Unmanned aerial vehicle operation maintenance decision-making method, computer equipment and readable storage medium

Abstract

The invention provides an unmanned aerial vehicle operation maintenance decision-making method, computer equipment and a readable storage medium, and relates to the technical field of unmanned aerial vehicle control. According to the method, the unmanned aerial vehicle operation and maintenance knowledge and hierarchical structure information thereof are carried through the unmanned aerial vehicle operation knowledge tree, and the operation knowledge text data of at least one target knowledge node strongly associated with the natural language query statement representing the unmanned aerial vehicle operation and maintenance requirements are retrieved in the unmanned aerial vehicle operation knowledge tree; and performing statement enhancement processing in cooperation with the natural language query statement to obtain a target enhanced query statement, so that the large language model inferes corresponding unmanned aerial vehicle operation decision information based on the target enhanced query statement, thereby improving the unmanned aerial vehicle operation decision information through organic combination of the tree data structure and the RAG technology. The knowledge recall rate of the unmanned aerial vehicle operation and maintenance knowledge in the big language model decision making process is improved, so that the decision making accuracy and decision making comprehensiveness of the unmanned aerial vehicle operation and maintenance decision making result are improved, and serious safety accidents and property losses are avoided.

Description

Technical Field

[0001] The present application relates to the field of drone control technology, and more specifically, to a drone operation and maintenance decision-making method, a computer device, and a readable storage medium. Background Art

[0002] With the rapid development of drone technology, the application scope of drones (for example, aerial photography, agricultural plant protection, logistics and transportation, etc.) and business volume are constantly expanding. The number and professional quality of drone operators today can hardly meet the growing talent demand in various industries. Therefore, the introduction of a large language model (LLM) with a high level of intelligence to assist in the realization of drone operation and maintenance decisions has become an important direction for various industries to solve the talent gap problem.

[0003] At present, when large language models are used to process complex tasks, RAG (Retrieval Augmented Generation) technology is usually used to provide more efficient and accurate support for large language models. However, it is worth noting that the existing RAG technology is mostly based on a flat data structure design, and usually uses a text slice chain storage method to realize the storage and retrieval of external knowledge bases. When faced with UAV operation and maintenance knowledge with obvious hierarchical characteristics (wherein, UAV operation and maintenance knowledge usually involves UAV operation specifications, safety inspections, fault diagnosis and troubleshooting, daily maintenance, emergency response and other aspects of knowledge, these knowledge have complex hierarchical relationships and associations), it often leads to the loss of complex hierarchical structure information, and it is difficult to effectively organize and present (recall) deep-level UAV operation and maintenance knowledge in the process of using large language models to assist in the realization of UAV operation and maintenance decisions, resulting in the final UAV operation and maintenance decision results being inaccurate and incomplete, which is easy to cause serious safety accidents and property losses. Summary of the invention

[0004] In view of this, the purpose of this application is to provide a UAV operation and maintenance decision-making method, computer equipment and readable storage medium, which can effectively carry UAV operation and maintenance knowledge and its hierarchical structure information through a tree data structure in the process of using a large language model to realize UAV operation and maintenance decisions, so as to combine RAG technology to improve the knowledge recall rate of UAV operation and maintenance knowledge in the large language model decision-making process, improve the decision accuracy and comprehensiveness of the UAV operation and maintenance decision results, and avoid serious safety accidents and property losses.

[0005] In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0006] In a first aspect, the present application provides a method for making decisions on operation and maintenance of a drone, the method comprising:

[0007] Obtain natural language query statements for characterizing the operation and maintenance requirements of drones;

[0008] Retrieving operation knowledge text data of at least one target knowledge node strongly associated with the natural language query statement in a drone operation knowledge tree, wherein the node hierarchical relationship of the drone operation knowledge tree is used to represent hierarchical structure information of drone operation and maintenance rules;

[0009] According to the operation knowledge text data of the at least one target knowledge node, the natural language query statement is subjected to sentence enhancement processing to obtain a target enhanced query statement;

[0010] The large language model is called to perform decision reasoning based on the target enhanced query statement to obtain drone operation decision information adapted to the drone operation and maintenance requirements.

[0011] In an optional embodiment, each tree node of the drone operation knowledge tree separately stores part of the operation knowledge text data of the drone operation and maintenance rules, and a node embedding vector matching the corresponding operation knowledge text data; the step of retrieving the operation knowledge text data of at least one target knowledge node strongly associated with the natural language query statement in the drone operation knowledge tree includes:

[0012] Performing embedding vector conversion on the natural language query statement to obtain an embedding vector to be matched corresponding to the natural language query statement;

[0013] Calculate the vector similarity of the node embedding vectors of all tree nodes in the drone operation knowledge tree with the embedding vector to be matched, and obtain the vector similarity results of all tree nodes arranged in descending order;

[0014] Based on the results of the descending arrangement of the vector similarities, at least one candidate knowledge node having the highest corresponding vector similarity is selected from all the tree nodes;

[0015] For each candidate knowledge node, a target knowledge node associated with the candidate knowledge node is determined in the drone operation knowledge tree, and the operation knowledge text data stored in the corresponding target knowledge node is extracted.

