5679 results about "Time information" patented technology

The invention discloses a multi-source data fusion method and system based on cloud computing, and particularly relates to the technical field of multi-source data fusion, and the method comprises the steps: obtaining the real-time information of multi-source data, generating a health score and a label through a health evaluation model, and marking a data state and an abnormality type; performing format conversion and semantic enhancement processing on the health data to generate structured data; based on a knowledge graph reasoning technology, extracting a logical relationship to generate a fusion rule, and calculating a rule weight and confidence in combination with a health label, a data source confidence and a semantic feature; a fusion result of conflict data is judged through a semantic consistency index, if the fusion result does not conform to expectation, historical data and rules are loaded according to a verification model, a verification score is calculated, an optimization process is triggered after the verification score is compared with a threshold value, health assessment model parameters, a semantic embedding model and fusion rule logic are optimized, and fusion accuracy and reliability are improved.

The embodiment of the invention relates to the technical field of video processing, and particularly discloses a video timestamp event identification and reasoning method based on a multi-modal large model. According to the embodiment of the invention, an input video is processed, and visual information and text information are extracted; performing interaction and fusion on the visual information and the text information to generate a video event semantic representation; and performing event identification on the semantic representation of the video event to obtain an event timestamp mark and event logical reasoning. According to the method, deep fusion and semantic association understanding of multi-modal information such as vision and texts can be realized, the problem that a traditional method is insufficient in semantic consistency and feature expression ability among modals is solved, and deep fusion of multi-modal features and accurate understanding of semantic association are realized; and the frame-level time information can be used as a visual code and a text feature to perform visual text feature alignment, and then video-level modeling is performed, so that the problem that a traditional model cannot directly capture an accurate time frame when a visual event occurs is solved.

The invention discloses a multi-agent cooperation system and method based on spatial calculation and multi-modal AI fusion, and the method comprises the steps: collecting all real-time data of a construction site through a multi-modal data collection and fusion unit, and carrying out the processing and integration of all collected data through a multi-modal AI algorithm, and generating a unified semantic association model; a three-dimensional virtual environment of a construction site is constructed through a space calculation unit, real-time synchronous mapping is formed for the construction site, and dynamic changes in the construction process are simulated and predicted; task allocation and behavior planning are carried out on multiple agents through an agent behavior management unit; and the intelligent agents are configured to perform real-time information sharing and task cooperation through a communication protocol through the multi-intelligent-agent cooperation unit. According to the scheme, the problems of difficulty in multi-source data integration, low cross-department cooperation efficiency, insufficient construction dynamic adjustment and the like in intelligent construction in the building industry can be solved.

The invention discloses a knowledge base generation and retrieval method based on the RAG technology and an intelligent question answering system, and belongs to the technical field of natural language processing, and the knowledge base generation and retrieval method based on the RAG technology and the intelligent question answering system based on the RAG technology have the beneficial effects that the accuracy and the real-time performance of information processing are improved through efficient combination of retrieval and generation. High-dimensional vector representation is generated through text slicing and vectorization, a vocabulary index is constructed, and hierarchical retrieval and keyword matching are achieved; meanwhile, a question generation and question-question retrieval method is utilized, a result is optimized and generated in combination with a large language model (LLM), the reasoning ability of the large model is further improved through chain reasoning (COT) and a cue word optimization module, and it is ensured that the generated answer has high accuracy and logicality; the method can be widely applied to the fields of intelligent question answering, knowledge base construction, real-time information retrieval and the like.

The invention provides a green agricultural product distribution path optimization and inventory control method, and the method comprises the steps: selecting a distribution path with the minimum agricultural product quality influence degree as an optimal path according to a prediction result of the agricultural product quality influence degree, and transmitting the optimized distribution path to a distribution vehicle terminal in real time; after the distribution vehicle terminal receives the optimal distribution route issued by the cloud, the distribution route is navigated in real time through a vehicle-mounted navigation system, a driver is guided to drive along the optimal route, and the distribution route is dynamically adjusted according to the real-time road condition information; meanwhile, according to the real-time position and the driving speed of the distribution vehicle, the arrival time of the agricultural products is dynamically estimated, and the estimated arrival time information is pushed to a destination customer to remind the customer to make preparation for receiving goods in advance.

