343 results about "Perception model" patented technology

The invention relates to the technical field of power systems, and provides a power distribution network multi-dimensional regulation and control decision method, system and device and a medium, and the method comprises the steps: inputting the preprocessed multi-source operation data into a preset state perception model, and obtaining a multi-dimensional state vector representing the operation state of a power distribution network; a multi-dimensional state vector is used as a state space, regulation and control operation is used as an action space, a composite reward function is established according to a power distribution network operation target, and modeling is carried out to obtain a Markov decision process framework; interacting with a power distribution network simulation environment by adopting a deep reinforcement learning algorithm, obtaining a current state from a state space, selecting and executing regulation and control operation in an action space according to a strategy network, updating strategy network parameters based on feedback of a composite reward function until an optimal regulation and control strategy network is obtained, and obtaining a deep reinforcement learning strategy model; and performing strategy rolling updating based on the real-time monitoring data to obtain a target regulation and control strategy. According to the invention, comprehensive optimal regulation and control of a complex operation scene can be realized.

The invention relates to an Internet of Vehicles multi-agent cooperative sensing method and system. The method comprises the following steps: constructing a collaborative sensing network, wherein the collaborative sensing network comprises a self-agent and a plurality of collaborative agents; acquiring and processing sensing data through the collaborative sensing network; performing feature extraction to obtain intermediate features; self-adaptive sparsification is carried out to obtain sparse features, and the sparse features are compressed and transmitted to a self-agent; performing time sequence feature enhancement on the features of all the agents at the self-agent end; fusing the features to obtain fused features; and constructing an Internet of Vehicles perception model, and realizing perception by the detection model according to the fused features. According to the method, the calculation complexity of traditional global attention is reduced from the square level to the linear level through an adaptive sparsification mechanism, the calculation overhead is remarkably reduced while the multi-agent feature interaction precision is kept, and the method is more suitable for real-time operation on the vehicle-mounted edge equipment with limited resources.

The invention discloses a radiation safety management method and system based on cloud platform data driving, and relates to the technical field of cloud platform radiation management. The method comprises the following steps: collecting radiation field data and environment state data of a target area through a deployed intelligent sensing node; uploading the radiation field data, the environment state data and the context data from the service module to a cloud platform, performing energy compensation and radiation unmixing, and constructing a multi-dimensional feature vector; and processing the multi-dimensional feature vector by using a pre-trained situational radiation perception model, identifying radiation field features and an environment situation, and generating a situational radiation safety early warning signal. The technical problem that in the prior art, radiation safety monitoring depends on single-point measurement, comprehensive judgment cannot be carried out in combination with environment and service context data, and consequently the radiation field anomaly recognition capability is insufficient is solved, and the purpose that the radiation field anomaly recognition capability is improved through cloud platform data driving and context awareness model fusion is achieved. And the technical effects of high-precision identification and situational safety early warning of the radiation field state are realized.

The invention discloses a classroom real-time learning emotion knowledge and intelligent intervention system, and relates to the technical field of artificial intelligence auxiliary education. Comprising a time sequence perception modeling module, a conflict identification analysis module, an intervention priority control module, a decision fusion output module, a stability prediction and early warning module and a closed-loop vibration suppression regulation and control module, the time sequence perception modeling module establishes a unified time baseline and a phase reference field, and performs joint modeling on multi-source classroom learning condition data under the constraint of the unified time baseline and the phase reference field; time sequence semantic fusion features are extracted, and a contradictory candidate trajectory set is generated. According to the method, multi-modal time sequence semantic features are fused, a learning situation conflict source is identified, conflict intensity is quantified, an intervention priority and a punishment mechanism are constructed to guarantee rhythm stability, an evidence chain and an attention mechanism are adopted to generate a consistent decision, oscillation detection and time reversal regulation are combined, closed-loop control of the intervention process is achieved, strategy oscillation is effectively avoided, and the method is suitable for popularization and application. And the intervention continuity and stability are improved.