[0016] In an optional embodiment, for each candidate knowledge node, the step of determining a target knowledge node associated with the candidate knowledge node in the drone operation knowledge tree includes:

[0017] Detecting whether the candidate knowledge node has a child node in the drone operation knowledge tree;

[0018] If it is detected whether the candidate knowledge node has a child node in the drone operation knowledge tree, each child node of the candidate knowledge node is used as a target knowledge node respectively; otherwise, the candidate knowledge node is directly used as a target knowledge node.

[0019] In an optional implementation, the step of performing sentence enhancement processing on the natural language query sentence according to the operational knowledge text data of the at least one target knowledge node to obtain a target enhanced query sentence includes:

[0020] Obtain historical enhanced query statements and historical decision information of the large language model for drone operation and maintenance operations;

[0021] The historical enhanced query statement, the historical decision information and the operational knowledge text data of the at least one target knowledge node are used as context statements of the natural language query statement for retrieval enhancement to obtain the target enhanced query statement.

[0022] In an optional implementation, the step of calling the large language model to perform decision reasoning based on the target enhanced query statement to obtain drone operation decision information adapted to the drone operation and maintenance requirements includes:

[0023] Performing named entity recognition on the natural language query statement to obtain all target entities involved in the natural language query statement;

[0024] Searching for actual entity relationships corresponding to all target entities and effective operation and maintenance decision information associated with all target entities in the UAV reliable decision knowledge graph;

[0025] The large language model is called to perform decision reasoning on the target enhanced query statement based on the actual entity relationship and the effective operation and maintenance decision information, so as to obtain the drone operation decision information adapted to the drone operation and maintenance requirements.

[0026] In an optional embodiment, the method further comprises:

[0027] Obtaining reliable operation and maintenance decision information for the operation and maintenance of the UAV, wherein the reliable operation and maintenance decision information is obtained by the UAV operator through correction of the UAV operation decision information inferred by the large language model;

[0028] Performing named entity recognition on the reliable operation and maintenance decision information to obtain a reliable entity relationship corresponding to the reliable operation and maintenance decision information;

[0029] According to the reliable entity relationship and the reliable operation and maintenance decision information, the knowledge graph of the drone reliable decision-making knowledge graph is updated.

[0030] In an optional embodiment, the method further comprises:

[0031] Obtaining the operational rule knowledge text data of the drone operation and maintenance rules;

[0032] A document tree is constructed for the operation rule knowledge text data to obtain a drone operation knowledge tree whose corresponding node hierarchical relationship matches the hierarchical structure information of the drone operation and maintenance rules.

[0033] In an optional implementation, the step of constructing a document tree for the operation rule knowledge text data to obtain a drone operation knowledge tree whose corresponding node hierarchical relationship matches the hierarchical structure information of the drone operation and maintenance rules includes:

[0034] Performing fine-grained text segmentation on the operation rule knowledge text data, and respectively performing embedding vector conversion on the segmented multiple operation rule text data to obtain text embedding vectors corresponding to each of the multiple operation rule text data;

[0035] For each piece of operation rule text data, a text node is constructed based on the piece of operation rule text data and the corresponding text embedding vector to obtain a text node, wherein each text node stores the corresponding operation rule text data and the text embedding vector;

[0036] According to the hierarchical structure information of the operation rule knowledge text data, the ancestor nodes of all text nodes are recursively constructed through text summarization to obtain the drone operation knowledge tree, wherein each ancestor node stores the text summary content of all its child nodes and the summary embedding vector matching the text summary content.

[0037] In a second aspect, the present application provides a computer device comprising a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and the processor can execute the computer program to implement the drone operation and maintenance decision-making method described in any one of the aforementioned embodiments.

[0038] In a third aspect, the present application provides a readable storage medium having a computer program stored thereon. When the computer program is executed by a computer device, the drone operation and maintenance decision-making method described in any one of the aforementioned embodiments is implemented.

[0039] In this case, the beneficial effects of the embodiments of the present application may include the following:

[0040] The present application carries the drone operation and maintenance knowledge and its hierarchical structure information through a drone operation knowledge tree, and retrieves the operation knowledge text data of at least one target knowledge node that is strongly associated with a natural language query statement that represents the drone operation and maintenance requirements in the drone operation knowledge tree, performs sentence enhancement processing in conjunction with the natural language query statement, and obtains a target enhanced query statement, so that the large language model can directly perform decision reasoning based on the target enhanced query statement to obtain the corresponding drone operation decision information, thereby improving the knowledge recall rate of the drone operation and maintenance knowledge in the decision-making process of the large language model through the organic combination of the tree data structure and the RAG technology, thereby improving the decision accuracy and comprehensiveness of the drone operation and maintenance decision results, and avoiding serious safety accidents and property losses.