Disclosed are an improved GNSS tropospheric delay (ZTD) real-time grid model construction method and system. Firstly, a global tropospheric delay elevation normalization grid model is constructed using global ERA5 reanalysis data from a plurality of years. Then, elevation normalization and grid horizontal interpolation are performed on GNSS real-time ZTDs to obtain a global tropospheric delay real-time grid model. A user only needs to provide position and time information and use the global tropospheric delay elevation normalization grid model and the global tropospheric delay real-time grid model to be able to obtain a ZTD at a target point. According to the present invention, the high spatial resolution of a numerical weather model (NWM) and the high precision of the GNSS tropospheric product are comprehensively used, and an elevation normalization function form is optimized, so that the construction of a tropospheric delay real-time grid model covering a larger height range is achieved, and high-precision tropospheric delay correction information can be provided for GNSS real-time high-precision data processing, thereby improving the data processing performance.

The invention discloses a supply chain intelligent management system based on dynamic collaborative optimization, and relates to the technical field of hotel supply chain management, and the system comprises a supply chain intelligent management platform which is in communication connection with the following modules: a multi-modal data sensing module, a multi-modal data processing module and a multi-modal data processing module. The multi-modal data acquisition module is used for acquiring multi-modal data in a hotel through a LoRaWAN + BLE hybrid sensor network, accessing an external data source and acquiring real-time information through an API (Application Program Interface); and the dynamic demand prediction module captures a time sequence trend through bidirectional LSTM based on an ASTGNN model. According to the method, external dynamic data such as social media public opinions and weather are fused, the multi-modal data analysis technology is combined, the accuracy of demand prediction is remarkably improved, the ASTGNN model is utilized, the time sequence trend and the demand association between the branches are analyzed in combination with bidirectional LSTM and GCN, and the feature weight is dynamically adjusted, so that the demand prediction error rate is greatly reduced, and the demand prediction efficiency is improved. A more reliable demand prediction basis is provided for enterprises, and optimization of inventory management and production plans is facilitated.

The invention relates to a man-machine hybrid autonomous navigation system in an unknown dynamic environment, which comprises an environment sensing module for acquiring environment information by using sensors such as a laser radar, a camera and an inertial measurement unit and performing data preprocessing; the local target point selection module selects a local target point set meeting requirements in the environment according to the real-time information provided by the environment sensing module; the decision planning module is used for generating an optimal path and a navigation strategy based on mixed training of a deep reinforcement learning algorithm and artificial experience; and the learning optimization module improves the generalization ability of the algorithm through online learning and transfer learning technologies, and realizes effective introduction and strategy adjustment of human experience in combination with the man-machine interaction module. The man-machine interaction module can work in cooperation with the local target point selection module so as to dynamically adjust target point selection according to external input and optimize path planning. According to the method, high decision stability can be kept in an unseen environment, the training cost is reduced, and the method is more suitable for autonomous navigation application in the real world.

The invention relates to the technical field of satellite communication, in particular to a high-orbit and low-orbit satellite communication resource scheduling method and system, which collects real-time state information of high-orbit and low-orbit satellites, performs digital simulation of a satellite network based on the real-time information, constructs a satellite network management model and predicts satellite trajectory change. The method comprises the following steps: performing multi-dimensional evaluation on a satellite network management model by using a pre-constructed network evaluation framework, generating cross-orbit link evaluation information, analyzing parameters such as link quality and bandwidth, obtaining current service demand information, analyzing a communication resource allocation strategy in combination with a cross-orbit link evaluation result, and formulating a resource scheduling strategy according to strategy analysis. And a high-orbit and low-orbit satellite communication network is constructed through execution of a hierarchical network management mechanism and a cross-orbit data transmission protocol.

The invention provides an end-to-end automatic driving track evaluation method and system based on a world model, and the method comprises the steps: generating bird's-eye view state real-time information according to multi-mode sensor data obtained in a current automatic driving scene; inputting the aerial view state real-time information into a strategy network to obtain a plurality of real-time track anchor points output by the strategy network; according to the bird's-eye view state real-time information and the real-time track anchor point, constructing a real-time state-action pair, and inputting the real-time state-action pair into a bird's-eye view world model to obtain a predicted state-action pair output by the bird's-eye view world model; according to the real-time state-action pair and the predicted state-action pair, the predicted driving track corresponding to the predicted state-action pair is evaluated, and an evaluation result of the predicted driving track is obtained. According to the method, the predicted driving track can be evaluated more accurately, and the safety and reliability of the automatic driving system are improved.

