999 results about "Semantics" patented technology

The invention discloses a data knowledge-based method based on semantic fusion, and relates to the technical field of computer information processing.The method comprises the steps that a requirement set serves as input, knowledge requirement analysis, concept modeling, relation modeling and constraint declaration are completed, and a semantic model OB is constructed; taking the original data set Sraw as input, completing standardization processing and structure segmentation under the support of a semantic model OB, and forming an entity corpus and a semantic unit set; based on the semantic unit set, structured and unstructured triple extraction, semantic verification and graph loading are executed, and an initial knowledge graph is constructed; performing entity alignment, relationship merging, rule reasoning and versioning release on the initial knowledge graph to generate a graph; introducing a quality evaluation mechanism, and outputting an optimized atlas and an evaluation document; the KG opt deployment is online, and query packaging, visualization, service arrangement and incremental maintenance are completed. The method aims at solving the problems that multi-source heterogeneous data are not uniform in structure and inconsistent in semantics.

The invention relates to a large language model construction method and system fused with spatial semantic understanding, and the method comprises the steps: obtaining a multi-source heterogeneous corpus, and extracting an entity, an attribute and a business rule; extracting a semantic feature vector set based on the multi-source heterogeneous corpus, and constructing an entity relationship network and an enhanced knowledge graph; generating an enhanced training sample, and training the general large language model to obtain a primary large language model; generating a verification sample set and performing verification; identifying a specific weakness pattern, and generating a corresponding confrontation sample and a knowledge enhancement sample; training the primary large language model to obtain an optimized large language model; in conclusion, the enhanced knowledge graph fusing the spatial semantic features and the business rules is constructed, and the gradient training samples are generated based on the graph to perform multi-stage model training and optimization, so that the method has the effects of improving the internalized understanding ability of the model for the spatial semantics and the business rules and enhancing the reliability of multi-step spatial reasoning.

The invention provides a natural language text data intelligent classification method and system based on deep learning, and relates to the technical field of natural language processing, and the method comprises the steps: 1, employing a context awareness mechanism to analyze the real semantics of a target vocabulary according to an antagonistic variant existing in a text, and obtaining a target vocabulary; in combination with a word meaning library and a pre-training process of a dynamic learning rate adjustment strategy, generating a candidate replacement vocabulary set with consistent semantics; and step 2, based on the candidate replacement vocabulary set, performing multi-dimensional semantic similarity calculation and emotional tendency discrimination, determining applicable vocabularies conforming to an original culture background through a context adaptation strategy, and generating a standardized text sequence. According to the method, through multi-dimensional semantic analysis, cultural context fusion, cross-granularity feature construction and dynamic parameter correction, the accuracy and adaptability of natural language text classification are realized.

The embodiments of the present application relate to the technical field of artificial intelligence. Provided are a prompt anomaly detection method and apparatus, and a device and a storage medium. The method comprises: acquiring a prompt to be subjected to detection and the perplexity of said prompt; if the perplexity exceeds a second threshold value and does not exceed a first threshold value, using an adversarial perturbation mode to process said prompt, in order to acquire a perturbed prompt; on the basis of prediction by a large language model, obtaining response semantics of said prompt and perturbed response semantics corresponding to the perturbed prompt; and if the response semantics and the perturbed response semantics are inconsistent, said prompt being anomalous. Two threshold values are set for perplexity detection, and after perturbation processing is performed on a prompt to be subjected to detection, whether the prompt is anomalous can be determined again by means of the comparison between response semantics and perturbed response semantics, so that the accuracy of prompt anomaly detection can be effectively improved, thereby enhancing the security of large language models.

The invention belongs to the technical field of industrial equipment life prediction, and discloses an equipment life prediction method and system of a large language model based on semantic alignment, and the method comprises the steps: obtaining original multi-dimensional sensor time sequence data, and obtaining an embedded matrix after preprocessing; constructing a prompt text with domain semantics to obtain a natural language embedded representation; constructing a semantic text prototype, and realizing alignment of the embedding matrix and the semantic text prototype to obtain a patch embedding sequence; and splicing the patch embedding sequence and the natural language embedding representation, inputting the spliced patch embedding sequence and the natural language embedding representation into a pre-trained large language model for forward propagation, extracting hidden vectors output corresponding to the patch embedding sequence, splicing and flattening the hidden vectors into a single vector, and outputting to obtain an equipment life prediction result. According to the method, cross-modal knowledge learned by the LLM in large-scale pre-training and the powerful reasoning ability are fully utilized, and accurate prediction of the residual life of the equipment is achieved.

