174 results about "Self-optimization" patented technology

The invention provides an intelligent question and answer method based on a preset multi-dimensional knowledge base and a large language model, and the method comprises the steps: receiving a natural language query, converting the natural language query into a query semantic vector, and recognizing a corresponding intention type and a first confidence coefficient; if the first confidence coefficient is higher than an intention threshold value and the highest similarity between the query semantic vector and a standard question semantic vector in a preset multi-dimensional knowledge base is higher than a matching threshold value, returning a standard answer; otherwise, retrieving a generated context based on the query semantic vector, fusing the user query and the generated context, inputting the fused user query and generated context into a large language model to generate a preliminary answer, checking fact consistency of the preliminary answer and a standard answer in the generated context to generate a credibility score, and obtaining a target answer when the credibility score exceeds a credibility threshold; and dynamically optimizing the system based on feedback data of the user to the target answer. The intelligent customer service system with controllability, flexibility and self-optimization capability is realized, and the semantic understanding depth and response accuracy of the system are improved.

The invention relates to the technical field of communication, and discloses a three-dimensional radio environment map construction method and system, a terminal and a storage medium, and the core is to construct a'base station unmanned aerial vehicle 'bidirectional interaction closed-loop optimization framework to realize efficient construction of a high-precision map. Self-adaptively fusing sparse radio measurement data and environmental building structure features; introducing a confidence evaluation mechanism based on adversarial learning and position weighting, and generating a pixel-by-pixel confidence map; an intelligent planning method based on a trajectory diffusion model is designed, local perception constraint and long-term information gain are cooperated with a classifier-free guide mechanism, and an optimal trajectory considering both safety and sampling efficiency is generated; and a continuously self-optimized closed-loop system is formed through newly acquired data of the unmanned aerial vehicle and periodical updating of the model. According to the invention, a high-reliability technical basis is provided for applications such as urban air communication and spectrum resource management.

The invention discloses a defect detection system and method for a glass bottle, and belongs to the field of industrial visual defect detection and analysis. The system comprises a GBDM module, the GBDM module is a defect detection module, and the GBDM module comprises a front attention enhancement structure, a rear attention enhancement structure and a defect detection structure, a multi-branch feature extraction structure is used for generating four feature branch diagrams; and the feature fusion structure is used for fusing the four feature branch diagrams. The GBDM module is embedded in the backbone network of the YOLOv8 network model, and multi-branch structures such as wavelet transform, Gabor direction filtering and specular reflection suppression are utilized, so that the feature extraction capability of multiple types of complex defects such as cracks, smudginess and damage on the surface of the glass bottle is enhanced, the generalization capability of the model in a small sample scene is remarkably improved, and the accuracy of the model is improved. The real-time requirement of an industrial scene is met, continuous self-optimization is achieved, and the method has high robustness, high accuracy and high application value.

The invention relates to a thermal power plant fault early warning and diagnosis method based on a nebula system, and the method comprises the steps: collecting the multi-dimensional operation time sequence data of a thermal power plant, carrying out the feature extraction and lexical element processing of the multi-dimensional operation time sequence data through an encoder, and obtaining a unified equipment state lexical element sequence; based on the equipment state lexical element sequence, constructing a dynamic star map representing the operation state of the whole power plant; inputting the dynamic star map into a space-time fusion backbone network; the space-time fusion backbone network performs iterative processing on the dynamic star map and generates a health degree attenuation trajectory; when the slope of the health degree attenuation trajectory exceeds a preset threshold value, dynamic early warning and system diagnosis are triggered, and a natural language diagnosis report containing a causal reasoning chain is generated; new multi-dimensional operation time sequence data are collected in real time, and an incremental learning algorithm is used to update the encoder and the space-time fusion backbone network online; compared with the prior art, the system can continuously adapt to working condition changes and has high self-optimization capacity.

The invention discloses a three-dimensional temperature field online reconstruction system based on receiving and transmitting integrated sound wave sensing, and relates to the field of boiler monitoring. Comprising an information acquisition module, an emission control module, a reconstruction sensing module, a path optimization module, a sound wave receiving module, an amplitude limiting filtering module, a decoding separation module, a feature extraction module, a physical simulation module, a multi-source fusion module, a three-dimensional reconstruction module, a state monitoring module and an online calibration module. According to the method, the transmitted signal can still keep high identifiability in a high-noise and strong-interference background, the credibility of subsequent inversion and data assimilation is improved, a sensing network can be self-optimized, the observation precision and the space coverage capability are improved, a wide dynamic range and high linearity are realized, a simulation result is closer to a real working condition, and the method is suitable for large-scale popularization and application. And the reliability of temperature field inversion is obviously improved.