[0041] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 A schematic diagram of the composition of a computer device provided in an embodiment of the present application;

[0044] Figure 2 One of the flow charts of the drone operation and maintenance decision-making method provided in the embodiment of the present application;

[0045] Figure 3 for Figure 2 A schematic flow chart of the sub-steps included in step S220;

[0046] Figure 4 for Figure 2 A schematic flow chart of the sub-steps included in step S240;

[0047] Figure 5 A second flowchart of the drone operation and maintenance decision-making method provided in an embodiment of the present application;

[0048] Figure 6 The third flowchart of the drone operation and maintenance decision-making method provided in the embodiment of the present application.

[0049] Icons: 10 - computer device; 11 - memory; 12 - processor; 13 - communication unit. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0051] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0052] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0053] In the description of the present application, it should be understood that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the product of the application is conventionally placed when in use, or are the orientations or positional relationships conventionally understood by those skilled in the art. They are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0054] In the description of this application, it should also be noted that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0055] In addition, in the description of the present application, it is understood that the relational terms such as the term "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only those elements, but also includes other elements that are not clearly listed, or also includes elements inherent to such process, method, article or equipment. In the absence of more restrictions, the elements defined by the sentence "comprise one..." do not exclude the existence of other identical elements in the process, method, article or equipment including the elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood by specific circumstances.

[0056] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0057] Please refer to Figure 1 , Figure 1 : is a schematic diagram of the composition of the computer device 10 provided in the embodiment of the present application. In the embodiment of the present application, the computer device 10 can assist the drone operator in making drone operation and maintenance decisions through natural language; the computer device 10 can be an electronic device that is connected to the terminal device held by the drone operator, wherein the electronic device can be but not limited to: a personal computer, a server, a gateway device, etc.; the computer device 10 can also be integrated with the terminal device held by the drone operator, wherein the terminal device can be but not limited to: a personal computer, a laptop, a tablet computer, a smart phone, etc.

[0058] In the embodiment of the present application, the computer device 10 may include a memory 11, a processor 12, and a communication unit 13. The memory 11, the processor 12, and the communication unit 13 are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, the memory 11, the processor 12, and the communication unit 13 may be electrically connected to each other via one or more communication buses or signal lines.

[0059] In the embodiment of the present application, the memory 11 may be, but not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), etc. The memory 11 is used to store a computer program, and the processor 12 may execute the computer program accordingly after receiving an execution instruction.

[0060] Optionally, in this embodiment, the memory 11 can also be used to build various vector databases (for example, chroma vector databases), so as to localize the data storage of the drone operation knowledge tree carrying the drone operation and maintenance knowledge and the hierarchical structure information of the drone operation and maintenance knowledge through the built vector database, so as to facilitate the external knowledge call in the decision-making process of the large language model. Among them, the node hierarchical relationship of the drone operation knowledge tree can be used to characterize the hierarchical structure information of the drone operation and maintenance knowledge (rules), and each tree node in the drone operation knowledge tree separately stores part of the operation knowledge text data of the drone operation and maintenance knowledge (rules); for each non-root node in the drone operation knowledge tree, the operation knowledge text data stored at the parent node of the non-root node includes the text summary content of the operation knowledge text data of the non-root node and its sibling nodes (for example, the text chapter title of the drone operation and maintenance knowledge).

[0061] Optionally, in this embodiment, the memory 11 can also be used to store a drone reliable decision knowledge graph between various drone-related entities (e.g., entities "drone", "drone airport", "drone emergency airport", etc.) related to drone operation and maintenance operations, so as to further improve the accuracy and reliability of decisions by calling the drone reliable decision knowledge graph in the large language model decision process, and avoid causing serious safety accidents and property losses. Among them, the drone reliable decision knowledge graph is used to describe the reliable entity relationships between the various drone-related entities, and the reliable operation and maintenance decision information of various drone-related entity combinations with reliable entity relationships, and the reliable operation and maintenance decision information is obtained by the drone operator correcting the drone operation decision information determined by the large language model (that is, by reviewing the drone operation decision information to correct errors or omissions in the corresponding decision information).

[0062] In the embodiment of the present application, the processor 12 may be an integrated circuit chip with signal processing capability. The processor 12 may be a general-purpose processor, including a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or at least one of other programmable logic devices, discrete gates or transistor logic devices, and discrete hardware components. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc., which may implement or execute the disclosed methods, steps, and logic block diagrams in the embodiment of the present application.