The invention relates to the technical field of enterprise management, in particular to an intelligent decision support system and method based on cognitive logic and scenarized semanteme, and the method comprises the steps: obtaining enterprise decision data through multi-source data monitoring, carrying out the preprocessing, and generating cross-modal enterprise scene cognitive information; analyzing cross-modal association among the data, fusing a cognitive logical reasoning rule and a semantic understanding model, and constructing scenarized semantic decision fusion features; training an enterprise decision-making quality evaluation model based on historical cases, performing quality perception on a current decision-making scene, and outputting decision-making quality information; and an execution effect is judged in combination with an expected target, an optimization mechanism is started if the target is deviated, candidate schemes are generated by using a historical knowledge base and a multi-target optimization algorithm, an optimal solution is screened, and intelligent adjustment of a decision scheme is realized. According to the method, multi-source heterogeneous data and cognitive logic are fused, semantic understanding, dynamic evaluation and adaptive optimization capabilities of a decision system are enhanced, and scientificity and real-time performance of enterprise decision are improved.

The invention relates to the technical field of artificial intelligence and natural language processing, in particular to a cue word compiling and generating method and system for a large language model.The method comprises the steps that in response to a compiling calling request, a PDL protocol is received to serve as a compiling input text; performing lexical and grammatical analysis on the compiled input text, and constructing a corresponding abstract syntax tree AST as an intermediate representation according to a context-independent method; semantic verification and structure correction are carried out on the AST, and when semantics are effective, structure optimization is carried out to generate an optimized AST; and traversing the optimized AST, matching a cue word template in a cue word generation rule base, mapping semantic nodes into structured text fragments, splicing the structured text fragments to form a final cue word text, and outputting a compilation result consistent with PDL protocol semantics. By introducing a PDL-oriented compiling mechanism, automatic analysis, structure optimization and high-quality cue word generation of an input protocol are realized before cue words of a large language model are generated, and the development efficiency and the output quality of the cue words are improved.

The invention discloses a knowledge base technical method and system based on RAG retrieval enhancement, and relates to the technical field of knowledge bases. The method comprises the steps that a source document is analyzed into structured text fragments, and semantic vectors are generated; splitting the text segments into minimum knowledge units, extracting concepts and behavior trigger words to construct a concept association graph, and storing session abstracts in a fixed-length annular structure; the method comprises the following steps: receiving user query, generating a query vector, retrieving a most relevant text fragment, constructing and de-duplicating candidate knowledge units by combining hierarchical diffusion of a concept association graph and an annular abstract matching result, deeply splicing original text paragraphs according to a graph path, generating a dynamic prompt box, calling a generative model to generate a preliminary answer, and executing verification. And the verification state is marked on the final answer. By constructing a concept association map, the activeness weight and the time sequence fingerprint of a map edge are updated in real time, and accurate capture of deep semantics and logical relationships of query intentions is realized.

The invention discloses a knowledge graph reasoning method based on dynamic rule perception memory. The method aims at solving the problems that an existing neural combination rule learning method is insufficient in expression ability and prone to splitting global semantics and local relation modes. The method comprises the following steps: introducing a lightweight dynamic relationship memory module, and executing self-attention on all relationships in a knowledge graph to capture global semantics; meanwhile, local relation interaction features are extracted through convolution, semantic fusion is carried out through a Transform encoder with multi-head attention and relative position coding, and unified and parallel-computing high-quality combination representation is constructed for each pair of relations in parallel. And meanwhile, a closed high-quality reasoning path is generated in combination with bidirectional breadth-first search. Through global semantic and local interactive collaborative modeling, the extendibility is ensured, the understanding of global semantics is enhanced, and the reasoning accuracy is improved. Global and local relation dependence is fused, an interpretable reasoning path is generated, and reasoning accuracy, expandability and robustness are improved.