The invention provides a dynamic updating method and system of a knowledge graph based on reasoning enhancement. The method belongs to the cross technical field of artificial intelligence, knowledge engineering and dynamic system modeling. The method comprises the following steps: performing multi-dimensional feature extraction on original knowledge data to generate a knowledge feature vector set; constructing an initial knowledge graph structure based on the knowledge feature vector set, and defining an initial association rule between knowledge nodes to form a basic knowledge graph model; building a multi-order causal inference engine according to the basic knowledge graph model, performing deep mining on potential causal relationships among knowledge nodes, and generating a knowledge causal relationship network; and performing confidence evaluation on each causal chain in the knowledge causal relationship network to obtain the knowledge causal relationship network subjected to quality verification. Based on a multi-order causal reasoning engine, the system can mine and verify potential causal relationships among knowledge nodes in real time, and continuous updating and self-optimization of the knowledge graph are ensured.

The invention discloses an offshore regional communication system, and relates to the technical field of offshore aquaculture platform communication, the offshore regional communication system comprises a multi-level communication platform, a distributed adaptive networking engine and a cross-domain cross-medium communication gateway, the multi-level communication platform is integrated with a satellite communication module, a water surface communication module and an underwater communication module, a local edge calculation function is realized; the distributed self-adaptive networking engine is in communication connection with the multi-level communication platform and is used for carrying out neighbor discovery, global topology convergence, dynamic routing calculation and hub automatic switching on the multi-level communication platform; and the cross-domain cross-medium communication gateway is in communication connection with the multi-level communication platform and the distributed self-adaptive networking engine, and is used for performing signal format conversion and intelligent routing among the satellite communication domain, the water surface communication domain and the underwater communication domain for the multi-level communication platform. The communication system provided by the invention can realize cross-domain cooperation of satellite, water surface and underwater communication, and has self-sensing, self-repairing and self-optimizing capabilities.

The invention provides an AI auxiliary code generation method and system applied to a low-code platform, and relates to the technical field of low-code platforms.The AI auxiliary code generation method comprises the steps that firstly, component demand scenarized information is obtained and analyzed to obtain a scenario feature set, and a dynamic evolution rule set is generated in combination with platform real-time metadata; calling an AI code generation module, inputting the scene feature set and the rule set, and generating an initial component code; then deploying the initial code to a test environment, collecting and analyzing operation data, generating an iterative optimization instruction adjustment code and a rule set, and obtaining a target component code; and finally, embedding the target code into a platform specified function module to complete code generation. The method can adapt to diversified scene requirements, has a self-optimization capability in combination with the real-time state of the platform, and can improve the development efficiency and code quality of the low-code platform.

The invention relates to the technical field of building safety, and discloses a safety transformation monitoring system for building construction and a method thereof. The system comprises a data acquisition and standardization module, a dynamic security entity relationship map construction module, an organization capability and cognitive load modeling module, a hybrid risk identification and conduction calculation engine, a self-adaptive security transformation strategy generation engine and a closed-loop self-calibration module. According to the method, a dynamic security entity relation graph is constructed and updated in real time, the cognitive load of management personnel is quantified, and the cognitive load is used as a dynamic adjustment factor of the conduction probability of risks on the graph; when a safety transformation strategy is generated, comprehensively evaluating a risk reduction effect, resource cost and additional cognitive load cost to output an optimal adaptive strategy; and finally, performing closed-loop self-calibration on the system model by utilizing execution feedback. According to the method, dynamic prospective risk prediction, self-adaptive strategy generation and continuous self-optimization are realized.

The invention relates to the technical field of fault detection, and provides a fault identification method and a parallel processing system, through organic combination of four steps, a complete system from data acquisition, feature extraction, fault diagnosis to fault trajectory prediction is constructed, comprehensive and accurate diagnosis and detection of equipment faults are realized, and the fault detection efficiency is improved. According to the system, fault information in voiceprint signals can be fully mined, the influence of various factors on fault diagnosis is comprehensively considered, the accuracy and reliability of fault diagnosis are greatly improved, and a powerful guarantee is provided for safe operation of equipment; all the steps are mutually associated and mutually influenced, self-optimization can be continuously carried out according to the real-time operation state of the equipment and newly obtained voiceprint data, and the recognition performance of fault detection is improved.