The invention relates to the field of data processing, in particular to a security agent construction system and method based on a large model. Constructing a low-temperature liquid leakage knowledge base; collecting normal operation condition data, and training the leakage sensing model in an unsupervised mode; configuring a large model with a tool calling interface as a security agent; multi-modal data are synchronously collected in real time, a preliminary abnormal alarm is generated through a perception model, and associated information is retrieved; fusing the multi-modal information by the intelligent agent to carry out leakage prediction analysis and comprehensive research and judgment; and for the confirmed leakage event, a professional tool is automatically called to carry out quantitative consequence simulation, and a scenario disposal scheme is generated. According to the invention, the dependence on scarce leakage samples is effectively overcome, and the detection reliability is improved; through knowledge enhancement and multi-modal fusion, the research and judgment accuracy is greatly improved; full-process automation from early perception, intelligent analysis to consequence severity division is realized, and the intelligent level and emergency response efficiency of industrial safety monitoring are remarkably improved.

The invention discloses a knowledge graph teaching path dynamic recommendation method and system based on reinforcement learning. The method comprises the following steps: constructing a cognitive adaptive dynamic knowledge graph; obtaining an individual multi-dimensional state vector of a student to be recommended; obtaining a trained student state perception model and a trained teaching path reinforcement learning model; inputting the individual multi-dimensional state vector of the to-be-recommended student into a trained student state perception model to obtain cognitive state information of the to-be-recommended student; and calling the trained teaching path reinforcement learning model by taking the cognitive adaptive dynamic knowledge graph as an environment and combining cognitive state information of the to-be-recommended student, so as to obtain a current personalized teaching path recommendation result of the student individual. The teaching content sequence can be dynamically adjusted according to the knowledge state and the learning behavior of the student, and the teaching integrating degree is improved; the teaching path is continuously optimized through learning, and different types of students can be adapted.

The invention belongs to the technical field of robot perception and decision making, and discloses a multi-modal fused intelligent target perception method and system, and the method comprises the steps: obtaining multi-modal data in a distribution network operation scene, carrying out the preprocessing, fusing the knowledge of the distribution network operation field, and generating knowledge-enhanced multi-modal features; carrying out cross-modal alignment processing on the knowledge-enhanced multi-modal features, carrying out graph structure-based joint semantic and space alignment on isomorphic modals, and carrying out feature projection-based binding alignment on heterogeneous modals to obtain consistent aligned multi-modal features in a shared semantic space; based on a dynamic adaptive strategy, screening and fusing the aligned multi-modal features to generate unified fusion features; and inputting the fusion features into a target perception model, and outputting an image segmentation result and a point cloud semantic segmentation result of the distribution network operation target. According to the method, high-precision segmentation and positioning of the target in a complex distribution network scene are realized, and safe and efficient operation of the robot is effectively supported.

The invention discloses a network topic hot spot extraction method based on bullet screen semantic recognition, and aims to solve the problems of bullet screen data semantic sparsity, semantic offset, noise interference and the like. The method is characterized by comprising the following steps: carrying out semantic coding on a bullet screen by utilizing a Transform structure; a dynamic semantic evolution perception model is constructed, semantic offset is measured through KL divergence, and topological correlation analysis is carried out through GCN; constructing a space-time density field in combination with a video time axis to realize space-time coupling feature fusion; automatically extracting a hot spot cluster by adopting an improved density peak clustering algorithm; and predicting a hotspot evolution trend by using an LSTM model. By means of the technical scheme, topic hotspots can be accurately captured, semantic evolution logic can be recognized, and the purity and predictability of hotspot extraction are improved.