The invention discloses a small sample target detection method based on target feature enhancement and semantic fusion perception, and relates to a computer vision technology. A data set is divided into a query set and a support set, after features are extracted through a backbone network, background noise in the support features is inhibited through a dynamic hypergraph construction module, and high-order semantic association of a target area is enhanced; fusing the category name text semantics and the image specific prototype by using a semantic fusion perception module to generate a high-discrimination category prototype; modeling semantic distribution by means of a variational auto-encoder, and extracting variational features; and fusing the region-of-interest features and the variation features through a channel attention mechanism to realize classification and regression. According to the method, the prototype characterization capability is effectively improved, and experiments show that the method remarkably improves the detection precision and is suitable for labeling sample scarce scenes. More accurate small sample target detection is realized by enhancing the feature expression of the support feature map and the semantic meaning of the category prototype, and higher robustness and recognition performance are shown in a complex scene.

The invention relates to the technical field of electrical digital data processing, and discloses a system conflict point intelligent review system based on a knowledge graph and a large language model, which comprises the following steps: constructing a dual-mode storage space containing an unstructured index and a structured logic graph, analyzing target text extraction features and triggering graph-based generation logic; converting the topological structure of the associated sub-atlas into a natural language instruction sequence to construct a forced logic constraint template, filling the template with a text, and inputting a pre-training language model to generate a verification result; according to the method, the discrete atlas topology is mapped into the linear logic constraint, random divergence of the generative model is restrained on the calculation principle, and precise decoupling and dynamic evolution of unstructured semantics and structured logic are achieved.

The invention discloses an intelligent Chinese text error correction method integrating spelling and semantics, and relates to the technical field of data processing. Comprising the following steps: S1, according to an original Chinese text to be corrected and a correction intention instruction, obtaining an original candidate correction set of each agent through parallel operation of each agent; s2, processing the original candidate correction set according to the correction intention instruction, and obtaining the score of each original candidate correction item; and S3, comparing the original candidate correction items with the comprehensive evaluation scores through a diversity optimization screening algorithm, and determining a final candidate correction item set. According to the method, the candidate correction set is generated in parallel through multiple agents and is combined with the conflict resolution and multi-dimensional evaluation mechanism, so that the accuracy of error correction can be ensured, diversified expression selections can be provided, and the problems of single and rigid output of a traditional error correction system are avoided.

The invention discloses a knowledge question-answering method and system based on topic knowledge graph retrieval enhancement, and the method comprises the steps: firstly extracting a local topic represented in a triple form based on an original document through employing a large language model, carrying out the clustering, and generating a global topic triple set representing the global perspective of the whole document; secondly, on the basis of the global topic triple set, topic-guided entity and relation extraction is adopted, and a mixed knowledge graph is constructed; secondly, providing a semantic perception personalized PageRank algorithm, matching query semantics with semantics of edges in the mixed knowledge graph, and dynamically adjusting the weight of score propagation between nodes; and finally, designing a three-level progressive retrieval mechanism, retrieving multi-level information related to user query from the mixed knowledge graph, and inputting the multi-level information into the large language model to generate a final answer. According to the method, the semantic integrity and retrieval precision of the knowledge graph are remarkably improved, and the accuracy, comprehensiveness and enabling performance of generated answers are ensured.

The invention discloses a method and system for extracting credible evidence of a document based on a large model, and the method comprises the following steps: carrying out the complete reasoning of the evidence of a damaged document: constructing an image and text double-branch model, capturing the edge texture of a damaged region of the document through an improved structure of a fusion cross-modal attention mechanism, and carrying out the image and text double-branch model; generating a pixel-level prediction mask of the missing region; according to the text branch, on the basis of BERT, a domain term embedding layer is added, mask language modeling is carried out on an incomplete text through a derivative model which is subjected to document evidence corpus fine adjustment, and candidate completion content is generated in combination with context semantics and a domain term library. By combining improved physical damage feature extraction with a domain adaptive semantic completion model, the integrity and accuracy of damaged document evidence extraction are remarkably improved, and dual optimization of damaged region pixel-level repair and semantic logic coherence is realized.

The invention discloses a text and image fused online comment toxicity detection and filtering method and system, and relates to the technical field of natural language processing, and the method comprises the steps: extracting a semantic vector of a comment text through a RoBERTa-Marge model; the visual features of the image are extracted through an OfficientNet-V2 model; aligning heterogeneous modal features by adopting a double-flow contrast loss function; self-adaptive decision making of culture sensitivity: loading a regional sensitive rule table according to a user IP address, and dynamically adjusting symbolic semantics; calculating an intimacy correction factor based on the social relationship between the publisher and the receiver; and performing context weighted toxicity scoring, calculating a user historical behavior weight, and outputting a final toxicity probability. According to the method, a deep dynamic mapping mechanism of text and visual features is constructed, heterogeneous features are extracted through RoBERTa-Large and OfficientNet-V2 double-flow architectures, semantic space alignment is forced by utilizing comparative learning, and the problem of image-text splitting detection in the traditional technology is solved.