[0063] In the embodiment of the present application, the communication unit 13 is used to establish a communication connection between the computer device 10 and other electronic devices through a network, and to send and receive data through the network, wherein the network includes a wired communication network and a wireless communication network. For example, the computer device 10 can obtain a natural language query statement entered by a drone operator through a terminal device through the communication unit 13 to characterize the operation and maintenance requirements of the drone.

[0064] In an embodiment of the present application, the computer device 10 may pre-store a specific computer program related to the UAV operation and maintenance decision-making function in the memory 11, and by driving the processor 12 to execute the specific computer program accordingly, in the process of using a large language model to implement UAV operation and maintenance decisions, the UAV operation and maintenance knowledge and its hierarchical structure information are effectively carried through a tree data structure, so as to combine RAG technology to improve the knowledge recall rate of UAV operation and maintenance knowledge in the large language model decision-making process, so that deep-level UAV operation and maintenance knowledge can be normally called in the large language model decision-making process, so as to improve the decision accuracy and comprehensiveness of the UAV operation and maintenance decision results, and avoid causing serious safety accidents and property losses.

[0065] Understandably, Figure 1 The block diagram shown is only a schematic diagram of a composition of the computer device 10. The computer device 10 may also include Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown. Figure 1 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0066] In the present application, in order to ensure that the computer device 10 can improve the knowledge recall rate of the drone operation and maintenance knowledge with hierarchical structure characteristics in the large language model decision-making process, so that the deep-level drone operation and maintenance knowledge can be normally called in the RAG technology implementation process, to improve the decision accuracy and comprehensiveness of the drone operation and maintenance decision results, and avoid causing serious safety accidents and property losses, the embodiment of the present application achieves the above-mentioned purpose by providing a drone operation and maintenance decision method. The drone operation and maintenance decision method provided by the present application is described in detail below.

[0067] Please refer to Figure 2 , Figure 2 This is one of the flow charts of the drone operation and maintenance decision method provided in the embodiment of the present application. In the embodiment of the present application, the drone operation and maintenance decision method may include steps S210 to S240.

[0068] Step S210: obtaining a natural language query statement for characterizing the operation and maintenance requirements of the drone.

[0069] Step S220, retrieving the operation knowledge text data of at least one target knowledge node strongly associated with the natural language query statement in the drone operation knowledge tree, wherein the node hierarchical relationship of the drone operation knowledge tree is used to represent the hierarchical structure information of the drone operation and maintenance rules.

[0070] In this embodiment, the computer device 10 can use a tree data structure to carry the UAV operation and maintenance knowledge and its hierarchical structure information, and on the basis of the UAV operation knowledge tree (i.e., the tree data structure that carries the UAV operation and maintenance knowledge and its hierarchical structure information), combine the retrieval technology included in the RAG technology to deeply explore the knowledge details in the UAV operation and maintenance knowledge that are strongly associated with the natural language query statements from the UAV operator, so as to extract the operation knowledge text data of all target knowledge nodes in the UAV operation knowledge tree that are strongly associated with the natural language query statements, so as to improve the knowledge recall rate of the UAV operation and maintenance knowledge with hierarchical structure characteristics in the large language model decision-making process, so that the deep-level UAV operation and maintenance knowledge can be normally called in the decision-making process.

[0071] Optionally, see Figure 3 , Figure 3 yes Figure 2 Flow chart of sub-steps included in step S220. In an embodiment of the present application, step S220 may include sub-steps S221 to S224 to dig deep into the knowledge details of the drone operation and maintenance knowledge that are strongly associated with the natural language query statements from the drone operator, and improve the knowledge recall rate of the drone operation and maintenance knowledge in the large language model decision-making process.

[0072] Sub-step S221 , performing embedding vector conversion on the natural language query statement to obtain an embedding vector to be matched corresponding to the natural language query statement.

[0073] Sub-step S222, respectively, calculates the vector similarity of the node embedding vectors of all tree nodes in the drone operation knowledge tree with the embedding vector to be matched, and obtains the vector similarity results of all tree nodes arranged in descending order.

[0074] In this embodiment, each tree node of the drone operation knowledge tree stores part of the operation knowledge text data of the drone operation and maintenance rules, and the node embedding vector that matches the corresponding operation knowledge text data. After obtaining the embedding vector to be matched corresponding to the natural language query statement, the computer device 10 can obtain the node embedding vectors of all tree nodes in the drone operation knowledge tree by accessing the vector database storing the drone operation knowledge tree, and then perform vector similarity calculation (for example, cosine similarity calculation) on the node embedding vectors of all tree nodes and the embedding vector to be matched, and obtain the vector similarity of all tree nodes and the embedding vector to be matched, and then arrange all tree nodes in descending order of vector similarity according to the vector similarity corresponding to all tree nodes, and obtain the corresponding vector similarity descending order arrangement result.