The invention discloses a hydropower station generating capacity prediction method and system based on knowledge graph guidance, and relates to the technical field of hydropower station generating capacity prediction. Comprising the following steps: S100, in a power generation capacity prediction process, establishing a threshold trigger audit baseline, recording an operation environment state, a knowledge graph edge weight change parameter and a prediction error symbol during prediction deviation triggering each time, generating a trigger fingerprint containing a time identifier, and storing the trigger fingerprint; according to the method, through whole-process auditing, triggering fingerprint tracing, gain analysis and risk label identification, weight dynamic limitation, double-track comparison and closed-loop correction, accurate control and self-optimization of abnormal dependence of the model are realized, the accuracy and the stability of power generation capacity prediction are improved, the correction and protection capability on abnormity is enhanced, and the power generation capacity is improved. And the risk of error accumulation and dependence on curing is effectively avoided.

The invention relates to a multi-mode quality inspection system based on industrial valve production, in particular to the field of industrial intelligent manufacturing, and can significantly improve the accuracy and reliability of industrial valve quality inspection by setting a space-time alignment module, a feature fusion module, a fusion strategy optimization module and a quality decision and model updating module. Various fine defects of the valve can be accurately identified through efficient fusion and intelligent analysis of multi-modal data; the self-adaptive optimization mechanism ensures the stable performance of the system under different production working conditions, effectively reduces the missed detection and misjudgment risks, guarantees the product quality, reduces the production waste and potential safety hazards, prolongs the effective service period of the system through continuous self-optimization, and achieves the efficient and high-quality closed-loop management in the intelligent manufacturing process.

The invention provides an intention-driven semantic evolution mechanism based on a tensor semantic field and an application system thereof, and relates to the technical field of artificial intelligence, semantic networks and cognitive computing. According to the method, five types of semantic primitives including data, information, knowledge, intelligence and intention are expressed as high-order tensor nodes, a semantic tensor field network is constructed, and dynamic evolution and intention driving of semantics are achieved. The system comprises a multi-scale semantic aggregation mechanism, an intention weight diffusion algorithm, a semantic tensor evolution operator and a white box interpretation interface, supports full-link semantic processing from original data to high-level wisdom to intention constraint, and overcomes the defects in semantic representation and evolution, intention fusion and system interpretability in the prior art. And the generative AI system has stronger intention perception, semantic self-optimization and process transparency capabilities, and is suitable for applications such as a semantic perception large model platform, an AI cognitive map system and an interpretable language generator.

The invention relates to the technical field of AI computing power scheduling, in particular to an AI computing power sharing scheduling system based on a cloud native architecture, which comprises a data acquisition and convergence module, a resource portrait construction module, a resource efficiency optimization module, an operation and maintenance self-healing module and a closed-loop feedback module, the data acquisition and convergence module is responsible for acquiring cluster dynamic operation data and AI task life cycle data; a resource portrait construction module respectively constructs resource portraits for the computing nodes and the AI tasks based on the data; the resource efficiency optimization module executes a resource optimization decision through a Kubernetes scheduler and a cluster management queue according to the resource portrait; the operation and maintenance self-healing module executes a predictive operation and maintenance decision through a Kubernetes API based on the resource portrait; and the closed-loop feedback module feeds back the execution result of the optimization and self-healing decision as new data to the front-end module, so that continuous updating of the resource portrait and self-optimization of the system are realized, and a complete intelligent closed loop is formed.

The invention discloses an AI knowledge base continuous optimization system and method based on man-machine cooperation and closed-loop feedback. The continuous optimization system for the AI knowledge base comprises a multi-mode co-emotion knowledge input module, a knowledge active discovery and construction module, a layered depth inference engine, a response generation module oriented to a specific scene, an output formal verification and interpretability analysis module, a dynamic resource scheduling and reflection module and a man-machine collaborative feedback interface. A feedback data deep analysis module, a knowledge base dynamic update and conflict resolution module and a model increment optimization and security alignment module. Through the mode, the systematic system which has the capabilities of active learning, deep reasoning, self-verification and efficient man-machine cooperation and can realize continuous self-optimization of the knowledge base is provided.

