Oil and gas field multi-robot cooperative operation method and system based on large language model

Abstract

This invention discloses a method and system for multi-robot collaborative operation in oil and gas fields based on a large language model, relating to the fields of robot collaboration and intelligent planning. The method acquires oil and gas field operation task information, pre-sets task construction rules, uses a large model to parse task instructions and form a hierarchical task tree, and then writes oil and gas field operation constraint information to form a constrained hierarchical task tree. Simultaneously, it acquires robot capability information and constructs a skill library, matches and selects execution robots, and thus determines the multi-robot collaborative operation arrangement to complete the oil and gas field operation task. The system includes a task acquisition module, a task rule pre-setting module, a task tree generation module, a constraint writing module, an execution subject determination module, and a collaborative execution module. This invention, through the collaborative design of the method and system, enhances the understanding and collaborative stability of multi-robot systems for complex tasks, improving their collaborative efficiency and task completion quality in oil and gas field scenarios.

Description

Technical Field

[0001] This invention relates to the field of robot collaboration and intelligent planning technology, specifically to a method and system for multi-robot collaborative operation in oil and gas fields based on a large language model. Background Technology

[0002] Oil and gas field operations require tasks such as inspection, equipment status monitoring, and anomaly identification. Multi-robot collaborative operation technology, by organizing multiple robots to work together within the same work area, can reduce the frequency of personnel entering high-risk areas and improve operational efficiency. With the development of large language modeling technologies, using large language models to semantically understand task instructions in natural language form and generate structured task representations under preset task construction rules has become a viable technological path for driving multi-robot collaborative operations. Hierarchical task trees can be used to describe the parent-child relationships between task nodes and the logical constraints between them. Furthermore, combining oil and gas field operational constraints with robot capability information can form collaborative execution schemes, enabling the division of labor and collaborative execution of complex tasks by multiple robots.

[0003] Existing multi-robot collaborative operation systems typically employ rule templates, script processes, or task planning models to achieve task modeling, task decomposition, and task allocation. Most related solutions use pre-structured work orders or standardized processes as input, manually configuring task templates, task dependencies, and execution order, and then scheduling based on robot capabilities, resource status, or task priority. Some solutions introduce natural language understanding modules to convert natural language commands into structured instructions, operation steps, or task lists, and the scheduling module then assigns tasks to different robots for execution; during execution, the system updates task status or handles anomalies based on status feedback. These solutions can achieve multi-robot collaborative operation under specific task types or fixed process conditions.

[0004] However, traditional multi-robot collaborative operation systems still have shortcomings in applications with strong constraints and complex environments such as oil and gas fields. Existing solutions mostly rely on manually configured task templates or fixed processes, resulting in low adaptability when faced with changes in task type, work area, and parameters. For natural language task instructions, the task representations generated by existing methods lack unified structured constraints and reusable task substructures, leading to insufficient consistency and controllability of task decomposition results. Furthermore, it is difficult to clearly express logical constraints such as sequential execution, concurrent execution, and selective execution within the same task representation, and to effectively link oil and gas field operation constraints to the task allocation and execution process, resulting in insufficient executability of collaborative execution schemes or the need for manual correction. Therefore, there is an urgent need for a multi-robot collaborative operation technology solution that can generate hierarchical task trees using a large language model under rule constraints, and combine oil and gas field operation constraints with robot capability information to form an executable collaborative scheme.

[0005] Reference 1 ([1] Chengdu University of Technology. A port air-ground collaborative intelligent autonomous inspection robot system and method: 202510952502.2[P]. 2025-10-17.) proposes an air-ground collaborative inspection scheme in port scenarios, which integrates localization mapping and target detection, and introduces a large model for task planning to achieve autonomous inspection. However, this scheme is mainly aimed at port inspection and does not fully consider the complex constraints such as explosion-proof level, restricted area, time window and communication availability of oil and gas field operations. It also lacks structured prompt words and structured task tree output constraints to ensure the controllable decomposition and executability of heterogeneous multi-robot collaborative operations. Summary of the Invention