The embodiment of the invention provides a model optimization method, electronic equipment and a storage medium. The method comprises the following steps: acquiring an initial quantization bit width and an initial pruning rate of each convolutional layer in a BEV perception model; on the basis of the initialized quantization bit width of each convolution layer, quantizing the weight of the convolution layer, and according to the initial pruning rate of each convolution layer and the importance score vectors of the multiple output channels, obtaining pruning masks of the output channels; based on the loss function, the quantized weight and the pruning mask training model, obtaining an initial deployment model; and carrying out online quantitative sensitivity evaluation training on the initial deployment model, and when an evaluation condition is met, taking the initial deployment model as a final deployment model. Therefore, the problems that the domain adaptation and generalization ability is limited, the detection precision is reduced, the domain adaptation and generalization ability is difficult to be efficiently utilized by NPU of edge chips such as horizon lines, model evolution in the training process cannot be responded, and error accumulation is serious are solved, and structured sparsity, dynamic adaptation and end-to-end collaborative optimization are achieved.

The governed interaction interface maintains an identity-preserving semantic-state for a knowledge entity and updates it only through validated semantic evidence generated autonomously by execution mechanisms. Semantic intentions describe semantic needs and are transformed into capability-requests containing no operational commands. Execution mechanisms—including environmental devices, perception models, symbolic analyzers, virtual actors, and software agents—interpret capability-requests independently through tool-side semantic interpreters and may act or decline to act. The system does not observe or evaluate tool behavior and receives only semantic evidence describing semantic meaning of any resulting effect. The governance engine evaluates each evidence fragment independently under identity, coherence, lineage, evidentiary sufficiency, deviation constraints, and contextual compatibility. Semantic-state transitions occur only when validated evidence satisfies governance constraints. The system performs no prediction, optimization, control computation, multimodal fusion, or supervisory coordination, remaining fully separated from device-level behavior.

The invention discloses an intelligent mobile robot navigation method for a dynamic environment, and relates to the technical field of computer processing, and the method comprises the following steps: S01, collecting the dynamic environment of a robot in real time through a laser radar, an RGB-D camera and a UWB module, building a robot real-time environment perception model, and carrying out the real-time environment perception of the robot; the robot real-time environment sensing model comprises parameterized description of metal reflection, dynamic obstacles and narrow terrains; s02, determining a plurality of key design parameters of navigation of the intelligent mobile robot; and S03, based on the plurality of key design parameters, constructing an environment perception accuracy objective function of the intelligent mobile robot and a real-time response and scene understanding depth objective function. According to the invention, high-precision environment perception, low-delay path re-planning and safe and efficient navigation are synchronously realized in a dynamic environment through metal reflection interference filtering, dynamic obstacle acceleration pre-judgment and narrow terrain adaptive passing strategies.

The invention discloses a government affair question-answering method, device and equipment based on image-text understanding and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining initial image data of a government affair question, preprocessing the initial image data to obtain to-be-processed image data, and storing the to-be-processed image data in a database; determining image information by using an optical character recognition technology, an image classification technology and a target detection technology; carrying out correlation analysis on the image information, carrying out semantic relation and logic relation matching to generate structured information, constructing a knowledge graph, matching the structured information with the knowledge graph to obtain a matching result, generating an initial government affair reply by utilizing a context sensing model and dialogue context information, and sending the initial government affair reply to a server; constructing a user portrait based on the historical behavior data of the user, and determining policy content in the knowledge graph based on the structured information to obtain government affair recommendation information; and integrating the initial government affair reply and the government affair recommendation information by using an image-text generation model to obtain a target government affair reply comprising an interactive diagram so as to improve the efficiency of government affair question answering.

The invention relates to a training data set migration method and device and a vehicle, and the method comprises the steps: obtaining first vehicle data and second vehicle data, the first vehicle data represents a perception hardware parameter of a first vehicle model, the second vehicle data represents a perception hardware parameter of a second vehicle model, and the first vehicle model and the second vehicle model are different vehicle models; generating a perception transformation matrix based on difference information of the first vehicle data and the second vehicle data; obtaining original training data applied to a perception model of the first vehicle type, and performing transformation processing on the original training data based on the perception change matrix to obtain perception transformation data; and obtaining target training data based on the perceptual transformation data. According to the method and the device, the existing training data can be migrated to perception model training of different vehicle types, the training cost of the multi-vehicle-type perception model is reduced, the reuse rate of the training data is improved, data fusion is carried out based on vehicle type differences during training data migration, and the accuracy of the training model after training data migration can be ensured.