A multi-degree-of-freedom drive-by-wire unmanned vehicle obstacle avoidance path planning method mainly comprises the following steps that based on environment perception information, surrounding obstacles are analyzed and processed, and an environment safety boundary is established; on the basis of an environment safety boundary, speed sampling is carried out based on dynamic window trajectory planning of hierarchical response, and an optimal local trajectory is calculated and selected; in the driving and obstacle avoidance process, according to obstacle distribution and road condition fluctuation real-time information, the posture of a vehicle body is dynamically adjusted, speed sampling is updated, and path planning is updated. According to the method, the optimal obstacle avoidance path can be planned in real time, so that the unmanned vehicle can quickly respond in a complex scene, and potential collision risks can be avoided in time; the characteristics that the multi-degree-of-freedom drive-by-wire unmanned vehicle is suspended and the vehicle body posture is adjustable are fully utilized, vehicle posture adjustment is fused into an obstacle avoidance strategy, and the feasibility of an obstacle avoidance planning track and the stability of vehicle execution are improved.

The invention discloses a time sequence knowledge graph reasoning method based on a large language model, which comprises the following steps of: performing vectorization expression on query entities and relationships by utilizing an embedded model, and generating vectors containing structural features and semantic information in combination with adjacent entities, relationships and time information; a candidate entity set most relevant to query is screened out from the large-scale entity set by calculating the conditional probability; constructing an event evolution tree, and performing multi-hop sampling on adjacent nodes of a query entity and a candidate entity in a time window to generate a tree structure capable of completely reflecting a historical event evolution path; the method comprises the following steps: designing a structured prompt template, and performing instruction fine tuning on a pre-trained large language model by adopting an LoRA (Low Rank fine tuning technology), so that the model better follows a task instruction to generate a reasoning result. And efficient and accurate time sequence knowledge graph reasoning is realized.

Systems and methods for dynamic knowledge integration in LLM systems including receiving multimodal input data comprising text, image, audio, video, and / or code data, extracting information by processing the multimodal input data through a document processor, storing the extracted information in a dynamic knowledge base, receiving a user query at a query processor, identifying knowledge domains related to the user query using domain-specific agents, retrieving real-time information from the dynamic knowledge base responsive to the identified knowledge domains, integrating the real-time information into the processing of an LLM by a dynamic knowledge integrator, and generating a response using the LLM with the real-time information.

The invention relates to the field of running state monitoring of gas turbines for power generation, in particular to a gas turbine dynamic parameter regression prediction method based on a fusion network, which comprises the following steps: S1, data acquisition: acquiring historical data of a gas turbine running atlas; s2, data preprocessing: performing steady-state data screening and abnormal value elimination on the collected historical data, determining input parameters and output parameters of a prediction model, setting a step length for reading the historical data each time, segmenting the data through a sliding window method, dividing data of a training set and a test set, and the like; s3, the prediction model extracts a spatial feature vector through a convolutional layer in a convolutional neural network CNN, then fully extracts time information through a mode of combining Transform and an LSTM network, and obtains potential spatial-temporal correlation; and S4, training test and storage of the prediction model: training the prediction model through the training data, judging whether the trained prediction model meets the requirement or not through the test result of the test set, and storing the prediction model meeting the requirement.

The invention discloses a Transform-based traffic flow prediction method and device capable of sensing a local-global time-space relationship, and the method comprises the steps: firstly obtaining the historical traffic data of all to-be-predicted road sections, the historical data comprising the traffic flow feature data of the predicted road sections, the static adjacent matrix data between road network nodes, and the sampling timestamp data of the historical data; secondly, a spatio-temporal information embedding layer is constructed to provide multiple types of embedding input for a model trunk, the learning ability of the model is enhanced, and three different types of embedding are respectively historical data information embedding, time information embedding and space node self-adaptive embedding; then, constructing a local-global time dependence extraction module, respectively learning short-time and long-time time dependence relationships in the data by using a multi-scale TCN and a self-attention mechanism in a time dimension, and meanwhile, introducing a double-path self-adaptive information gating fusion technology to realize effective fusion of time features of different hierarchies; then constructing a local-global spatial dependency extraction module, respectively learning local and global spatial dependency relationships in the data by using a dynamic-static graph convolutional network and a self-attention mechanism in spatial dimension, and realizing effective fusion of spatial features of different levels based on a two-way adaptive information gating fusion technology; and finally, mapping the potential spatial-temporal feature representation into a prediction result through a full connection layer network.