[0006] To address the problems in existing technologies for multi-robot collaborative operations in oil and gas fields under strong constraints, such as reliance on manual configuration for task modeling, insufficient consistency and controllability in the conversion of natural language task instructions into executable task representations, difficulty in integrating operational constraints into the task organization process, and lack of unified support for determining the execution subject, this invention proposes a method and system for multi-robot collaborative operations in oil and gas fields based on a large language model. By pre-setting task construction rules and combining them with a reusable task substructure library, a hierarchical task tree is formed using a large language model. Oil and gas field operational constraint information is then written into task nodes and the sequential content between nodes. Finally, the execution subject is determined for basic task nodes using a robot skill library, thereby achieving standardized modeling and collaborative implementation of multi-robot collaborative operations in oil and gas fields.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: On the one hand, this invention provides a method for multi-robot collaborative operation in oil and gas fields based on a large language model, comprising the following steps: S1. Obtain oil and gas field operation task information, which includes task instructions in natural language form and oil and gas field operation constraint information. Furthermore, the oil and gas field operation constraints include explosion-proof level constraints, restricted area constraints, high-risk area constraints, operation priority constraints, operation window constraints, and communication availability constraints.

[0008] S2. Pre-set task construction rules for generating a hierarchical task tree, the task construction rules including a reusable task substructure library, task node connection rules and structured representation rules for the hierarchical task tree; Furthermore, the reusable task substructure library includes task substructures for typical oil and gas field operations. The task substructures include inspection substructures, leak detection substructures, equipment status acquisition substructures, and emergency response substructures. Each task substructure is equipped with operation parameter items to adapt to specific operation scenarios. The operation parameter items include operation area, operation object, detection index threshold, collection point set, and operation time window parameters.

[0009] Furthermore, the hierarchical task tree includes task nodes organized in a parent-child hierarchy. The task nodes include root task nodes, non-leaf task nodes, and leaf task nodes. The leaf task nodes are basic task nodes that can be executed by the robot.

[0010] S3. Based on the large language model, perform semantic parsing on the task instructions, and under the constraints of the task construction rules, call the corresponding task substructure from the reusable task substructure library to organize and hierarchically expand the task nodes to form a hierarchical task tree. Furthermore, the task construction rules include node generation content for generating task nodes, hierarchy setting content for determining the hierarchical affiliation of task nodes, continuation setting content for limiting the continuation method of task nodes, and representation content for limiting the output fields and format requirements of the task tree. The large language model determines the task nodes to be included in the hierarchical task tree according to the task instructions, and completes the organization of task nodes according to the node generation content, hierarchy setting content and continuation setting content, and outputs the hierarchical task tree according to the representation content.

[0011] S4. Write the oil and gas field operation constraint information into the task nodes and the connection content between nodes to form a constrained hierarchical task tree. Furthermore, the oil and gas field operation constraint information is written into the task node or the connection content between nodes, including: limiting the access conditions of the operation area corresponding to the task node, limiting the safety conditions for the task node to be activated, limiting the collaborative connection requirements between related task nodes, and limiting the operation time window corresponding to the task node.

[0012] S5. Obtain the capability information of each robot in the multi-robot system and build a robot skill library; Furthermore, the robot skill library includes information on each robot's operational capabilities, operational area adaptability, and operational condition adaptability.

[0013] S6. Based on the task requirements of the basic task nodes, candidate execution robots are matched from the robot skill library, and the matching results are filtered in combination with the oil and gas field operation constraint information to determine the execution robot for each basic task node. Furthermore, the determination of the execution subject for the basic task node includes: matching the operational capability information, operational area adaptability information, and operational condition adaptability information of the candidate execution robots with the execution requirements of the basic task node based on the task requirements of the basic task node, and filtering the matching results in conjunction with the oil and gas field operation constraint information, thereby determining the execution robot for the basic task node.

[0014] S7. Determine the multi-robot collaborative operation arrangement based on the constrained hierarchical task tree, and complete the oil and gas field operation task accordingly.