The invention relates to a mechanical arm disordered grabbing algorithm, and belongs to the technical field of robotics, and the method comprises the steps: firstly, obtaining a target material image and a point cloud sample, and training a multi-modal attention fusion perception model through sample enhancement processing and a difficult case mining mechanism after the target material image and the point cloud sample are labeled with categories, postures and grabbing areas; secondly, a mechanical arm obtains operation area data, the operation area data are input into a multi-modal attention fusion perception model to obtain initial attitude parameters, errors are corrected by combining with a visual data fusion calibration model, and the material stacking state is judged through a dynamic reference comparison method; thirdly, based on the correction parameters and the stacking state, an evaluation function is constructed to screen collision-free and stress-balanced grabbing points, and a collision-free path is planned by fusing kinematics constraints of the mechanical arm; and finally, the grabbing action is executed, the strength and the posture are adjusted in real time according to the contact force and the position deviation, the material stability is evaluated through vision after grabbing, and the stability and the operation efficiency of the mechanical arm in a complex disordered scene are improved.

The invention relates to the technical field of automatic driving, in particular to an automatic driving robust planning system and method based on dual-channel perception, and the system and method construct a standby input channel independent of a main perception model at the planning level. And the problem of failure of sensing a single point is fundamentally solved. Automatic and seamless mode switching is realized through confidence monitoring, and high performance in a normal state and high safety in an abnormal state are considered at the same time. The design target of the second sensing path (occupying the grid) is more single (geometric occupation), and the second sensing path can be made lighter and more robust and is complementary with the first path. The architecture provided by the invention meets the requirements on redundancy and safety architecture in the functional safety standard (such as ISO 26262) of the automobile industry, and is beneficial to improving the overall safety level of the automatic driving system.

The invention discloses an urban water supply prediction method based on a non-stationary perceptual Transform-BiLSTM model, and the method comprises the steps: independently calculating the mean value and variance of each sequence sample as non-stationary statistical information through introducing a reversible normalization mechanism, enabling a sequence to depend on weight learning reversible mapping on the basis of maintaining the original distribution characteristics of data, and carrying out the calculation of the mean value and variance of each sequence sample as the non-stationary statistical information; and the perception capability of non-stationary components is enhanced, and the training stability is improved. In addition, discrete cosine transform (DCT) is adopted for frequency domain modeling, and the feature extraction capacity of the model for the periodicity and trend of the urban water supply sequence is enhanced. In the aspect of feature modeling, the model fuses the global attention mechanism of Transform and the time sequence modeling capability of BiLSTM, cooperatively captures the long-term dependency relationship and local time sequence dependency features in urban water supply data, and has a remarkable capturing capability effect on the change of a non-stationary structure under the background of multi-scale fluctuation of an urban water supply sequence.

The invention discloses a scene perception information fusion decision-making method for industrial field multi-source heterogeneous data, which comprises an edge layer, a cloud layer and an interaction layer which form a three-level collaborative architecture, and belongs to the technical field of industrial field data processing. The method comprises the specific steps that S1, an edge layer collects industrial field multi-source heterogeneous original data and carries out self-adaptive preprocessing to obtain cleaned multi-source feature data; s2, the edge layer processes the multi-source feature data through a lightweight scene perception model, and outputs a preliminary scene judgment result; s3, the cloud layer constructs and updates a causal relationship graph based on industrial business knowledge and incremental data of the edge layer, processes data uploaded by the edge layer through a causal reasoning fusion model, outputs a decision result and issues the decision result through the interaction layer; and S4, the cloud layer optimizes parameters of the causal reasoning fusion model and the lightweight scene perception model based on a decision execution result fed back by the edge layer, and issues the optimized parameters to the edge layer through the interaction layer to realize dynamic iteration.