Described herein are systems, devices, methods, and media for an emergency response data exchange system. The system provides an operations center emergency management application with a textual and / or graphical notification of emergency communications initiated on a premises managed by the operations center. The system exchanges operations center data (e.g., floorplans, video feeds, point of contact information, building access controls) with an emergency communications center (ECC) and / or with a first responder device to facilitate reducing emergency response times. The system selectively establishes communications (e.g., messaging windows) between the operations center, ECC, and / or first responder device to facilitate real-time information sharing between relevant parties of an emergency.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of financial science and technology, medical health and the like, and discloses an agent-based webpage information processing method, device and equipment and a medium. And distributing the data to a capturing agent to obtain webpage original data. According to the method, model context information is generated by extracting webpage source information and release time information, page structure representation is constructed, key information nodes are positioned, structured data are generated, then multi-scene decision data are generated in combination with the structured data and the model context information in the analysis process, and a decision result is output through a decision reasoning module. By constructing a uniform data processing flow, webpage capture, context generation, structured analysis, analysis processing and decision reasoning are organically linked, efficient acquisition and intelligent processing of webpage information are realized, and timeliness and accuracy of decision are improved.

Techniques for processing with respect to a user input as contextual information is available are described. A system generates a first task prediction using first context data that is available when a user input is received. The system generates a second task prediction (e.g., updated first task prediction) when second context data is received, and then further generates a third task prediction when third context data is received. Example first context data may include device type information, time information, location, etc. Example second context data may include automatic speech recognition (ASR) data. Example third context data may include natural language understanding (NLU) data. Using the third task prediction, the system generates an output responsive to the user input.

The invention provides a member information processing method and system based on an intelligent door lock and a storage medium. A first intelligent door lock responds to a first unlocking operation, a first face image and a second face image are captured, the first face image is recognized as a family member corresponding to the first intelligent door lock, the second face image is not a family member corresponding to the first intelligent door lock, and a first message is sent to a door lock management platform; the first unlocking operation is an external unlocking operation; the first message comprises room number information, time information and second face information corresponding to arrival of the first intelligent door lock; according to the door lock management platform, determining that the second face image is a family member corresponding to a second intelligent door lock, and sending a second message to the second intelligent door lock; the second message comprises room number information, time information and family member information corresponding to arrival of the first intelligent door lock. Compared with the prior art, based on interconnection of community door locks, the door locks report door number information of members, and anxiety of parents is eliminated.

The invention provides a three-dimensional space track identification method based on multi-view three-dimensional reconstruction, which comprises the following steps of: shooting the same scene from different angles by using a plurality of cameras, and determining a picture intersection part of the plurality of cameras as a research scene; frame extraction and pedestrian detection are carried out on the video of each camera, pedestrians and position information of the pedestrians in each frame of video are identified, a unified identity number is distributed to the same pedestrian, registration is carried out on video frame images of different cameras, the pixel position of each pedestrian in each camera is obtained, and a multi-view-angle triangulation method is adopted, and internal and external parameters of the cameras are combined, so that the pedestrian positioning accuracy is improved. Solving a three-dimensional point coordinate of each pedestrian in a camera coordinate system; the three-dimensional point coordinates are converted into coordinates in a world coordinate system, and world coordinate system coordinates of pedestrian point locations of all frames and all numbers are obtained; and according to the time information, connecting the pedestrian point locations of each frame and each number to form a track, and completing reconstruction of the pedestrian three-dimensional space track in the world coordinate system. According to the invention, manpower and calculation cost can be reduced.

The invention relates to the technical field of data linkage calculation, and provides a multistage document dependent freight trial calculation data linkage calculation method and system, and the method comprises the steps: expanding the attribute of a document node from a single business field to a composite vector containing a spatial-temporal feature through time information and geographic identification spatial-temporal dimension modeling; the dependency graph can capture the fine-grained dependency relationship of the freight service, and the space-time accuracy of data linkage is improved; when the basic information or the document information changes, the priority and the calculation path of the task queue are updated in real time through deep reinforcement learning, and calculation errors caused by change lag are avoided; through tensor synthesis operation, parallel subtask division, and a dual verification mechanism of a domain knowledge graph and a forced dependency edge, it is ensured that the whole process from dependency graph construction to calculation result output conforms to industry specifications, and the compliance and risk control ability of freight services are significantly improved.