[0015] On the other hand, the present invention also provides a multi-robot collaborative operation system for oil and gas fields based on a large language model, comprising: The task acquisition module is configured to receive oil and gas field operation task information, which includes task instructions in natural language and oil and gas field operation constraint information. The task rule preset module is configured to preset task construction rules for generating a hierarchical task tree. The task construction rules include a reusable task substructure library, task node connection rules, and structured representation rules for the hierarchical task tree. The task tree generation module is configured to perform semantic parsing of the task instructions based on a large language model, and under the constraints of the task construction rules, call the corresponding task substructure from the reusable task substructure library to organize and hierarchically expand the task nodes to form a hierarchical task tree. The constraint writing module is configured to write the oil and gas field operation constraint information into the task node or the continuation content between nodes to form a constrained hierarchical task tree. The execution entity determination module is configured to acquire the capability information of each robot in the multi-robot system and construct a robot skill library, wherein the robot skill library includes the operational capability information, operational area adaptability information, and operational condition adaptability information of each robot; and, based on the task requirements of the basic task node, match the operational capability information, operational area adaptability information, and operational condition adaptability information of the candidate execution robots with the execution requirements of the basic task node, and filter the matching results in conjunction with the oil and gas field operation constraint information to determine the execution robot of the basic task node; The collaborative execution module is configured to determine the multi-robot collaborative operation arrangement based on the constrained hierarchical task tree, and complete the oil and gas field operation task accordingly.

[0016] Compared with existing technologies, the above-mentioned method and system for multi-robot collaborative operation in oil and gas fields based on a large language model has the following advantages: 1. This invention utilizes pre-defined task construction rules and a reusable task substructure library to perform semantic parsing of task instructions using a large language model, forming a hierarchical task tree. Since the task tree generation process is constrained by the task construction rules, and the reusable task substructure library provides a reusable structural foundation for task decomposition, it reduces the reliance on manually configured task templates in existing technologies and improves the consistency, standardization, and controllability of converting natural language task instructions into executable job representations.

[0017] 2. This invention incorporates oil and gas field operational constraints into task nodes and the connections between them, enabling constraints such as restricted areas, explosion-proof ratings, operational windows, and safety conditions to be embedded in a hierarchical task tree during the task organization phase. Since operational constraints are not added and modified later in the task execution phase, but are integrated with task nodes and their connections immediately after the task tree is formed, the adaptability of the task organization results to highly constrained oil and gas field scenarios is enhanced, reducing subsequent manual adjustments and improving the rationality of operational scheduling.

[0018] 3. This invention acquires the capability information of each robot in a multi-robot system and constructs a robot skill library. It then matches the task requirements of basic task nodes with the operational capability information, operational area adaptability information, and operational condition adaptability information in the robot skill library, and further filters the matching results by incorporating oil and gas field operational constraints. Because the determination of the executor considers task requirements, robot capabilities, and operational constraints simultaneously, it can more accurately select suitable executor robots for each basic task node, improving the relevance and rationality of the executor determination. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0020] Figure 2 This is a schematic diagram of oil and gas field operations according to an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of a hierarchical task tree structure.

[0022] Figure 4 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0023] To further describe the multi-robot collaborative operation method for oil and gas fields based on a large language model proposed in this invention, an embodiment of the invention is described in conjunction with the accompanying drawings. This embodiment is merely a further explanation of the invention; those skilled in the art can make non-essential improvements and adjustments based on this invention, which should also fall within the scope of protection of this invention.

[0024] This embodiment uses "oil and gas pipeline inspection" as an example: the pipeline P1 in area A is inspected from the starting point S1 to the ending point S2. If any abnormality is found, it is located, marked and reported. During the operation, the restricted area R1 must be avoided.

[0025] like Figure 1 The diagram shown is a flowchart illustrating the multi-robot collaborative operation method for oil and gas fields based on a large language model, according to the present invention. Specifically, it includes the following steps: S1. Obtain oil and gas field operation task information, which includes task instructions in natural language form and oil and gas field operation constraint information.

[0026] In this embodiment, the natural language task instruction is, for example, (to inspect and collect data at points V3 and W2 of pipeline P1, identify and report anomalies, and avoid the restricted area R1). Oil and gas field operation constraints include: inspection point set (V3, W2), restricted area constraints (R1), data acquisition type (image / video / sensor readings), anomaly identification and judgment rules, and reporting methods.

[0027] like Figure 2 The diagram shown is a schematic of an oil and gas pipeline inspection operation scenario in an embodiment of the present invention. Pipeline P1 is located in area A, extending from starting point S1 to ending point S2. Inspection points V3 and W2 are set along the pipeline. During the operation, the restricted area R1 must be avoided. Robot A follows the planned path to bypass the restricted area R1 and proceeds to the corresponding inspection point to perform inspection data collection, anomaly identification, and result reporting tasks. The collaborative operation system is used to parse the tasks and complete the corresponding operation arrangements.