The invention relates to the technical field of computer application, in particular to an industry document generation method based on a large language model and retrieval enhancement, which comprises the following steps: acquiring a first user demand of a building engineering project to be input, performing data preprocessing on the first user demand, and acquiring the preprocessed first user demand; performing semantic alignment on the preprocessed first user demand by adopting a cross-modal embedding model to generate a multi-modal fusion vector; according to a dynamic semantic slicing strategy, performing slice segmentation on the multi-modal fusion vector, and obtaining a plurality of slice units containing complete semantics as segmented multi-modal fusion vectors; inputting the segmented multi-modal fusion vector into a pre-trained special retrieval enhancement large model for the building industry, and generating an industry document which conforms to a building engineering specification and a first user demand; according to the method, the problems that the generated building related document content in the building industry is low in professionality and accuracy and cannot meet the industry requirements are solved.

The invention relates to an automatic program repairing method based on data flow driving, which takes an error file and a method signature as input, and reports error information in a defect positioning tool or an integrated development environment. Secondly, handing over error information and original defect codes to a large language model, extracting an element list related to defects from the large language model, then constructing defect contexts, using a search tool to retrieve key element definitions and dependency relationships, integrating control flow and error mode information, and finally obtaining defect information; then, the retrieved information, the original defect codes and the potential defect types are integrated into cue words, and the cue words are submitted to a large language model to generate candidate patches. And then the candidate patch is verified, and if the candidate patch does not pass the verification, error reporting information during the verification is combined with the candidate patch to carry out iterative optimization until the candidate patch successfully passes all test verification or the maximum number of iterations is reached. According to the method, context semantics are enhanced by means of data flow analysis, and the repairing accuracy and efficiency are improved.

The invention relates to a multi-modal virtual human real-time generation method based on an expression control embedding space, and belongs to the field of artificial intelligence. According to the method, an expression control embedding space is constructed and used for fusing voice semantics, a rhythm structure and multi-dimensional emotion information, and continuous and controllable multi-modal driving vectors are generated. The whole system has an end-to-end linkage mechanism from audio input to expression and action output. Semantic features, rhythm structures and emotional states jointly act on generation paths of lip and upper body postures and expression modalities, and all modal features are fused and expressed in a unified control space through a collaborative coding and time sequence alignment mechanism. And finally, a high-consistency and high-fidelity virtual human video is generated in real time through an output scheduling mechanism. The method has remarkable advantages in the aspects of modal fusion consistency, generation expression naturalness and emotion control flexibility, and can be widely applied to key scenes such as virtual human broadcasting, voice interaction agency and meta-universe digital identity construction.

The invention belongs to the technical field of semantic understanding systems, and particularly relates to a semantic understanding system based on a large language model.The semantic understanding system is characterized in that firstly, a data preprocessing module is used for conducting cleaning, denoising and cross-modal conversion on input multi-modal data such as texts and images, and standardized data is generated; a semantic feature extraction module extracts general semantic features by using a pre-training model, adapts to field requirements through dynamic learning rate fine adjustment, and outputs scenarized semantic vectors; then, a dynamic semantic-knowledge bidirectional fusion module adjusts token weight according to a dynamic semantic weight algorithm, realizes real-time alignment of semantics and a knowledge graph by means of a knowledge entity association strength algorithm, and a knowledge enhancement fusion module further optimizes knowledge weight and dynamically updates association; then, the semantic reasoning module performs multi-round reasoning based on fusion information, and evaluates the result reliability in combination with a confidence coefficient algorithm; and finally, the output and optimization module generates a structured result, and iteratively optimizes parameters of each module according to feedback data to complete a semantic understanding processing flow.

The invention relates to the technical field of data processing, and particularly provides an address text correlation analysis method, system and device and a storage medium, and the method comprises the steps: receiving a target address text; based on the semantics of the target address text, retrieving a plurality of candidate standard address texts from a pre-constructed address knowledge base; on the basis of the target address text and the retrieved candidate standard address text, a cue word is constructed, and the cue word comprises an instruction used for guiding a large language model to conduct address relevance judgment and context information; and inputting the cue word into a large language model trained in an address field, and obtaining and analyzing the output of the large language model to determine a standard address text associated with the target address text and the association degree thereof. According to the method, the domain knowledge base and the large language model are fused, so that the accuracy and practicability of address correlation analysis are remarkably improved.