The invention discloses a complex city river network hydrological data prediction method, optimization scheduling method and system, and the method comprises the steps: firstly constructing a space-time diagram neural network model, which comprises the steps: constructing a three-dimensional space-time diagram matrix reflecting the dynamic state of a river network and an adjacent matrix capable of aggregating the space-time information at the same time, and forming a complete space-time diagram data sample; extracting space-time fusion features by adopting a space-domain graph convolution method, and introducing an attention mechanism to calculate the correlation degree between nodes; and then training the space-time diagram neural network model, and carrying out multiple iterations to fit the model and achieve the optimum. And further according to predicted hydrological data obtained by the time-space diagram neural network model, based on the determined scheduling target and constraint conditions, constructing a river network optimization scheduling model, and solving the river network optimization scheduling model by adopting an NSGA-III algorithm to realize optimization scheduling of urban river network water environment improvement. The method has the advantages of being high in scheduling optimization quality and high in timeliness, and construction of a complex river network hydrological hydrodynamic physical model is avoided.

The invention relates to the technical field of information, and provides an intelligent power consumption behavior inspection method and device based on multi-source data and a storage medium, and the method comprises the steps: processing a mass power consumption data flow collected by a sensor and an intelligent electric meter in real time through employing a distributed calculation framework, completing the integration and standardization of the multi-source data, and generating a dynamically updated power consumption monitoring original data set; extracting a multi-dimensional feature vector from time series data and historical electricity utilization records containing the basic identifier, the time information, the real-time change feature and the abnormal mark, and outputting a classification result of the abnormal behavior; finally confirmed abnormal behavior data are extracted, a real-time monitoring instrument panel is generated through the visualization technology, and the abnormal behaviors of power utilization inspection of the power supply enterprise are dynamically displayed. According to the invention, multi-source heterogeneous data are effectively integrated, intelligent analysis and real-time monitoring of power consumption behaviors are realized, the accuracy and efficiency of power consumption inspection of power supply enterprises are improved, and powerful support is provided for safe and stable operation of a power system.

The invention relates to the technical field of edge computing, in particular to a safety production monitoring method based on edge collaboration, which comprises the following steps: collecting multi-node clock pulses, calculating adjacent time differences to form a pulse difference parameter set, carrying out statistical distribution, setting a dynamic tolerance boundary, and constructing an effective deviation parameter set; the method comprises the following steps of: generating a weighted smooth deviation result by combining clock stability and link delay weighted normalization, sorting monitoring data according to a timestamp to identify duplication abrupt change exception, constructing a cross-node exception time sequence by associating nodes and time information, and judging a safety state transition result of equipment. A time difference parameter is constructed through multi-node reference pulse arrival time, a dynamic deviation constraint is formed by combining distribution characteristics, clock stability and link measurement weighted smoothing are introduced, cross-node time reference consistency is kept, operation data unified time sequence rearrangement is carried out, a sudden change event reliably reappears in a time window, and an abnormal identifier is associated with a node identity. And forming a cross-node event evolution link.

The invention relates to the technical field of video processing, and relates to a video generation method and device, a computer program product and electronic equipment. The video generation method comprises the steps that an event description text sequence is received, a reference frame feature vector of a video to be generated is determined according to the event description text sequence, and the event description text sequence comprises text description information of multiple events and corresponding starting and ending time information; performing time sequence position coding processing on the reference frame feature vector according to the starting and ending time information to obtain a video frame time sequence fusion feature, and performing time sequence position coding processing on the text description information to obtain a text time sequence fusion feature; based on a time sequence cross attention mechanism, fusing the video frame time sequence fusion feature and the text time sequence fusion feature to obtain a target fusion feature; and generating a dynamic image according to the target fusion feature corresponding to each reference frame feature vector to obtain a target video. The quality of the generated multi-event video can be improved.