[0028] S2. Pre-set task construction rules for generating hierarchical task trees, the task construction rules including a reusable task substructure library, task node connection rules and structured representation rules for hierarchical task trees.

[0029] S21. Define the representation rules for the hierarchical task tree. In one implementation, the hierarchical task tree includes task nodes organized in a parent-child hierarchy. The task nodes include a root task node, non-leaf task nodes, and leaf task nodes, where the leaf task nodes are basic task nodes that can be executed by the robot. The representation rules for the hierarchical task tree are used to define the fields and format requirements used when outputting the task tree.

[0030] In one implementation, the task node contains at least the following fields: 1) Node identifier id; Node type type (root / non_leaf / leaf); Parent node identifier parent_id; Task name name; 2) Task parameter items (params, such as location number, path segment, collection threshold, etc.); 3) Task requirements (e.g., skill tags for arrival navigation, image acquisition, anomaly detection, and reporting communication); 4) Constraints (e.g., "Avoid R1"); 5) Execution status (not started / in progress / completed / failed); S22. Establish a reusable task substructure library, which contains task substructure templates corresponding to oil and gas pipeline inspection operations. In this embodiment, the inspection task substructure may include the following node sequence (arrival at point, data collection, anomaly identification, result generation and reporting), wherein: the point set is used as a parameterized slot; the data collection type, threshold, and reporting object are used as configurable parameter fields; and the restricted area R1 is used as a constraint field bound to the relevant nodes or path planning parameters.

[0031] S23. Establish task node connection rules. The task node connection rules are used to determine the organization and succession of each task node in the hierarchical task tree after the task substructure is invoked.

[0032] In this embodiment, a hierarchical task tree can be formed around the inspection task of pipeline P1, and corresponding lower-level task nodes can be generated according to points V3 and W2. Then, basic task nodes such as arrival at the point, inspection data collection, anomaly identification, and result reporting can be organized under each point task node. The restricted area R1 can be written as oil and gas field operation constraint information into the constraint items of relevant task nodes and the continuation content between nodes, which is used to limit the access conditions of the corresponding task node's operation area and the avoidance requirements during the execution process.

[0033] S3. Based on the large language model, perform semantic parsing on the task instructions, and under the constraints of the task construction rules, call the corresponding task substructure from the reusable task substructure library to organize and hierarchically expand the task nodes to form a hierarchical task tree.

[0034] S31. Semantic parsing of the task instruction. In this embodiment, the large language model performs semantic parsing on the natural language task instruction "Inspect and collect data at points V3 and W2 of pipeline P1, identify and report anomalies, and avoid the restricted area R1". It identifies the work object as pipeline P1, the work points as V3 and W2, the work content as inspection and data collection, anomaly identification and result reporting, and the work constraint as avoiding the restricted area R1.

[0035] Based on the semantic parsing results, the task content to be included in the hierarchical task tree and the corresponding basic execution requirements can be further determined.

[0036] S32. Call the corresponding task substructure and organize the task nodes. In this embodiment, based on the semantic parsing result, the task substructure corresponding to the inspection operation is called from the reusable task substructure library, and the task nodes are organized in conjunction with the point set {V3, W2}; Specifically, a root task node can be formed around "pipeline P1 inspection operation", and corresponding subordinate task nodes can be formed around points V3 and W2 respectively. Under each point task node, basic task nodes such as arrival at the point, inspection data collection, anomaly identification and result reporting can be further organized to form a hierarchical task tree organized according to the parent-child hierarchy.

[0037] In one implementation, the task construction rules include node generation content for generating task nodes, hierarchy setting content for determining the hierarchical affiliation of task nodes, continuation setting content for limiting the continuation method of task nodes, and representation content for limiting the requirements of task tree output fields; the large language model determines the task nodes to be included in the hierarchical task tree according to the task instructions, and completes the organization of task nodes according to the node generation content, hierarchy setting content and continuation setting content.

