing results.

[0055] In the process of hierarchical decomposition, task characteristic labels are generated for each subtask.

[0056] Specifically, the system receives a list of inspection tasks, each including a task ID, task type (e.g., line inspection, transformer inspection), inspection area, and time constraints. A pre-trained large language model is used to parse each inspection task. First, the target object is identified, such as specific power equipment types like transformers, switches, and lines. Then, the inspection area is determined, such as coordinate range or topological location. Next, constraints such as time windows, task priorities, environmental conditions, and equipment state dependencies are extracted. Finally, the parsed results are converted into machine-processable structured information, such as JSON or database tables.

[0057] Based on the analysis results, the system decomposes the task level by level. In one feasible implementation, the inspection task is decomposed into a top-level task, intermediate-level tasks, and bottom-level sub-tasks. The top-level task is the complete inspection task, serving as the root node of the task tree. The intermediate-level tasks are sets of sub-tasks divided according to the inspection area or equipment type. The bottom-level sub-tasks are specific operations that can be directly assigned to the robot for execution, such as "checking the transformer oil level" or "taking photos of the line." Each level of task node retains parent node information to track the task hierarchy and execution order.

[0058] During the hierarchical decomposition process, task characteristic labels are generated for each subtask, including at least the task type (detection, photography, maintenance assistance, etc.); execution requirements (required sensor types, tools, and robot capability requirements); time constraints (task execution time period and priority); location or area information (used for subsequent scheduling and path planning); and dependencies (preceding tasks or related tasks that must be completed simultaneously). These characteristic labels are stored in the task tree nodes for easy matching of computational capabilities and task allocation.

[0059] The task decomposition module can automatically parse complex inspection tasks into multi-level task trees and generate detailed feature annotations for each sub-task, thereby clarifying and structuring task objectives, inspection areas, and constraints.

[0060] The candidate set generation module is used for each robot to calculate the capability matching degree based on its own capabilities and the characteristics of the task. When the capability matching degree is greater than zero, the robot is added to the candidate execution set of the corresponding sub-task. Multiple robots are allowed to enter the candidate set for the same sub-task at the same time.

[0061] Furthermore, the steps for generating the candidate execution set include:

[0062] Each robot is matched with the task characteristics of the sub-task based on its capability tags, and the capability matching degree is calculated.

[0063] When the capability matching degree is greater than zero, the corresponding robot is added to the candidate set for this sub-task;

[0064] In the candidate set, multiple robots are allowed to perform capability matching simultaneously for the same subtask to form a candidate robot set for that subtask.

[0065] Furthermore, the calculation steps for the capability matching degree are as follows:

[0066] The required capabilities of the subtask are compared with the robot's capability tags one by one. If the robot has the required capability, a positive matching value is assigned; otherwise, a zero value is assigned and the corresponding robot is excluded.

[0067] Based on the comprehensive comparison results, a capability matching score is generated.

[0068] Specifically, the system first reads the task characteristics of each subtask, such as a subtask requiring the skill "visual recognition" and a battery level greater than 50%, with the working area being area A. Simultaneously, it reads the capability tags of each robot, such as robot... It possesses visual recognition capabilities, has 60% battery, and is located in Zone A. Lacking visual recognition skills, battery at 40%, located in area A; Possessing visual recognition skills, with 80% battery, and located in area B, the system compares each robot's capabilities with the task requirements, generating a capability matching score. For example, The match degree is 1. The match score is 0 (excluded). The match rate is 0.5 (the location is not in the mission area, but the skill is available).

[0069] The system adds robots with a capability matching degree greater than zero to the candidate set. For example, the candidate set for subtask T1 is... The robots in this set can participate in task allocation competition simultaneously, and the system can make an initial allocation based on task priority and actual execution cost. It is important to note that during task execution, if a robot's state changes (e.g., ...), the system will not assign tasks. When the battery level drops to 45%, the system recalculates the capability matching degree and updates the candidate set and task assignment to ensure that the task can still be completed smoothly.