The invention relates to the technical field of resource scheduling, and particularly discloses a government affair collaborative emergency resource dynamic scheduling method and system.The method comprises the steps that event spatio-temporal information is obtained through a government affair data sharing platform, multi-source data are fused to construct a three-dimensional situation map, and a dynamic demand list is generated; positioning distributed resource nodes and establishing supply capability portraits; constructing an optimization model with response time efficiency, resource utilization efficiency and demand satisfaction balance degree as targets, and generating a multi-target scheduling scheme set by adopting a particle swarm algorithm; generating a visual decision billboard based on the scheme set, determining a final scheme through multi-party collaborative consultation, and analyzing the final scheme into an operable instruction; monitoring an instruction execution state and a resource track in real time, and automatically triggering scheme re-optimization when deviation or situation change occurs to form a closed-loop scheduling system; according to the invention, the method achieves the precise, intelligent and adaptive dynamic adjustment of resource scheduling, and improves the efficiency and robustness of emergency response.

The invention relates to the technical field of navigation, in particular to a bicycle refined intelligent navigation method, system and device and a storage medium, and the method comprises the following steps: obtaining information such as gradient and length of a ridable route; a physical exhaustion result is obtained through a physical exhaustion calculation model in combination with user weight, bicycle type and the like; calculating a time consumption result according to the riding ability of the user and evaluating a route difficulty level; acquiring real-time information of weather and road conditions, and generating detailed prompt information according to the real-time information; and sorting routes are screened according to user preferences, recommended and displayed. According to the method, refinement and intellectualization of bicycle navigation can be improved, physical power and time consumption are accurately calculated by integrating multiple factors, comprehensive and detailed route prompt and personalized recommendation are provided for users, different user requirements are met, and riding experience and efficiency are improved.

The invention relates to the technical field of Internet of Things security, and discloses an Internet of Things control system and a security early warning method, and the method comprises the steps: firstly collecting equipment basis and real-time information in an Internet of Things system, then evaluating equipment risk, network security and data encryption conditions, setting an early warning threshold according to multi-aspect evaluation results and historical security data, and finally, carrying out early warning. And finally, real-time monitoring is carried out, and early warning is triggered when the risk index exceeds a threshold value. The method has obvious advantages, equipment aging and safety risks can be accurately evaluated, equipment maintenance is planned in advance, and the maintenance cost is reduced; network attacks can be effectively identified, and network security is guaranteed; the data encryption security is enhanced, and the data leakage risk is reduced; early warning is given out in time, and the system response speed is increased; and the equipment safety state can be dynamically updated, the system safety management is continuously optimized, the safety and reliability of the Internet of Things system are comprehensively improved, and the stable operation of the Internet of Things system is promoted.

The invention relates to the technical field of water conservancy management, and discloses a water conservancy dynamic decision-making method and system based on digital twinning and multi-source heterogeneous data, and the method comprises the steps: carrying out the real-time information capturing of a target water conservancy system, and generating a system operation information scheduling matrix based on a grid geographic model and a time dislocation mark; performing hierarchical slicing on the system operation information scheduling matrix according to a multi-scale time window to form a multi-channel flow framework of regional hydrological dynamic behaviors; analyzing the evolution trends of the water level, the flow and the water quality by using a multi-channel flowing framework and a dynamic fusion analysis operator; dividing a river reach, a reservoir and a regulation and control facility corresponding to the potential risk event as candidate scheduling areas; and based on the local risk coupling network, dynamically optimizing a scheduling strategy, a pump gate operation sequence and a data acquisition frequency, and generating a real-time scheduling and emergency response scheme of the candidate region. The method has the advantage of improving flood early warning accuracy and water supply scheduling efficiency.

The invention relates to a cross-attention mechanism optimization method and a cross-attention mechanism optimization system for multi-modal feature fine-grained alignment. The method comprises the following steps: constructing a space-time topological graph of a dynamic scene, and representing a space-time relationship among intelligent agents through nodes and various connecting edges; coding the heterogeneous spatio-temporal information by using a plurality of groups of parallel recurrent neural networks, and converting the heterogeneous spatio-temporal information into uniform dimension feature representation; establishing a multi-modal cross attention alignment component, and quantifying feature quality from multiple dimensions through a multi-criterion evaluation unit; an adaptive weight fusion system is adopted to dynamically integrate evaluation results, and a unified quality score is generated; constructing a progressive optimization architecture based on the quality score, and performing multi-task cooperative training on a double-layer attention mechanism by combining real-time sampling and a directional optimization strategy; and finally generating a movement decision. According to the method, fine-grained alignment and optimization of multi-modal features are realized, and the accuracy and adaptability of intelligent agent navigation and trajectory prediction in a complex dynamic environment are remarkably improved.