The invention discloses a subway passenger and freight co-transport simulation method and system based on multiple agents, and belongs to the technical field of urban traffic and logistics simulation. The simulation method comprises the following steps: constructing a multi-source heterogeneous urban traffic basic database; generating a passenger agent and a cargo agent, and allocating an initial travel plan or a logistics strategy to each agent; simulating a passenger transport benchmark equilibrium state, loading a passenger agent through iteration, calculating a utility score, and re-planning a low-score agent until the passenger transport benchmark equilibrium state is converged to a user equilibrium state; constructing a passenger and freight co-transport conflict interaction and nonlinear perception model, and constructing a coupling congestion perception model and a nonlinear comfort penalty function; passenger and cargo co-evolution iterative simulation is carried out, passengers and cargo intelligent agents are loaded at the same time, multi-dimensional utility evaluation and dynamic collaborative re-planning are carried out, and a system is simulated through a double-layer coupling iterative mechanism to find a new equilibrium process; and calculating and outputting a passenger flow influence index, a logistics efficiency index and a system comprehensive benefit.

The invention relates to the technical field of image processing, and discloses a fish school density determination method and device based on multi-range adaptive partition, computer equipment and a storage medium. According to the invention, the multi-modal perception model is used to replace the traditional mode of reading and marking the range of the sonar image depending on manual visual reading, and the polar coordinate-Cartesian coordinate mapping model is constructed based on the analyzed range parameters, so that the self-adaptive geometric partitioning and geometric distortion correction of the original sonar image are realized. A sonar image is decomposed into a plurality of microcosmic local block masks by generating local block masks, local effective fish school and local water volume are independently calculated in the local block masks, and finally the overall fish school density is determined. The method can be applied to the field of cultivation insurance insurance claim settlement business, and the capability of eliminating image distortion when an image system identifies and scans an image is improved.

The invention relates to the field of transformer overload control, in particular to a method and system for dynamically evaluating the overload capacity of a transformer, and the method comprises the steps: firstly, precisely extracting load step features and eliminating interference through a high-precision collection and sliding window differential verification technology; then, operating a finite-state machine based on a fluid dynamics principle, and switching among three modes of steady-state coupling, inertial decoupling and linear recovery according to load change; especially, a decoupling mode is entered during load impact, adiabatic temperature rise is utilized to deduce and replace lagged actually measured top oil temperature, and the problem of calculation distortion of a traditional model during oil flow lagging is effectively solved. In addition, the system executes double-thread calculation of the standard model and the inertial sensing model in parallel, outputs the final hot spot temperature through the maximum arbitration logic, and deduces the remaining allowable operation time according to the final hot spot temperature so as to generate a protection control signal. According to the invention, the evaluation accuracy and the operation safety of the transformer under the short-time overload condition are obviously improved.

The invention relates to the technical field of automatic driving, in particular to a dynamic obstacle trajectory prediction planning method and system based on a time sequence occupancy grid, constructs a standby input path independent of a main sensing model at the planning level, and fundamentally solves the problem of failure of a sensing single point. And performing confidence evaluation on the structured sensing result output by the BEVTransform model to judge whether the current sensing result is reliable or not. And entering a dynamic obstacle trajectory prediction planning scheme of a standby input path when the dynamic obstacle trajectory prediction planning scheme is not reliable. The method completely bypasses the dependence on obstacle detection, classification and tracking, directly senses and predicts the motion from the geometric level, and is particularly effective when the BEV fails. Static and dynamic elements are represented by a uniform occupied grid, and an input interface of the planner is simplified. By predicting the future occupancy situation, a more prospective and safer decision can be made, such as speed reduction in advance for giving way.

The invention provides a lifting machine operation scene understanding method and system fused with a multi-mode perception model, and relates to the technical field of artificial intelligence, and the method comprises the steps: firstly collecting scene triggering type perception information of vision, mechanics, position and the like of a lifting machine operation scene to form a triggering perception information set; executing cross-modal interactive mapping processing on the trigger sensing information set to obtain interactive mapping sensing information; scene adaptation calibration processing is carried out based on the scene adaptation perception model to obtain a scene adaptation perception model; inputting the interactive mapping perception information into a scene adaptive perception model to carry out multi-dimensional situation deduction to generate panoramic situation description; and finally, based on the panoramic situation description, a dynamically-adaptive lifting machine operation regulation and control instruction is generated and transmitted to the execution control unit, and accurate understanding and intelligent regulation and control of the lifting machine operation scene are achieved.