[0038] S33. Generate a hierarchical task tree output. In one implementation, the output of the hierarchical task tree can be represented in a fielded data format, for example, as the following data fragment: { "nodes": [ {"id":"T0","type":"ROOT","name":"Pipeline P1 Inspection Operation"}, {"id":"T1","type":"NON_LEAF","parent_id":"T0","name":"V3 Point Inspection Sub-task"} {"id":"T2","type":"NON_LEAF","parent_id":"T0","name":"W2 location inspection sub-task"}, {"id":"V3_Arrive","type":"LEAF","parent_id":"T1","name":"Arrived at valve chamber V3","req":["Navigation to arrive"],"params":{"target":"V3"},"constraints":["Avoid R1"]}, {"id":"V3_Acq","type":"LEAF","parent_id":"T1","name":"V3 point acquisition","req":["image acquisition"],"params":{"target":"V3"}}, {"id":"W2_Arrive","type":"LEAF","parent_id":"T2","name":"Arrived at wellhead W2","req":["Navigation arrived"],"params":{"target":"W2"},"constraints":["Avoid R1"]}, {"id":"W2_Acq","type":"LEAF","parent_id":"T2","name":"W2 Location Acquisition","req":["Image Acquisition"],"params":{"target":"W2"}} ] }

[0039] S4. Write the oil and gas field operation constraint information into the task nodes and the connection content between nodes to form a constrained hierarchical task tree.

[0040] In this embodiment, the oil and gas field operation constraint information includes at least the restricted area R1, operation area access requirements, operation safety requirements, and operation time window requirements. The oil and gas field operation constraint information can be written into the constraint items of the corresponding task node, or into the continuation content between nodes, to limit the entry conditions, start conditions, continuation conditions, and operation time periods of the relevant task nodes.

[0041] Specifically, the restricted area R1 can be written into the constraints of the "reaching point V3" node and the "reaching point W2" node to limit the access conditions of the corresponding task node's work area. At the same time, the restricted area R1 can also be written into the node continuation content related to the movement implementation to limit the avoidance requirements of the robot when performing the task of reaching the point, so as to ensure that the robot does not enter the restricted area R1 during the movement.

[0042] In this embodiment, the basic task nodes formed around points V3 and W2 can be organized sequentially according to a preset node succession method. For example, under the same point sub-task, after the "arrive at point" node is completed, the subsequent basic task nodes such as "inspection and collection", "anomaly identification" and "result reporting" are implemented sequentially according to the succession content between nodes. For the lower-level task nodes corresponding to different points, the corresponding node succession content can be written according to the actual resource conditions and constraints to form a constrained hierarchical task tree.

[0043] For ease of understanding, such as Figure 3As shown, the root task node T0 represents the pipeline P1 inspection operation. Under the root task node T0 are subordinate task nodes T1 and T2, representing the V3 and W2 point inspection sub-tasks respectively. Under T1 and T2, an "Arrival Point" node is first organized, and then further organized into basic task nodes such as "Inspection Data Collection," "Anomaly Identification," and "Result Reporting." The operational constraint "Avoid Restricted Area R1" can be written into the constraint items of the corresponding arrival point node and the continuation content between related nodes, used to limit the access conditions of the corresponding task node's work area and the avoidance requirements during execution.

[0044] S5. Obtain the capability information of each robot in the multi-robot system and build a robot skill library.

[0045] In one implementation, the capability information includes at least the robot's operational capability information, operational area adaptability information, and operational condition adaptability information; Among them, the operational capability information can be used to characterize the basic types of tasks that the robot can perform, such as reaching the location, inspection and data collection, anomaly identification, and result reporting; the operational area adaptability information can be used to characterize the robot's adaptability to the operational area environment; and the operational condition adaptability information can be used to characterize the robot's adaptability to safety conditions, communication conditions, or other operational constraints.

[0046] In this embodiment, the system constructs a robot skill library based on the capability information of each robot in the multi-robot system, and writes the capability information of different robots into the robot skill library in a corresponding relationship, so that it can be matched and filtered according to the execution requirements of the basic task nodes in the future.

[0047] S6. Based on the task requirements of the basic task nodes, candidate execution robots are matched from the robot skill library, and the matching results are filtered in combination with the oil and gas field operation constraint information to determine the execution robot for each basic task node.

[0048] S61. Demand-capability matching: For each leaf node, select robots that meet its req from the candidate set; S62. Constraint Screening: Based on operational constraints such as "avoiding restricted area R1", further screen candidate robots and eliminate candidates that cannot avoid R1 in path planning or cannot complete the task as required. S63. Determine the execution robot. After completing the matching of execution requirements and constraint screening, determine the execution robot for each basic task node.