The invention discloses a classroom multi-modal data processing method and system based on deep learning, and aims to solve the composite technical problem of poor classroom multi-modal data fusion analysis effect caused by global asynchronism of a data source, a nonlinear sequential relationship between modals and lack of perception for teaching semantics in the prior art. The method comprises the following steps: acquiring a classroom multi-modal data stream; innovative hierarchical cross-modal time alignment processing is carried out, coarse-grained global alignment is carried out firstly to correct initial time migration between devices, then fine-grained local alignment is carried out, and the core of the processing is that an innovative dynamic time warping algorithm is adopted. Higher alignment weights are given to semantic key segments such as teacher explanation key points and teacher-student interaction, and nonlinear alignment conforming to the teaching rhythm is achieved; and finally, the aligned features are sent to a fusion network to generate a unified fusion feature vector.

The invention discloses a question and answer knowledge base updating method and device, equipment and a medium, and the method comprises the steps: inputting a platform help document into a large language model when it is monitored that the platform help document is updated, and controlling the large language model to generate a plurality of candidate question and answer entries; calculating the similarity between every two standard questions in the plurality of candidate question and answer items and standard questions in reference question and answer items stored in a question and answer knowledge base, and when the similarity meets a preset newly-added condition, storing the corresponding candidate question and answer items in the question and answer knowledge base; when the similarity does not meet a preset newly-added condition, calling a large language model to judge whether the full-text semantics of the corresponding candidate question entries and the reference question and answer entries are the same, and if not, storing the corresponding candidate question entries into a question and answer knowledge base; and if yes, marking the corresponding candidate question item as a to-be-audited state. According to the method, the updating efficiency and accuracy of the question and answer knowledge base are remarkably improved, and the user experience of an intelligent customer service system is enhanced.

The invention provides a method and system for constructing a Neo4j knowledge graph based on an LLM natural language, and belongs to the technical field of smart tourism and knowledge graphs. The method comprises the steps that S1, text entities are recognized, and semantics are understood; s2, analyzing an entity relationship and generating a Cypher statement; s3, extracting entity attributes and performing constraint verification; s4, constructing a Neo4j knowledge graph; and S5, performing dynamic updating and quality control. According to the method, the natural language understanding ability of the large language model and the relation modeling advantage of the Neo4j graph database are fused, so that automatic construction, dynamic updating and culture compliance verification of the tourism knowledge graph are realized.

The invention provides a code reuse method and device based on AI drive, equipment and a medium. When a code submission event is detected, a code scanning process is triggered, a code analysis module driven by AI carries out multi-dimensional analysis on submitted codes, quantitative evaluation is carried out on the basis of a preset reusability evaluation model, and code snippets meeting a reuse standard are stored in a code knowledge base with a semantic index structure; analyzing annotation semantics based on a natural language processing technology through an AI retrieval engine, calculating semantic similarity between demand description and code snippets in a knowledge base in combination with a deep learning model, and positioning the code snippets which are optimally matched; after code multiplexing execution, performing incremental modification on the multiplexing code snippets and automatically generating change records; and carrying out value layering on the code snippets based on the Pareto analysis principle by using frequency trend prediction. Through combination of the AI technology and code reuse, the development efficiency can be remarkably improved.

The invention relates to the technical field of knowledge maps, and discloses a knowledge map completion method based on semantic-structure multi-level fusion. According to the technical scheme, for a training set positive example triple, a structured rationality score and a semantic correlation score are fused to screen high-quality training samples; in order to complement a query search relation path, screening a guiding path through a uniqueness index; integrating the query text, the entity description, the neighbor facts and the guide path to construct an enhanced input prompt; dynamic structure embedding is generated by using a relational graph convolutional network, the dynamic structure is mapped by an adapter module and then injected into a large language model, and completion prediction is completed by combining a fusion input fine tuning model. According to the method, through multi-level fusion of semantics and structural information, intelligent sample screening, prompt enhancement construction and structural information dynamic injection are achieved, the accuracy, reasoning ability and training efficiency of a large language model in a knowledge graph completion task are effectively improved, and the method is especially excellent in performance in complex relation reasoning and inductive completion scenes.