The invention discloses an electric meter intelligent communication detection method and system based on a power line carrier, and relates to the technical field of electric meter communication detection, and the method comprises the steps: collecting the communication data of an electric meter at different time periods, inputting an electric meter communication state dynamic sensing model, analyzing features, and obtaining an initial evaluation result; channel characteristic parameters are calculated by using a carrier channel characteristic dynamic estimation algorithm based on the result, and the channel characteristic parameters are input into a power carrier signal demodulation optimization algorithm to generate an optimized demodulation scheme; signals are demodulated and data are verified according to the scheme, related results are uploaded to an electricity meter communication performance optimization management platform, and the platform evaluates performance and generates optimization suggestions and fault early warning. The system comprises a data acquisition unit, a state analysis unit, a channel calculation unit, a demodulation optimization unit, a data processing unit and a performance management unit which operate cooperatively. The communication state can be dynamically sensed, the channel characteristics can be accurately estimated, the demodulation process can be optimized, the integrated management of the communication performance can be realized, and the accuracy and reliability of the electric meter communication can be improved.

The invention discloses a multi-unmanned aerial vehicle cooperative search method based on dynamic region division and a multi-constraint virtual force field, and aims to solve the problems of low search efficiency, unbalanced load, difficulty in guaranteeing the motion safety of multiple unmanned aerial vehicles and the like in multi-unmanned aerial vehicle coverage search. Firstly, a task area, an unmanned aerial vehicle kinematics model and a sensor sensing model are constructed; then, dynamic region division is carried out based on a weighted Voronoi diagram, the search progress of the unmanned aerial vehicle, the responsibility region area and the obstacle distribution condition are comprehensively considered, and self-adaptive balancing of task loads is achieved; and selecting a target point based on a utility function, and finally designing a multi-constraint virtual force field control method consisting of target attraction, obstacle avoidance force, regional barrier force and exploration force, generating an expected speed meeting speed constraint and obstacle avoidance requirements, and realizing stable and efficient coverage search. The result shows that the multi-machine collaborative search efficiency can be effectively improved, and the method is suitable for disaster search, routing inspection monitoring, autonomous detection in a complex environment and other scenes.

The invention discloses a three-dimensional occupancy perception method and system suitable for multiple tasks, and belongs to the field of robot perception, and the method comprises the steps: carrying out the manual labeling of point cloud data and image data, obtaining a point-by-point label and a pixel-by-pixel label, carrying out the data verification, converting the point cloud data with the labels into a 3D semantic occupancy label of a scene through 3D semantic reconstruction, and carrying out the recognition of the 3D semantic occupancy label. Forming sample data; training a three-dimensional occupancy perception model by using the sample data and performing three-dimensional occupancy perception prediction; sequentially performing feature enhancement extraction, feature dimension conversion, feature time sequence fusion and feature compensation correction on the image data based on a feature extraction module to obtain multi-scale voxel features; the feature information of the multi-scale voxel features is independently adjusted through a semantic segmentation head and a target detection head, then a 3D semantic occupancy result and an instance detection result are output, the two detection results are converted into an instance segmentation result and a target tracking result through a post-processing module, and low-cost and high-efficiency multi-task three-dimensional occupancy perception is achieved.