[0070] With the candidate set generation module, each subtask is executed by a robot with matching capabilities. It also supports dynamic updates of candidate robots, improving the efficiency of global task completion and reducing the risk of task delays caused by changes in the state of a single robot.

[0071] The subtask exchange module is used to exchange subtasks among multiple candidate robots for the same subtask based on the actual execution cost, so as to achieve optimal global subtask allocation.

[0072] Furthermore, the steps for exchanging subtasks include:

[0073] The set of candidate robots for the same subtask is analyzed, and their performance is compared based on the benefit function of each robot.

[0074] Based on the comparison results, subtasks are swapped among candidate robots to optimize the overall subtask completion performance.

[0075] After the subtasks are swapped, update the task list and execution plan for each robot.

[0076] Furthermore, the subtask exchange takes place within a temporary alliance, the establishment of which includes the following steps:

[0077] Identify robots with overlapping functions from the set of candidate execution robots;

[0078] A temporary alliance is established among the robots, and their respective status information is shared within the alliance;

[0079] The alliance uses a Nash equilibrium negotiation mechanism to allocate subtasks, so as to form a stable allocation result among the robots.

[0080] Furthermore, the implementation steps of the Nash equilibrium negotiation mechanism include:

[0081] Each robot in the alliance is modeled as a game participant, and a payoff function is constructed for each robot. The payoff function is determined based on the difference between the value of the subtask and the execution cost.

[0082] An iterative update approach is adopted, which allows each robot to select sub-tasks that can improve its own benefits while keeping the task selections of other robots unchanged.

[0083] When no single robot within the alliance can gain a higher benefit by changing its task selection, a stable task allocation is determined within the alliance, and this allocation is used as the alliance's execution plan.

[0084] Specifically, during the subtask exchange phase, the system first exchanges the same subtask... We will analyze the set of candidate robots for a certain subtask. Each robot has a different execution cost. and task value In one feasible implementation, the system can construct a benefit function for each robot as follows:

[0085] ;

[0086] ;

[0087] ;

[0088] in, , and These represent the urgency of the subtask, the importance of the task, and the suitability of the task, respectively. , and These represent the time required to execute the subtask, the energy required to execute the task, and the distance between the robot and the task execution location, respectively. All these parameters are processed using a dimensionless method and uniformly linearly mapped to the interval [0,10] for weighted calculation. and These represent the weighting coefficients for each parameter. It should be noted that the above profit function calculation scheme is only one implementation method; other feasible schemes for calculating the profit function are also within the scope of protection of this invention. By comparing the profit functions, the system can initially determine the optimal allocation scheme of subtasks among candidate robots, thereby performing necessary subtask exchanges to optimize the overall task completion effect. For example, if... The profit is 8. It is 10. If the value is 7, then prioritize... Assigned to .

[0089] For robots with overlapping functions, the system identifies robots that can form temporary alliances from the candidate execution set and shares their respective state information, such as location, battery level, and capability tags, within the alliance. After the alliance is formed, each robot uses a Nash equilibrium negotiation mechanism to allocate sub-tasks. The system treats each robot in the alliance as a game participant, and each robot constructs a payoff function based on its capability tags and sub-task characteristics. The process involves iterative updates. During iteration, each robot selects a subtask that improves its own reward while keeping the task choices of other robots constant. If no single robot can gain a higher reward by changing its task choice, the alliance is considered to have reached a stable Nash equilibrium.

[0090] For example, there are three robots in the alliance. With three sub-tasks Initial allocation is , , The corresponding return is , , After system iteration, it was found that if Select the task to perform This can increase the profit to 9, and If the revenue remains unchanged, then the allocation will be updated to... , , After several iterations, none of the robots were able to unilaterally increase their profits. The system then used this allocation as an internal execution plan within the alliance and updated it to the task list of each robot.

[0091] Through the subtask exchange module, the system can achieve dynamic subtask exchange and stable task allocation within the alliance, taking into account robot capabilities, task value, and execution costs, thereby optimizing the overall task performance of heterogeneous cluster robots in collaborative execution.