The invention discloses a stability control method for a tunnel adjacent to a water-rich soft rock foundation pit based on artificial intelligence. Comprising the steps of prior prediction based on a finite element model, arrangement of monitoring sensors, construction and training of a neural network model, scheme safety verification, initial decision based on the neural network model, initial grouting scheme execution, circulating intelligent decision based on real-time monitoring and online incremental learning. Training data are generated by establishing a three-dimensional finite element model containing a water-soil coupling mechanism, a neural network model capable of mapping a complex nonlinear relation between monitoring parameters and an optimal grouting scheme is constructed and trained, dynamic decision making is carried out, the grouting scheme subjected to safety verification is output, and the model is continuously optimized; according to the method, whole-process closed-loop intelligent control from intelligent prediction, real-time decision making and safety control to self-optimization is achieved, and the guarantee capacity and the risk management and control level for the stability of the adjacent tunnel when foundation pit excavation is conducted in the water-rich soft rock complex stratum are remarkably improved.

The invention provides an intelligent service execution method based on a dialogue mechanism, and the method comprises the steps: receiving multi-modal input data of a user, and carrying out the preprocessing of the multi-modal input data of the user; a multi-modal graph network is constructed hierarchically to realize cross-modal reasoning of user intentions; constructing a historical attention weight, associating historical request data of the user, and optimizing an answer result of the current dialogue; constructing a task hypergraph to dynamically disassemble the complex task and planning a specific execution path between the subtasks; designing a feedback mechanism based on reinforcement learning to realize self-optimization of the system; interaction logs of the user and the system are maintained, and personalized services are achieved. Through multi-modal data fusion, dynamic task planning and historical feedback optimization technologies, the request understanding accuracy, the task execution efficiency and the system adaptive capacity are remarkably improved, and efficient and flexible service processing support is provided for an intelligent dialogue system.

The invention relates to the field of data security, in particular to a network security intelligent detection method based on big data, and the method comprises the steps: collecting network flow data, a terminal system call sequence and a user operation behavior log, and carrying out the data fusion processing to generate a unified behavior event flow; selecting a key behavior event based on an information entropy threshold value, performing time alignment through a dynamic time warping algorithm, and constructing a behavior gene map containing a communication association gene, an operation sequence gene and a behavior time sequence gene; a dynamic behavior baseline model is established by using unsupervised learning, and gene mutation detection and alarm are realized by calculating the deviation degree of each gene dimension; the detection performance is evaluated based on the false alarm rate, model parameters are optimized through a negative feedback mechanism, and acknowledged attack features are stored in a sharable threat gene feature library through a positive feedback mechanism. According to the method, the detection accuracy is continuously improved through a closed-loop learning mechanism, and a self-adaptive safety protection system with self-optimization capability is constructed.

The invention discloses a cascade large and small model architecture for construction scheme generation and an intelligent generation method, and relates to the crossing field of artificial intelligence and building construction technologies, and the architecture comprises an input and scheme structured definition module which is used for converting user demands into structured tasks; the multi-modal data sensing and understanding module is used for analyzing design drawings and documents; the intelligent knowledge retrieval and enhancement module retrieves related information from each knowledge base based on the RAG technology; the central scheduler is used for task decomposition and scheduling; the professional small model cluster comprises a structure safety checking calculation model, an intelligent drawing model, a construction period optimization model and the like and is used for executing professional calculation and generation; the result aggregation and scheme generation module is used for integrating all results to form a final scheme; and the output and feedback learning module is used for outputting a scheme and realizing system self-optimization. According to the cascaded large and small model architecture for construction scheme generation and the intelligent generation method, intelligent, efficient and high-quality generation of the construction scheme is realized.

The invention discloses an underwater acoustic sensor network clustering method, system and device and a medium, and relates to the technical field of underwater acoustic communication, and the method comprises the steps: obtaining the current network state characteristics of each node of a target underwater acoustic sensor network, inputting the current network state characteristics into an underwater acoustic sensor network clustering model, and obtaining a current optimal clustering scheme; based on the current optimal clustering scheme, electing cluster head nodes through a dynamic role rotation mechanism, and generating an adaptive network topology structure; performing performance evaluation on the adaptive network topology structure to obtain a group of performance indexes; and if the performance index does not meet the performance threshold value, or the confidence coefficient of the current optimal clustering scheme output by the underwater acoustic sensor network clustering model is lower than a confidence coefficient threshold value, starting a complete optimization learning process. According to the technical scheme, the technical problem that it is difficult to balance multi-target performance online and adaptively and carry out continuous self-optimization to adapt to the dynamic underwater environment through an existing clustering method is effectively solved.