An intelligent optimization method for an electrical-carbon coupled virtual power plant based on meta-reinforcement learning and distributed robust optimization relates to the field of virtual power plants, and comprises the following steps: sensing and preprocessing multi-source heterogeneous data, and constructing a multi-dimensional high-quality feature matrix; constructing an electricity-carbon deep coupling model, and quantifying a stepped carbon price risk and user preference synergistic effect; performing element intelligent hierarchical decision optimization, and outputting an optimal cooperation strategy of the power-carbon market; according to the method, the correlation characteristics of dynamic fluctuation and multi-source uncertainty of the electricity and carbon market are fully considered, a stepped carbon price-risk perception model and a user implicit preference collaborative model are constructed, cross-scene dynamic bidding strategy optimization is realized through a Meta-SAC element reinforcement learning algorithm, and the method has the advantages of being high in robustness, high in robustness and high in accuracy. The output randomness of renewable energy sources is processed through Wasserstein distributed robust optimization, multi-agent distributed collaborative scheduling is completed in combination with a federal ADMM algorithm, and finally, strategy dynamic evolution and optimization targets are achieved through a closed-loop feedback mechanism.

The invention relates to the technical field of data processing, in particular to a localized low-power-consumption computing power aggregation scheduling system and method based on a heterogeneous SoC, and the system comprises a local networking module, a modeling module, a scheduling distribution module, an execution module, a result aggregation module and an optimization scheduling module. According to the method, a technical chain from dynamic perception to intelligent decision and then to closed-loop control is constructed, multi-dimensional parameters of computational nodes are deeply coupled with demand parameters such as the calculated amount of tasks and the data dependency relationship, the coupled parameters are input into a scheduling model with the total energy consumption of a system as an optimization target, and in a local dynamic network with power supplied by a battery, the optimal energy consumption of the system is obtained. The system can adaptively select the node combination with the lowest energy consumption cost and the task decomposition mode, and the problem that reliable and efficient persistent computing power aggregation cannot be realized in the scene due to the fact that an optimization target and a resource sensing model do not conform to the real constraint of a low-power-consumption dynamic network is effectively solved.

The application relates to a conversation recommendation method based on a sequence growth perception model and belongs to the technical field of network big data information recommendation. In the method, all conversations are divided into multiple growth states according to a window increment rule with a size of one. The same growth state of all conversations is established as a global dynamic graph in a discrete form to model sequential information and dynamic transition patterns. Then, a dynamic graph neural network containing a growth perception layer and a dynamic perception layer is used to capture the global dynamic transition patterns. Meanwhile, a single conversation is established as a local static graph to capture local static transition patterns. Finally, conversation recommendation is realized based on the captured global dynamic transition patterns and local static transition patterns. The method greatly enriches the item transition pattern information in the conversation and effectively improves the recommendation effect.

The invention discloses a data preprocessing method and system for edge computing, and relates to the technical field of edge device data processing, and the method comprises the steps: determining the device state of an edge device and the data value of to-be-transmitted data, and outputting a processing capability score based on the device state and the data value through employing a pre-constructed capability perception model; outputting a processing strategy parameter through a pre-constructed strategy relation table according to the processing capability score, and associating the processing strategy parameter with the to-be-transmitted data to form a to-be-transmitted data packet; and setting an activity execution channel based on the edge device to transmit the to-be-transmitted data packet, and performing preprocessing control by using the processing parameter packet through the activity execution channel in the transmission process. According to the method, the occupation of redundant strategy parameters on transmission bandwidth and storage resources can be reduced, the data preprocessing efficiency and the resource utilization rate of edge equipment are remarkably improved, the local processing advantages of edge computing are fully played, and the requirements of accurate data processing and efficient resource utilization in an edge computing scene are better met.

The present application relates to a kind of based on multi-source industrial data's car body welding quality intelligent perception, prediction and decision-making method and system, belong to the technical field of intelligent manufacturing and industrial data analysis.The present application obtains multi-source high-frequency synchronous original signal by sensor array and robot bus;Perform space-time alignment and multidimensional feature mining;Quality perception model is constructed based on multi-path parallel network and cross-modal attention;Quality evolution trend and electrode cap life are evaluated using time series prediction model;Integrate expert rules and reinforcement learning to make process compensation decision.The present application realizes the accurate perception of welding quality, forward warning and closed-loop real-time compensation, significantly improves the perception accuracy, production robustness and prolongs the service life of electrode.