[0092] The fault task processing module is used for each robot to execute assigned sub-tasks and report status information in real time. When a power equipment fault is detected, a fault location task tree is generated based on the fault information, and a local candidate set is generated and tasks are exchanged on the fault location task tree.

[0093] Furthermore, the steps for generating local candidate sets and exchanging tasks include:

[0094] Based on the fault information, identify the faulty equipment, and infer the potentially associated faulty areas or equipment according to the type of the faulty equipment, so as to expand the coverage of the fault location task tree.

[0095] Select a set of candidate robots from the heterogeneous robot queue that are located in or near the fault area and whose capability labels match.

[0096] Local candidate set generation and task exchange are performed within the candidate robot set to achieve rapid allocation of fault location tasks;

[0097] When the local task allocation cannot meet the task requirements, the fault location task tree is incorporated into the global candidate set generation and task exchange process.

[0098] Specifically, when a robot discovers a power equipment fault during an inspection, the system first determines the faulty equipment's number, type, and location based on the fault report information. Then, the system invokes the fault reasoning module to infer potentially affected related fault areas or equipment based on the equipment type and its location in the power grid topology, combined with pre-built causal dependencies. For example, when a transformer fails, the system can infer that its upstream switches, downstream distribution lines, and related protection equipment may be simultaneously affected, thus expanding these devices into nodes in the fault location task tree.

[0099] Based on the expanded task tree, the system needs to select a suitable robot to execute each faulty subtask. To this end, robots meeting the following two conditions are selected from the current heterogeneous robot queue: the robot is located in or near the faulty equipment area (e.g., less than 500 meters away); and the robot's capability tags match the task characteristics (e.g., a UAV with infrared temperature measurement is suitable for equipment overheating detection, and a ground inspection vehicle with a high-definition camera is suitable for visual inspection). Robots meeting these conditions are included in the candidate set for local tasks. Each candidate robot calculates its local reward function based on parameters such as its remaining battery power, current position, and estimated task execution time, and submits a candidate application. The implementation of the local reward function is consistent with that in the subtask exchange module. Within the candidate set, the system first compares the results of the reward functions and assigns the subtask to the optimal robot. When multiple feasible candidates exist for a subtask, a task exchange is allowed: the overall execution cost between robots is compared; if the exchange reduces the total cost within the alliance, the exchange is performed and the task list is updated. For example, if robot A and robot B can simultaneously complete two tasks, "checking switch status" and "infrared temperature measurement", and if the path length of A is reduced by 30% and the execution time of B is reduced by 20% after the swap, the system will perform the swap and write the updated task allocation into the execution plan.

[0100] If, after the local candidate and task exchange, some critical tasks still lack suitable robots to perform (e.g., the drone's battery is insufficient to cover the upstream line), the system will integrate the entire fault location task tree into the global candidate set generation and exchange process to ensure that the task can eventually be completed.

[0101] Through the above process, the system can quickly organize local robot collaboration after a fault is detected, shorten task scheduling time, and automatically switch to global resource allocation when local resources are insufficient, ensuring the timeliness and robustness of fault location.

[0102] Furthermore, the steps for reasoning about potentially related fault areas or devices include:

[0103] A causal graph is constructed based on the operating topology of power equipment, historical fault logs, and expert knowledge to describe the causal dependencies between equipment.

[0104] When a device fails, the system uses a causal graph to infer its potential associated fault links, generates fault candidate nodes, and adds them to the fault location task tree.

[0105] Specifically, based on the operating topology of the power system, the physical connections and operational dependencies between devices are obtained. Secondly, historical fault logs are collected, and statistical analysis is performed on common fault modes and their propagation paths. Simultaneously, combined with the experience and rules of operation and maintenance experts, the causal relationships between key devices are supplemented and corrected. Based on this, a causal graph is constructed to describe the causal dependencies between devices.

[0106] When a device malfunctions, the system can retrieve nodes with direct or indirect causal relationships to that device from the causal graph and infer possible fault propagation links. The resulting potentially related devices are added as candidate nodes to the fault location task tree for further processing by the subsequent candidate set generation and task exchange modules.