[0049] For example, in this embodiment, the corresponding execution robots that undertake basic task nodes such as "arriving at point V3", "inspection and data collection at point V3", "identification of anomalies at point V3", "reporting of results at point V3", and "arriving at point W2", "inspection and data collection at point W2", "identification of anomalies at point W2", and "reporting of results at point W2" can be determined respectively, so as to provide the basis for the execution subject for subsequent multi-robot collaborative operation arrangements.

[0050] S7. Determine the multi-robot collaborative operation arrangement based on the constrained hierarchical task tree, and complete the oil and gas field operation task accordingly.

[0051] In this embodiment, after determining the execution robots for each basic task node, the system arranges collaborative operations for each basic task node according to a constrained hierarchical task tree. The collaborative operation arrangement may include the execution robot corresponding to each basic task node, the tasks undertaken by each execution robot, and the implementation order of the relevant basic task nodes.

[0052] For example, for the basic task nodes corresponding to points V3 and W2, the corresponding work arrangements can be determined according to the execution robots of each basic task node, so that the corresponding robots can undertake basic tasks such as "arriving at the point", "inspection and collection", "anomaly identification" and "result reporting", and complete the corresponding tasks under the constraints of the hierarchical task tree.

[0053] In this embodiment, once the basic task nodes such as "reaching point V3" or "reaching point W2" meet the corresponding implementation conditions, the corresponding execution robot performs subsequent basic task nodes such as inspection and data collection, anomaly identification, and result reporting according to the determined collaborative operation plan, thereby completing the inspection task of pipeline P1 at points V3 and W2. Finally, the system summarizes the operation results corresponding to each basic task node to complete the oil and gas field operation task.

[0054] like Figure 4 As shown, this invention also provides a multi-robot collaborative operation system for oil and gas fields based on a large language model, including a task acquisition module, a task rule presetting module, a task tree generation module, a constraint writing module, an execution subject determination module, and a collaborative execution module. The implementation process of each module of the system below corresponds to steps S1 to S7 of the aforementioned method embodiment. The way the system solves problems is similar to the above method. For specific implementation, please refer to the aforementioned method embodiment. Repeated parts will not be described again.

[0055] At the system level, the task acquisition module is used to receive oil and gas field operation task information, which includes task instructions in natural language form and oil and gas field operation constraint information. The task rule preset module is used to preset task construction rules for generating a hierarchical task tree. The task construction rules include a reusable task substructure library, task node connection rules, and structured representation rules for the hierarchical task tree. The task tree generation module is used to perform semantic parsing of the task instructions based on the large language model, and under the constraints of the task construction rules, call the corresponding task substructure from the reusable task substructure library to organize and hierarchically expand the task nodes to form a hierarchical task tree. The constraint writing module is used to write the oil and gas field operation constraint information into the task node or the continuation content between nodes to form a constrained hierarchical task tree. The execution entity determination module is used to acquire the capability information of each robot in the multi-robot system and construct a robot skill library. The robot skill library includes the operational capability information, operational area adaptability information, and operational condition adaptability information of each robot. Based on the task requirements of the basic task node, the module matches the operational capability information, operational area adaptability information, and operational condition adaptability information of the candidate execution robots with the execution requirements of the basic task node. The matching results are then filtered in conjunction with the oil and gas field operation constraint information to determine the execution robot of the basic task node. The collaborative execution module is used to determine the multi-robot collaborative operation arrangement based on the constrained hierarchical task tree, and to complete the oil and gas field operation tasks accordingly.

[0056] The specific embodiments of the present invention have been described above with reference to the accompanying drawings. However, the above embodiments are only for explaining the present invention and do not constitute a limitation thereof. Without departing from the technical features defined in the claims of the present invention, those skilled in the art can make necessary equivalent substitutions or partial modifications to the implementation details such as the generation and representation of the hierarchical task tree, the method of writing oil and gas field operation constraint information into task nodes and the continuation content between nodes, the construction of the robot skill library, the determination of the execution subject of the basic task node, and the formation of multi-robot collaborative operation arrangements. Such equivalent substitutions or partial modifications should all fall within the protection scope of the present invention.