[0075] Sub-step S223, based on the results of the descending arrangement of the vector similarities, at least one candidate knowledge node with the highest corresponding vector similarity is selected from all the tree nodes.

[0076] In this embodiment, the computer device 10 can extract a preset number (for example, 5) of tree nodes with the highest corresponding vector similarities from the vector similarity descending arrangement results, and use each extracted tree node as a candidate knowledge node strongly associated with the natural language query statement.

[0077] Sub-step S224, for each candidate knowledge node, determine the target knowledge node associated with the candidate knowledge node in the drone operation knowledge tree, and extract the operation knowledge text data stored in the corresponding target knowledge node.

[0078] In this embodiment, for each determined candidate knowledge node, by detecting whether the candidate knowledge node has a child node in the drone operation knowledge tree, it can be determined whether the operation knowledge text data stored in the candidate knowledge node belongs to the deep-level knowledge details in the drone operation and maintenance knowledge, so as to further confirm whether the candidate knowledge node should be used as a target knowledge node strongly associated with the natural language query statement, or each child node of the candidate knowledge node should be used as a target knowledge node strongly associated with the natural language query statement. Among them, when a candidate knowledge node does not have a child node in the drone operation knowledge tree, it indicates that the operation knowledge text data stored in the candidate knowledge node belongs to the deep-level knowledge details in the drone operation and maintenance knowledge, and at this time, the candidate knowledge node can be directly used as a target knowledge node strongly associated with the natural language query statement; and when a candidate knowledge node has a child node in the drone operation knowledge tree, it indicates that the operation knowledge text data stored in the candidate knowledge node does not belong to the deep-level knowledge details in the drone operation and maintenance knowledge, and at this time, each child node of the candidate knowledge node can be used as a target knowledge node strongly associated with the natural language query statement.

[0079] Thus, the step of “for each candidate knowledge node, determining a target knowledge node associated with the candidate knowledge node in the drone operation knowledge tree” may include:

[0080] Detecting whether the candidate knowledge node has a child node in the drone operation knowledge tree;

[0081] If it is detected whether the candidate knowledge node has a child node in the drone operation knowledge tree, each child node of the candidate knowledge node is used as a target knowledge node respectively; otherwise, the candidate knowledge node is directly used as a target knowledge node.

[0082] Therefore, the present application can, by executing the above-mentioned sub-steps S221 to S224, deeply explore the knowledge details in the drone operation and maintenance knowledge that are strongly associated with the natural language query statements from the drone operator, and improve the knowledge recall rate of the drone operation and maintenance knowledge in the large language model decision-making process, so as to effectively improve the decision accuracy and comprehensiveness of the drone operation and maintenance decision results.

[0083] Step S230, performing sentence enhancement processing on the natural language query sentence according to the operational knowledge text data of at least one target knowledge node to obtain a target enhanced query sentence.

[0084] In this embodiment, after the computer device 10 retrieves the operation knowledge text data of at least one target knowledge node that is strongly associated with the natural language query statement from the drone operation knowledge tree, it can use the enhancement technology included in the RAG technology to textually combine the operation knowledge text data of the at least one target knowledge node with the natural language query statement to obtain a corresponding target enhanced query statement, so as to use the target enhanced query statement to improve the understanding and analysis capabilities of the large language model in the drone operation and maintenance scenario.

[0085] Optionally, in one implementation of the present embodiment, the computer device 10 may obtain the target enhanced query statement by using the operational knowledge text data of the at least one target knowledge node as the context statement of the natural language query statement, and performing retrieval enhancement (i.e., prompt word expansion) on the natural language query statement in combination with the context statement under the same prompt template.

[0086] Optionally, in another implementation of this embodiment, the computer device 10 may obtain historical enhanced query statements and historical decision information of the large language model for drone operation and maintenance operations, and then use the acquired historical enhanced query statements and historical decision information, as well as the operational knowledge text data of the at least one target knowledge node, as context statements of the natural language query statements to perform retrieval enhancement on the natural language query statements, thereby obtaining target enhanced query statements that are more in line with the drone operation and maintenance needs of drone operators.

[0087] Step S240, calling the large language model to perform decision reasoning based on the target enhanced query statement to obtain drone operation decision information adapted to the drone operation and maintenance requirements.

[0088] In one implementation of the present embodiment, after the computer device 10 determines the target enhanced query statement using the enhancement technology included in the RAG technology, it can directly input the target enhanced query statement into the large language model, and the large language model adaptively generates drone operation decision information that is adapted to the drone operation and maintenance requirements based on the generation technology included in the RAG technology.