[0107] Through the construction and reasoning mechanism of causal graphs, the system can quickly and interpretably locate potentially related fault areas and devices after the initial fault is detected, minimize the scope of ineffective investigation, and prioritize the assignment of high-value location tasks to appropriate robots, thereby improving the timeliness of fault handling.

[0108] The results output module is used to output the task list and alliance structure of each robot, and record the equipment inspection and fault location.

[0109] Specifically, the system summarizes the sub-task numbers, execution order, time information, and inspection results of each robot into a task list, and records the alliance members and their division of labor into an alliance structure. At the same time, the equipment status and fault location results during the inspection process are stored in the database or log file for subsequent retrieval or analysis. Finally, the results can be provided to maintenance personnel through interface display or report export, thereby achieving complete traceability and visualization of the inspection and fault handling process.

[0110] Example 2

[0111] Currently, in power equipment inspection, single-type robots face problems such as uneven task coverage, rigid scheduling, and insufficient resource utilization, resulting in limited inspection efficiency and fault response speed. To address these issues, this invention provides a heterogeneous cluster robot collaborative power equipment inspection and fault location system. The system's implementation process is as follows: Figure 2 As shown. The specific implementation process of this system is as follows:

[0112] The system is applied to power equipment inspection scenarios involving land and air collaboration. The cluster initialization module first receives the power line corridor inspection task and the robot queue, which includes multiple quadruped inspection robots and several drones. It then initializes their position, power level, and capability tags. For example, the drones have aerial photography and infrared detection capabilities, while the ground robots have proximity image acquisition and local contact detection capabilities.

[0113] Subsequently, the task decomposition module parses the inspection task into a multi-level task tree. For example, the task of inspecting the top components of a high-voltage transmission tower is assigned to a drone, while the inspection of ground foundation components and cable connection points is assigned to a ground robot. Each sub-task is accompanied by task characteristic labels, such as "high-altitude image acquisition" and "contact temperature measurement".

[0114] During the candidate set generation process, UAVs and ground robots calculate their capability matching degree based on their own capability tags and are added to the candidate set. When multiple UAVs are simultaneously matched for the high-altitude inspection sub-task, the sub-task exchange module analyzes their battery level and flight distance to form a reasonable allocation, ensuring overall inspection efficiency.

[0115] In actual operation, if a drone discovers a suspected insulator overheating anomaly during inspection, the fault task processing module will generate a fault location task tree for that device. The system prioritizes local task capability matching and task exchange within the nearest set of ground robots, assigning the contact temperature measurement sub-task to the nearest quadruped robot. When this set is insufficient to complete all location tasks, the system incorporates the task tree into the global candidate set generation process, allowing more robots to participate in the collaborative task allocation.

[0116] Finally, the results output module summarizes and outputs the inspection task lists of the UAV and ground robot, the formed land-air collaborative alliance structure, and the location information of abnormal faults, providing maintenance personnel with visualized inspection results and fault reports.

[0117] Compared to existing technologies that rely solely on drones for power equipment inspection, this embodiment achieves refined task allocation and multimodal verification of fault location through heterogeneous collaboration between drones and ground robots. Experimental results show that, under the same task scale, this system improves inspection task coverage by approximately 13% and reduces inspection time by approximately 17%, effectively solving the problem of blind spots in traditional methods.