The invention provides a data processing method, electronic equipment and a vehicle, and relates to the technical field of data processing, and the method comprises the steps: obtaining the correlation evaluation data of each stored memory data, carrying out the multi-dimensional value evaluation of the memory data according to the correlation evaluation data, and determining a value evaluation score. And determining a corresponding forgetting strategy according to the value evaluation score, and performing storage management on the memory data according to the forgetting strategy. Different forgetting strategies are adopted for storage management of data with different evaluation values, high-value data are reserved, low-value data are deleted or compressed, part of storage space is released, and continuous expansion of the storage space is avoided. In the subsequent data retrieval process, the retrieval efficiency of the memory data can also be improved, and continuous reduction of the system efficiency is avoided. Through dynamic storage management of the memory data in the method, self-optimization and evolution of the memory management system are realized, and the requirements of users for data management are met.

The invention relates to the technical field of intelligent control, and discloses an edible mushroom cultivation environment multi-parameter intelligent cooperative regulation and control system and method. A distributed sensor array; a distributed actuator; the processor is provided with a data acquisition and preprocessing module; an optimal environment trajectory generation module; a multi-agent collaborative execution module; and a state evaluation and refining module. The method comprises the following steps: acquiring and processing environment data to generate a state vector; generating a dynamic target trajectory by using the generative adversarial network; controlling an actuator to cooperatively track the trajectory based on multi-agent reinforcement learning; and the execution entropy is calculated according to the execution process, so that a dual feedback loop is formed to restrain trajectory generation and adjust coordination actions. According to the method, through prospective track generation, deep collaboration and closed-loop self-optimization, the problems that a traditional method is poor in adaptability and collaboration and difficult to self-optimize are solved, and the intelligence and precision of edible fungus cultivation environment regulation and control of multiple parameters are improved.

The present application relates to a kind of intelligent interview evaluation and feedback system based on multi-modal data fusion, specifically relates to data processing field, by meta-learning mechanism dynamic perception interview scene and generate initial fusion weight, subsequently utilize the complex interaction relationship between modalities modeled by graph neural network to carry out fine-grained correction to weight, so as to significantly improve the accuracy and scene adaptability of multi-modal evaluation, further introduce reinforcement learning, link evaluation decision and long-term performance of talents, continuously optimize weight generation strategy, ensure that evaluation standard and business goal are aligned, finally, through closed-loop iteration mechanism, make the whole scheme can be updated automatically according to new data and performance feedback, continuous evolution, with strong self-optimizing ability and long-term robustness, realize the fundamental change from static rule to dynamic intelligent decision.

The application provides a self-evolution system for research report generation, comprising an Agentic RAG module and a multi-agent dynamic evaluation module; the Agentic RAG module is used for actively planning a retrieval strategy through a large language model, intelligently retrieving and screening the most relevant historical reports and experiences from a historical task report and experience library according to the current dialogue context of a user; the multi-agent dynamic evaluation module is used for performing multi-dimensional and interpretable automatic scoring on the retrieved historical reports and corresponding experiences, generating fine-grained and high-reliability reward signals to drive the system to realize self-evolution without training. The system can realize continuous self-optimization based on user interaction history without any parameter update or external training, significantly reduce the illusion rate, improve the factuality, structure and professionalism of the generated content, and is suitable for financial, scientific research, policy analysis and other scenes.

The invention relates to the technical field of license plate recognition, and discloses a closed-loop license plate recognition method and system based on a self-adaptive data set. The method includes the steps that S01, a license plate area is defined in an original image of a target vehicle; s02, segmenting the license plate area into character images; s03, marking a target character string from the license plate character string; s04, a matching quality score is calculated, and whether data set updating operation is triggered or not is judged; s05, updating the preset data set; s06, outputting license plate recognition information of the target vehicle; according to the invention, the linkage effect of comparison of the preset data set and updating of the preset data set can be realized, so that the license plate recognition system has the capabilities of dynamic learning and self-optimization, and the coverage range of the preset data set is ensured to be continuously expanded along with the passage of the timeline; the robustness and generalization ability of the license plate recognition system in coping with various interference factors are improved, and the adaptability and recognition precision of the license plate recognition system are enhanced.