[0089] In another implementation of this embodiment, considering that the direct combination of RAG technology and the tree data structure may still have problems such as inaccurate entity recognition and missing information (for example, the drone-related entity "drone emergency airport" in the natural language query statement is incorrectly identified as "drone airport", resulting in tree node data retrieval errors and text missing), it is impossible to meet the high accuracy required for reliable drone operation and maintenance operations. This application can improve the decision reliability and accuracy of the decision results of the large language model by introducing a drone reliable decision knowledge graph in the decision generation stage.

[0090] On this basis, please refer to Figure 4 , Figure 4 yes Figure 2 Schematic diagram of the flow of sub-steps included in step S240. In the embodiment of the present application, step S240 may include sub-steps S241 to S243 to improve the decision reliability and accuracy of the decision result of the large language model by introducing the drone reliable decision knowledge graph.

[0091] Sub-step S241 , performing named entity recognition on the natural language query statement to obtain all target entities involved in the natural language query statement.

[0092] Sub-step S242, searching the UAV reliable decision knowledge graph for actual entity relationships corresponding to all target entities, as well as effective operation and maintenance decision information associated with all target entities.

[0093] In this embodiment, the computer device 10 can use the Aho-Corasick automaton to request the entity relationships related to all the target entities and the recorded reliable operation and maintenance decision information from the drone reliable decision knowledge graph to obtain the actual entity relationships between all the target entities and the effective operation and maintenance decision information associated with all the target entities.

[0094] Sub-step S243, calling the large language model to perform decision reasoning on the target enhanced query statement based on the actual entity relationship and the effective operation and maintenance decision information, and obtaining the drone operation decision information adapted to the drone operation and maintenance requirements.

[0095] Therefore, the present application can improve the decision reliability and accuracy of the decision results of the large language model by executing the above sub-steps S241 to S243 and introducing the drone reliable decision knowledge graph.

[0096] The present application can execute the above steps S210 to S240, and in the process of using a large language model to realize UAV operation and maintenance decisions, effectively carry the UAV operation and maintenance knowledge and its hierarchical structure information through a tree data structure, so as to combine RAG technology to improve the knowledge recall rate of UAV operation and maintenance knowledge in the large language model decision-making process, so that deep-level UAV operation and maintenance knowledge can be normally called in the large language model decision-making process, thereby improving the decision accuracy and comprehensiveness of the UAV operation and maintenance decision results, and avoiding serious safety accidents and property losses.

[0097] Optionally, see Figure 5 , Figure 5 This is the second flow chart of the drone operation and maintenance decision-making method provided in the embodiment of the present application. Figure 2 Compared with the UAV operation and maintenance decision-making method shown in the figure, Figure 5 The drone operation and maintenance decision method shown may also include steps S250 to S270 to ensure that the corresponding constructed reliable decision knowledge graph can effectively capture the entity relationships between various drone-related entities and carry reliable operation and maintenance decision content approved by drone operators, so as to improve the decision accuracy and reliability of the decision results in the large language model decision process.

[0098] Step S250, obtaining reliable operation and maintenance decision information for the operation and maintenance task of the UAV, wherein the reliable operation and maintenance decision information is obtained by the UAV operator by correcting the UAV operation decision information inferred by the large language model.

[0099] Step S260: performing named entity recognition on the reliable operation and maintenance decision information to obtain a reliable entity relationship corresponding to the reliable operation and maintenance decision information.

[0100] Step S270, updating the knowledge graph of the UAV reliable decision-making knowledge graph according to the reliable entity relationship and the reliable operation and maintenance decision information.

[0101] In this embodiment, the knowledge graph update operations performed by the computer device 10 on the drone reliable decision knowledge graph may include but are not limited to: adding drone-related entities, deleting drone entity relationships, adding drone entity relationships, deleting drone entity relationships, adding operation and maintenance decision information, changing operation and maintenance decision information, and other operations.

[0102] Therefore, the present application can ensure that the corresponding constructed reliable decision knowledge graph can effectively capture the entity relationship between various drone-related entities and carry reliable operation and maintenance decision content approved by drone operators by executing the above steps S250 to S270, so as to improve the decision accuracy and reliability of the decision results in the large language model decision-making process.

[0103] Optionally, see Figure 6 , Figure 6 This is the third flow chart of the drone operation and maintenance decision-making method provided in the embodiment of the present application. Figure 2 or Figure 5 Compared with the UAV operation and maintenance decision-making method shown in the figure, Figure 6 The drone operation and maintenance decision method shown may also include steps S280 to S290 to ensure that the corresponding constructed drone operation knowledge tree can effectively carry the drone operation and maintenance knowledge and its hierarchical structure information based on the tree data structure, so as to facilitate the combination of RAG technology to improve the knowledge recall rate of drone operation and maintenance knowledge in the large language model decision-making process.