Claims

1. A method for multi-robot collaborative operation in oil and gas fields based on a large language model, characterized in that, Includes the following steps: S1. Obtain oil and gas field operation task information, which includes task instructions in natural language form and oil and gas field operation constraint information. S2. Construct task construction rules for generating hierarchical task trees, wherein the task construction rules include a reusable task substructure library, task node connection rules, and structured representation rules for hierarchical task trees. S3. Based on the large language model, perform semantic parsing on the task instructions, and under the constraints of the task construction rules, call the corresponding task substructure from the reusable task substructure library to organize and hierarchically expand the task nodes to form a hierarchical task tree. S4. Write the oil and gas field operation constraint information into the task nodes and the connection content between nodes to form a constrained hierarchical task tree. S5. Obtain the capability information of each robot in the multi-robot system and build a robot skill library; S6. Based on the task requirements of the basic task nodes, candidate execution robots are matched from the robot skill library, and the matching results are filtered in combination with the oil and gas field operation constraint information to determine the execution robot for each basic task node. S7. Determine the multi-robot collaborative operation arrangement based on the constrained hierarchical task tree, and complete the oil and gas field operation task accordingly.

2. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, The oil and gas field operation constraints mentioned in step S1 include explosion-proof level constraints, restricted area constraints, high-risk area constraints, operation priority constraints, operation window constraints, and communication availability constraints.

3. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, The reusable task substructure library mentioned in step S2 includes task substructures for typical oil and gas field operations. The task substructures include inspection substructures, leak detection substructures, equipment status acquisition substructures, and emergency response substructures. Each task substructure is equipped with operation parameter items to adapt to specific operation scenarios. The operation parameter items include operation area, operation object, detection index threshold, collection point set, and operation time window parameters.

4. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, The hierarchical task tree in step S2 includes task nodes organized in a parent-child hierarchy. The task nodes include root task nodes, non-leaf task nodes, and leaf task nodes. The leaf task nodes are basic task nodes that can be executed by the robot.

5. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, In step S3, the task construction rules include node generation content for generating task nodes, hierarchical setting content for determining the hierarchical affiliation of task nodes, continuation setting content for limiting the continuation method of task nodes, and representation content for limiting the requirements of task tree output fields. The large language model determines the task nodes to be included in the hierarchical task tree according to the task instructions, and organizes the task nodes according to the node generation content, hierarchical setting content and succession setting content, and outputs the hierarchical task tree according to the representation content.

6. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, In step S4, the oil and gas field operation constraint information is written into the task node or the connection content between nodes, including: limiting the access conditions of the operation area corresponding to the task node, limiting the safety conditions for the task node to be activated, limiting the collaborative connection requirements between related task nodes, and limiting the operation time window corresponding to the task node.

7. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, The robot skill library mentioned in step S5 includes the operational capability information, operational area adaptability information, and operational condition adaptability information of each robot.

8. The method for multi-robot collaborative operation in oil and gas fields based on a large language model according to claim 1, characterized in that, In step S6, determining the execution subject of the basic task node includes: matching the operational capability information, operational area adaptability information, and operational condition adaptability information of the candidate execution robots with the basic task node based on the task requirements of the basic task node, and filtering the matching results in conjunction with the oil and gas field operation constraint information, thereby determining the execution robot of the basic task node.

9. A multi-robot collaborative operation system for oil and gas fields based on a large language model, characterized in that, include: The task acquisition module is used to receive oil and gas field operation task information, which includes task instructions in natural language and oil and gas field operation constraint information. The task rule preset module is used to preset task construction rules for generating a hierarchical task tree. The task construction rules include a reusable task substructure library, task node connection rules, and structured representation rules for the hierarchical task tree. The task tree generation module is used to perform semantic parsing of the task instructions based on the large language model, and under the constraints of the task construction rules, call the corresponding task substructure from the reusable task substructure library to organize and hierarchically expand the task nodes to form a hierarchical task tree. The constraint writing module is used to write the oil and gas field operation constraint information into the task node or the continuation content between nodes to form a constrained hierarchical task tree. The execution subject determination module is used to acquire the capability information of each robot in the multi-robot system and build a robot skill library. Based on the task requirements of the basic task nodes, it matches candidate execution robots from the robot skill library and filters the matching results in combination with the oil and gas field operation constraint information to determine the execution robot for each basic task node. The collaborative execution module is used to determine the multi-robot collaborative operation arrangement based on the constrained hierarchical task tree, and to complete the oil and gas field operation tasks accordingly.

Images