The invention discloses an intelligent legal contract generation method and device based on a contract template and contract term semanteme, and belongs to the technical field of intelligent contracts, and the method comprises the steps: analyzing a legal contract text in combination with a plurality of natural language processing technologies, and constructing a contract meta-model; dynamically generating an intelligent legal contract template based on the contract meta-model; an executable intelligent legal contract code is generated through the contract template; and finally carrying out consistency check on the generated code and the original contract. Therefore, the problems of unclear structure, difficulty in semantic mapping, poor universality, law compliance and the like in the traditional process of converting the contract into the intelligent contract are solved, structured expression of the law text, automatic generation of the contract code and logic consistency verification are realized, and the generation efficiency and reliability of the intelligent law contract are improved.

The invention discloses a multi-modal fusion art teaching management system and a method thereof, and relates to the field of artificial intelligence and education informatization. The system comprises a modal feature extraction module, a context cue generation module, a context guide fusion module, a semantic alignment optimization module, a teaching feedback output module and a cue updating module. The system firstly performs modal feature extraction on work images submitted by students and teacher comment texts, and encodes the work images and the teacher comment texts into semantic vectors; generating a context prompt semantic vector based on a teaching task and a user behavior, guiding image-text semantics to perform two-stage fusion, and constructing a fusion semantic expression; a semantic alignment effect is optimized through a structure perception loss function; and multi-dimensional teaching feedback content is generated based on the fused semantic vector, prompt words are dynamically updated according to a feedback result and user behaviors, and self-adaptive optimization of the system is realized. The image-text fusion quality, the feedback generation precision and the system adaptability are improved, and the method is suitable for an intelligent art teaching scene.

The invention is suitable for the technical field of natural language processing, relates to a hierarchical memory and context awareness retrieval method of a large role model and a related product, and aims to solve the problems of limited model memory duration, insufficient retrieval correlation and insufficient personality consistency in a long dialogue. According to the invention, a short-term-middle-term-long-term three-level memory architecture is adopted, and a memory attenuation and migration mechanism is combined, so that dynamic metabolism of memory is realized; related memories are recalled accurately through a context semantics and role personality double-sensitive double-stage retrieval algorithm; relying on a personality-linked memory fusion and response generation strategy, the reply is ensured to fit personality setting; and a closed-loop adaptive learning mechanism of dialogue-memory-retrieval-generation-feedback is constructed, and the memory quality is continuously optimized. According to the method, the role large model can have the human-like continuous memory ability, the continuity, retrieval accuracy and personality consistency of long dialogues are remarkably improved, and the long-term personalized interaction requirements of scenes such as digital personality assistants and dialogue agents are met.

The invention discloses a multi-modality-based minority non-abandoned pattern knowledge graph construction method and system, and relates to the technical field of cultural heritage digital protection. Deep association of multi-modality knowledge is realized through a double-path entity relationship extraction mechanism, context semantics are coded by a text path by utilizing a pre-training language model, and a multi-modality knowledge graph is constructed; the method comprises the following steps: accurately extracting entities such as a pattern and an inheritor, a semantic relationship and a visual path, analyzing a pattern topological structure through a graph convolutional network, converting visual features such as lines and contours into structured relationship data, calculating cosine similarity of a text and a visual feature vector through comparative learning, establishing cross-modal mapping of the visual features and culture description, and obtaining a visual feature model; according to the mechanism, the knowledge graph simultaneously contains semantic logic and visual feature association, and construction of a complete knowledge chain from a pattern form to cultural connotation is realized.

The invention belongs to the technical field of artificial intelligence task planning, and provides a task planning PDDL file automatic generation method based on natural language input, and the method comprises the steps: (10) constructing task environment description words: analyzing environment elements and participants, and designing a diversified scene framework; (20) building a knowledge enhancement fine tuning model: fusing a vector RAG knowledge base retrieval result and pre-training model parameters by the model, and inhibiting logic illusion; (30) PDDL file generation and knowledge constraint: driving the fine tuning model to output a correct PDDL file in combination with scene semantics and knowledge base rules; (40) multi-stage dynamic evaluation: executing a feasibility index verification instruction through a knowledge base rule matching degree; and (50) model closed-loop iterative optimization: updating RAG knowledge base content and model parameters based on execution feedback, and constructing a closed-loop iterative optimization model. The method has the beneficial effects that the instruction logic deviation rate is reduced through RAG knowledge base constraint, the complex scene instruction generation accuracy is improved, and the dynamic decision response time is shortened.