[0104] Step S280, obtaining the operation rule knowledge text data of the drone operation and maintenance rules.

[0105] In this embodiment, the computer device 10 can obtain a large number of operation and maintenance rule documents representing the operation and maintenance knowledge of the drone, and obtain operation rule knowledge text data matching the operation and maintenance rules of the drone by converting the acquired operation and maintenance rule documents into text format.

[0106] Step S290, constructing a document tree for the operation rule knowledge text data to obtain a drone operation knowledge tree whose corresponding node hierarchical relationship matches the hierarchical structure information of the drone operation and maintenance rules.

[0107] In this embodiment, after acquiring the operation rule knowledge text data of the drone operation and maintenance rules, the computer device 10 can perform fine-grained text segmentation on the operation rule knowledge text data according to the distribution of symbols such as periods, semicolons, commas, and line breaks in the operation rule knowledge text data, and cluster sentences with the same context according to the text meanings of the segmented sentences to obtain multiple operation rule text data, wherein each operation rule text data is formed by a combination of multiple sentences in the same context; then, each segmented operation rule text data can be embedded in a vector conversion to obtain a text corresponding to the operation rule text data. This embedding vector is used to construct a text node based on the operation rule text data and the corresponding text embedding vector to form a leaf node of the drone operation knowledge tree, so that each text node can store an operation rule text data and a corresponding text embedding vector separately, and can serve as a leaf node in the drone operation knowledge tree. At this time, for any text node, the operation rule text data stored in the text node is the operation knowledge text data stored in the corresponding leaf node in the drone operation knowledge tree, and the text embedding vector stored in the text node is the node embedding vector stored in the corresponding leaf node in the drone operation knowledge tree.

[0108] After the computer device 10 constructs all text nodes for the operation rule knowledge text data, it will characterize the hierarchical structure information of the operation rule knowledge text data by extracting the title association relationship between the text chapter titles in the operation rule knowledge text data, and recursively construct ancestor nodes (including parent nodes and grandparent nodes, etc.) of all text nodes in a text summarization manner according to the subordinate relationship between each segmented sentence and each text chapter title in the operation rule knowledge text data and the hierarchical structure information of the operation rule knowledge text data, so that each constructed ancestor node stores the text summary content of all its child nodes, and The summary embedding vector matching the text summary content is constructed in sequence to construct each non-leaf node in the UAV operation knowledge tree. At this time, for any non-leaf node, the text summary content stored in the non-leaf node is the operation knowledge text data of the non-leaf node at the UAV operation knowledge tree, and the summary embedding vector stored in the non-leaf node is the node embedding vector of the non-leaf node at the UAV operation knowledge tree, thereby ensuring that the node hierarchical relationship of the constructed UAV operation knowledge tree matches the hierarchical structure information of the UAV operation and maintenance rules (knowledge), and at the same time, the UAV operation knowledge tree can effectively carry the UAV operation and maintenance knowledge through the tree data structure.

[0109] In this case, the above-mentioned step S290 may include: performing fine-grained text segmentation on the operation rule knowledge text data, and respectively performing embedding vector conversion on the multiple segments of operation rule text data to obtain text embedding vectors corresponding to each of the multiple operation rule text data; for each operation rule text data, constructing a text node based on the operation rule text data and the corresponding text embedding vector to obtain a text node, wherein each text node stores the corresponding operation rule text data and the text embedding vector; according to the hierarchical structure information of the operation rule knowledge text data, recursively constructing ancestor nodes for all text nodes through text summarization to obtain the drone operation knowledge tree, wherein each ancestor node stores the text summary content of all its child nodes, and a summary embedding vector matching the text summary content.

[0110] Therefore, the present application can ensure that the corresponding constructed drone operation knowledge tree can effectively carry the drone operation and maintenance knowledge and its hierarchical structure information based on the tree data structure by executing the above steps S280 to S290, so as to facilitate the combination of RAG technology to improve the knowledge recall rate of drone operation and maintenance knowledge in the decision-making process of the large language model.

[0111] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic, for example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to the embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or the flowchart, and the combination of boxes in the block diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0112] In addition, each functional module in each embodiment of the present application can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part. If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a readable storage medium. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a readable storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk.

[0113] The above are only various implementations of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for decision-making on operation and maintenance of unmanned aerial vehicles, characterized in that: The method comprises: Obtain natural language query statements for characterizing the operation and maintenance requirements of drones; Retrieving operation knowledge text data of at least one target knowledge node strongly associated with the natural language query statement in a drone operation knowledge tree, wherein the node hierarchical relationship of the drone operation knowledge tree is used to represent hierarchical structure information of drone operation and maintenance rules; According to the operation knowledge text data of the at least one target knowledge node, the natural language query statement is subjected to sentence enhancement processing to obtain a target enhanced query statement; The large language model is called to perform decision reasoning based on the target enhanced query statement to obtain drone operation decision information adapted to the drone operation and maintenance requirements.