[0118] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A heterogeneous cluster robot collaborative power equipment inspection and fault location system, characterized in that, include: The cluster initialization module is used to receive the inspection task list and the heterogeneous robot queue, and initialize the status information of the robots, which includes at least location, power level and capability tags. The task decomposition module is used to decompose the inspection task into a multi-level task tree using a large language model. Each sub-task in the task tree is labeled with task characteristics. The candidate set generation module is used for each robot to calculate the capability matching degree based on its own capabilities and the characteristics of the task. When the capability matching degree is greater than zero, the robot is added to the candidate execution set of the corresponding sub-task. Multiple robots are allowed to enter the candidate set for the same sub-task at the same time. The subtask exchange module is used to exchange subtasks among multiple candidate robots for the same subtask based on the actual execution cost, so as to achieve optimal global subtask allocation. The fault task processing module is used for each robot to execute assigned sub-tasks and report status information in real time. When a power equipment fault is detected, a fault location task tree is generated based on the fault information, and a local candidate set is generated and tasks are exchanged on the fault location task tree. The results output module is used to output the task list and alliance structure of each robot, and record the equipment inspection and fault location. The subtask exchange takes place within a temporary alliance, and the steps for establishing the temporary alliance include: Identify robots with overlapping functions from the set of candidate execution robots; A temporary alliance is established among the robots, and their respective status information is shared within the alliance; The alliance uses a Nash equilibrium negotiation mechanism to allocate subtasks, so as to form a stable allocation result among the robots. The implementation steps of the Nash equilibrium negotiation mechanism include: Each robot in the alliance is modeled as a game participant, and a payoff function is constructed for each robot. The payoff function is determined based on the difference between the value of the subtask and the execution cost. An iterative update approach is adopted, which allows each robot to select sub-tasks that can improve its own benefits while keeping the task selections of other robots unchanged. When no single robot within the alliance can gain a higher benefit by changing its task selection, a stable task allocation is determined within the alliance, and this allocation is used as the alliance's execution plan.

2. The heterogeneous cluster robot collaborative power equipment inspection and fault location system according to claim 1, characterized in that, The steps to break down inspection tasks into a multi-level task tree include: The large language model parses the input inspection task, identifies the target object, inspection area and constraints of the task, and decomposes it into executable sub-tasks based on the parsing results. In the process of hierarchical decomposition, task characteristic labels are generated for each subtask.

3. The heterogeneous cluster robot collaborative power equipment inspection and fault location system according to claim 1, characterized in that, The steps for generating a candidate execution set include: Each robot is matched with the task characteristics of the sub-task based on its capability tags, and the capability matching degree is calculated. When the capability matching degree is greater than zero, the corresponding robot is added to the candidate set for this sub-task; In the candidate set, multiple robots are allowed to perform capability matching simultaneously for the same subtask to form a candidate robot set for that subtask.

4. The heterogeneous cluster robot collaborative power equipment inspection and fault location system according to claim 3, characterized in that, The calculation steps for the capability matching degree are as follows: The required capabilities of the subtask are compared with the robot's capability tags one by one. If the robot has the required capability, a positive matching value is assigned; otherwise, a zero value is assigned and the corresponding robot is excluded. Based on the comprehensive comparison results, a capability matching score is generated.

5. A heterogeneous cluster robot collaborative power equipment inspection and fault location system according to claim 1, characterized in that, The steps for performing subtask swapping include: The set of candidate robots for the same subtask is analyzed, and their performance is compared based on the benefit function of each robot. Based on the comparison results, subtasks are swapped among candidate robots to optimize the overall subtask completion performance. After the subtasks are swapped, update the task list and execution plan for each robot.

6. The heterogeneous cluster robot collaborative power equipment inspection and fault location system according to claim 1, characterized in that, The steps for generating local candidate sets and exchanging tasks include: Based on the fault information, identify the faulty equipment, and infer the potentially associated faulty areas or equipment according to the type of the faulty equipment, so as to expand the coverage of the fault location task tree. Select a set of candidate robots from the heterogeneous robot queue that are located in or near the fault area and whose capability labels match. Local candidate set generation and task exchange are performed within the candidate robot set to achieve rapid allocation of fault location tasks; When local task allocation cannot meet task requirements, the fault location task tree is incorporated into the global candidate set generation and task exchange process.

7. A heterogeneous cluster robot collaborative power equipment inspection and fault location system according to claim 6, characterized in that, The steps for reasoning about potentially related fault areas or devices include: A causal graph is constructed based on the operating topology of power equipment, historical fault logs, and expert knowledge to describe the causal dependencies between equipment. When a device fails, the system uses a causal graph to infer its potential associated fault links, generates fault candidate nodes, and adds them to the fault location task tree.

Images