The application discloses a body intelligent agent interaction system and a control method, and relates to the technical field of artificial intelligence. The method comprises the following steps: a terminal sensor module is configured to collect physical signals of environment or user interaction, and generate standardized event data; a reflection processing module comprises a lightweight decision unit, which is used for receiving scene labels and user feedback information to self-optimize protected core behavior logic, and deep evolution is performed after a growth authorization signal is received; a narrative processing module is used for processing complex interaction cognition, generating scene labels and sending the scene labels to the reflection processing module, sending an inhibition instruction and a growth authorization signal to the reflection processing module, and performing context interpretation; wherein, the reflection processing module and the narrative processing module communicate through a preset interface, and the narrative processing module has no right to directly modify the core behavior logic; the system guarantees the explanation right and adjustment right of the narrative layer to the reflection layer, and prevents tampering of the core instinct at runtime through strict permission isolation.

The invention belongs to the technical field of natural language processing and information retrieval, and relates to an intelligent question answering method based on structured semantic indexing and double-layer memory enhancement. The method comprises six steps of query preprocessing, Agent-based multi-tool dynamic routing, structured and semantic enhanced index construction, self-adaptive context assembly, double-layer memory management, and data closed loop and self-evolution, a differential segmentation strategy is adopted for texts, codes and multi-modal data, query optimization, intelligent tool routing and a mixed retrieval algorithm are combined, and the multi-modal data are subjected to self-adaptive context assembly. The technical defects of semantic rupture, low retrieval accuracy, no long-term memory and lack of self-optimization of an existing RAG system are overcome. According to the method, the code retrieval accuracy and the context utilization rate are improved, personalized long-term service and automatic operation and maintenance are achieved, and the method is suitable for scenes such as technology research and development, code library maintenance and intelligent question and answer.

The invention relates to an agent task decomposition and dynamic arrangement system based on reinforcement learning, in particular to the field of reinforcement learning, according to the architecture, firstly, high-quality strategy and reward initialization is achieved through expert demonstration, the problems of cold start and sparse reward are effectively solved, and then simulation and reinforcement learning signals are fused in online operation, so that the real-time performance of the system is improved. Stable refining and rapid adaptation of a strategy are achieved through mixed experience, flexible decomposition and efficient parallel execution of complex tasks are achieved through dynamic collaboration and hierarchical optimization based on an attention mechanism among multiple agents, and finally, a meta-learning mechanism monitors macroscopic performance indexes to achieve the purpose of efficient and efficient parallel execution of the complex tasks. According to the method, the key parameters of each learning and collaboration module at the bottom layer are adaptively regulated and controlled, so that the whole agent group can be continuously self-optimized, and excellent robustness, efficient collaboration capability and long-term performance maintenance and improvement are shown in the face of dynamic and changeable environments and tasks.

The application provides a complex document intelligent extraction method, system and device based on a multi-modal large model and a storage medium, and relates to the technical field of automatic data processing and information extraction. The disclosed technology provides an efficient, accurate and self-optimizing information extraction solution, which includes a combination of technologies: a two-stage extraction strategy using preliminary screening and deep information extraction balances the efficiency and depth of information extraction; a cross-page dependency graph construction mechanism solves the problem of information fragmentation; a context-aware overall information extraction method ensures that the model reasons in a complete visual context; dynamic task decomposition and dimension reduction techniques are also used to address performance challenges caused by complex tasks; and a prompt word self-optimization closed loop based on artificial feedback is used to achieve continuous learning and precision improvement of the system.

The invention discloses an intelligent scheduling control method based on artificial intelligence, and relates to the technical field of intelligent traffic, real-time images are analyzed through a first artificial intelligence model, the current traffic state of a crossroad can be accurately obtained, a solid data basis is provided for subsequent decision making, and the traffic efficiency is improved. Then identifying the current traffic state through a second artificial intelligence model to determine a signal lamp execution phase, performing signal control according to the signal lamp execution phase, and constructing the current traffic state, the signal lamp execution phase, awards and a next traffic state into playback experience; the second artificial intelligence model is updated by adopting playback experience, new experience can be continuously accumulated and self-optimized in actual operation, so that a scheduling strategy can adapt to long-term change of a traffic mode, and the method has strong self-adaption and sustainable evolution capabilities.