2. The method according to claim 1, characterized in that Each tree node of the drone operation knowledge tree separately stores part of the operation knowledge text data of the drone operation and maintenance rules, and a node embedding vector matching the corresponding operation knowledge text data; The step of retrieving the operation knowledge text data of at least one target knowledge node strongly associated with the natural language query statement in the drone operation knowledge tree includes: Performing embedding vector conversion on the natural language query statement to obtain an embedding vector to be matched corresponding to the natural language query statement; Calculate the vector similarity of each node embedding vector of all tree nodes in the drone operation knowledge tree with the embedding vector to be matched, and obtain the vector similarity results of all tree nodes arranged in descending order; Based on the results of the descending arrangement of the vector similarities, at least one candidate knowledge node having the highest corresponding vector similarity is selected from all the tree nodes; For each candidate knowledge node, a target knowledge node associated with the candidate knowledge node is determined in the drone operation knowledge tree, and the operation knowledge text data stored in the corresponding target knowledge node is extracted.

3. The method according to claim 2, characterized in that For each candidate knowledge node, the step of determining a target knowledge node associated with the candidate knowledge node in the drone operation knowledge tree includes: Detecting whether the candidate knowledge node has a child node in the drone operation knowledge tree; If it is detected whether the candidate knowledge node has a child node in the drone operation knowledge tree, each child node of the candidate knowledge node is used as a target knowledge node respectively; otherwise, the candidate knowledge node is directly used as a target knowledge node.

4. The method according to claim 1, characterized in that The step of performing sentence enhancement processing on the natural language query sentence according to the operational knowledge text data of the at least one target knowledge node to obtain a target enhanced query sentence comprises: Obtain historical enhanced query statements and historical decision information of the large language model for drone operation and maintenance operations; The historical enhanced query statement, the historical decision information and the operational knowledge text data of the at least one target knowledge node are used as context statements of the natural language query statement for retrieval enhancement to obtain the target enhanced query statement.

5. The method according to claim 1, characterized in that The step of calling the large language model to perform decision reasoning based on the target enhanced query statement to obtain drone operation decision information adapted to the drone operation and maintenance requirements includes: Performing named entity recognition on the natural language query statement to obtain all target entities involved in the natural language query statement; Searching for actual entity relationships corresponding to all target entities and effective operation and maintenance decision information associated with all target entities in the UAV reliable decision knowledge graph; The large language model is called to perform decision reasoning on the target enhanced query statement based on the actual entity relationship and the effective operation and maintenance decision information, so as to obtain the drone operation decision information adapted to the drone operation and maintenance requirements.

6. The method according to claim 5, characterized in that The method further comprises: Obtaining reliable operation and maintenance decision information for the operation and maintenance of the UAV, wherein the reliable operation and maintenance decision information is obtained by the UAV operator through correction of the UAV operation decision information inferred by the large language model; Performing named entity recognition on the reliable operation and maintenance decision information to obtain a reliable entity relationship corresponding to the reliable operation and maintenance decision information; According to the reliable entity relationship and the reliable operation and maintenance decision information, the knowledge graph of the drone reliable decision-making knowledge graph is updated.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Obtaining the operational rule knowledge text data of the drone operation and maintenance rules; A document tree is constructed for the operation rule knowledge text data to obtain a drone operation knowledge tree whose corresponding node hierarchical relationship matches the hierarchical structure information of the drone operation and maintenance rules.

8. The method according to claim 7, characterized in that The step of constructing a document tree for the operation rule knowledge text data to obtain a drone operation knowledge tree whose corresponding node hierarchical relationship matches the hierarchical structure information of the drone operation and maintenance rules includes: Performing fine-grained text segmentation on the operation rule knowledge text data, and respectively performing embedding vector conversion on the segmented multiple operation rule text data to obtain text embedding vectors corresponding to each of the multiple operation rule text data; For each piece of operation rule text data, a text node is constructed based on the piece of operation rule text data and the corresponding text embedding vector to obtain a text node, wherein each text node stores the corresponding operation rule text data and the text embedding vector; According to the hierarchical structure information of the operation rule knowledge text data, the ancestor nodes of all text nodes are recursively constructed through text summarization to obtain the drone operation knowledge tree, wherein each ancestor node stores the text summary content of all its child nodes and the summary embedding vector matching the text summary content.

9. A computer device, characterized in that: It includes a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and the processor can execute the computer program to implement the drone operation and maintenance decision method described in any one of claims 1 to 8.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a computer device, the drone operation and maintenance decision method described in any one of claims 1 to 8 is implemented.

Images