444 results about "Undirected graph" patented technology

The invention discloses a multi-modal abnormal data detection and restoration method and system oriented to a power business scene. The method comprises the following steps: collecting multi-source heterogeneous power data, abstracting a power system into a weighted undirected graph, uniformly mapping the multi-source heterogeneous data into a graph signal, and preprocessing the collected data; extracting spatial features of nodes in a topological structure by adopting a graph convolutional network, and capturing a time dependency relationship in combination with a time sequence encoder; identifying various types of data abnormal points through an abnormal scoring function fusing the time sequence prediction error and the neighborhood consistency; a prediction-reconstruction combined repair strategy is adopted, time sequence prediction and neighborhood diffusion estimation are fused, and a preliminary repair value is generated; a lightweight parameter adapter is introduced, a scene feature vector is used as input, a repair weight and a regularization coefficient are dynamically generated, and a repair strategy is automatically adjusted; and performing physical consistency verification on a data result, wherein the physical consistency verification comprises power injection conservation constraint, voltage amplitude range constraint and time sequence continuity constraint.

The invention provides a method for automatically centering and assembling a bearing on a long-axis part, and belongs to the technical field of bearing assembly.The method comprises the steps that position information of the long-axis part and the bearing is collected in real time through a binocular camera and a laser distance measuring sensor, a central line equation is calculated through the least square method, and position deviation is determined; a centering adjustment vector is generated to control a multi-degree-of-freedom precise attitude adjustment platform to perform preliminary adjustment, a mechanical arm is matched with high-precision laser ranging to perform precise centering, the centering process is modeled into a weighted undirected graph, an optimal adjustment path is solved, optimal press fitting parameters are calculated based on a bearing press fitting physical mechanics equation, and press fitting is performed. Meanwhile, the assembling process is monitored through a force sensor and a displacement sensor, an assembling quality evaluation matrix is constructed, finally, the assembling result is detected and evaluated, and high-precision automatic assembling of the long-axis part and the bearing is achieved.

The invention discloses an electromagnetic field simulation grid adaptive generation method based on a neural network. The method comprises the following steps: performing mesh generation on a semiconductor device simulation model to obtain original generation data, and constructing node features; an undirected graph is constructed, graph nodes of the undirected graph adopt node features of grid points, and connecting edges adopt Euclidean distances between the grid points and adjacent grid points; inputting the undirected graph into a double-branch neural network, and outputting a predicted node position; and correcting each node of the original subdivision data by using the predicted node position obtained by the double-branch neural network to form new subdivision data for electromagnetic field simulation. According to the method, physical field information can be fused to efficiently adjust the node density, and the geometric boundary and topological structure characteristics of the device can be strictly kept.

The invention relates to an unmanned aerial vehicle control system and method for formation flight in a wind environment. The method comprises the following steps: generating a wind disturbance environment with a specified map scale and containing random obstacles; acquiring external image data and inertial measurement data of the unmanned aerial vehicle, and generating a grid map through point cloud extraction and conversion operation; the method comprises the following steps: modeling a spatial topological structure of an unmanned aerial vehicle cluster by adopting an undirected graph mode, and solving and generating a collision-free initial trajectory by adopting an unconstrained discrete optimization mode; optimizing the initial trajectory by adopting a task redistribution strategy, and re-planning a global target and a local target to obtain a re-planned optimized trajectory; and based on the optimized trajectory, solving by adopting a model prediction control mode to obtain an optimal control quantity for unmanned aerial vehicle bottom layer control so as to correspondingly control the motion state of each unmanned aerial vehicle and realize formation flight. Compared with the prior art, the unmanned aerial vehicle cluster can efficiently and accurately realize formation flight in an obstacle environment with wind influence.

The invention discloses an air-ground integrated network spectrum power joint distribution method combining hypergraph and meta reinforcement learning. The method comprises the following steps: step 1, establishing a multi-unmanned aerial vehicle communication network model; 2, constructing an undirected graph and hypergraph structure; 3, introducing a graph attention network to realize bidirectional aggregation of node features; 4, the aggregated node features are input into a Meta-DQN strategy network, so that the strategy has rapid adaptive capacity when facing environment changes and new tasks; according to the air-ground integrated network spectrum power joint allocation method combining the hypergraph and meta reinforcement learning, node feature representation with global perception capability can be acquired by each unmanned aerial vehicle agent on the premise of only depending on local observation, and rapid adaptation is realized in the face of dynamic environment change, so that efficient migration of a resource allocation strategy is realized, and the resource allocation efficiency is improved. The spectrum utilization rate and the communication success rate of the system are improved, and stable operation of the air-ground converged communication system in a complex scene is effectively guaranteed.

The invention provides a power distribution network fault intelligent positioning and self-healing recombination method and system. The method comprises the steps that S1, operation data of a power distribution network are acquired and preprocessed; s2, modeling the power distribution network topology into a weighted undirected graph, and constructing transient characteristics including node characteristics and edge characteristics; s3, on the basis of the transient characteristics and the power distribution network topology, the fault type is judged in real time by adopting a graph convolutional network, and a fault section is determined in combination with a traveling wave distance measurement and impedance method; s4, when entering an isolation fault state, automatically generating a power supply reconstruction scheme meeting current, voltage and switch constraints, and selecting an optimal scheme through multi-objective optimization; and S5, issuing a switching instruction according to the optimal scheme to complete topology reconstruction, and updating a fault identification and recombination strategy according to an actual execution result to improve positioning precision and recovery efficiency. According to the invention, a novel fault processing scheme integrating multi-source measurement, an efficient algorithm and online learning can be adopted, so that the power failure time is shortened, the power failure range is reduced, and the toughness of the power distribution network is improved.

The invention discloses a medical data center network risk assessment method based on a random walk model. The method comprises the following steps of 1, obtaining topological structure information and a node set of a medical data center network; 2, constructing a weighted undirected graph according to the topological structure and the data traffic information between the nodes; 3, in the weighted undirected graph, performing simulation based on a random walk model to obtain a transition probability matrix between nodes; 4, calculating a node risk score of each node according to the transition probability matrix; 5, performing aggregation processing on the node risk scores of the nodes to obtain an overall network risk score; and step 6, generating an assessment report according to the overall network risk score. According to the method, the risk conduction probability between the nodes is quantified by constructing the combination of the weighted undirected graph and the random walk model, the limitation of a traditional static assessment method is broken through, and a cross-node risk propagation path caused by data flow can be identified.

The invention discloses a construction method and device of a metal material performance prediction model, equipment and a medium. The method comprises the following steps: constructing an initial graph neural network; acquiring a three-dimensional volume element structure of the specified metal material and mechanical property data corresponding to the three-dimensional volume element structure; describing the three-dimensional volume element structure as an undirected graph structure; constructing a target data set by using the undirected graph structure and the mechanical property data; and performing iterative training on the initial graph neural network through the target data set to obtain a metal material performance prediction model. Therefore, the undirected graph structure of the three-dimensional volume element structure is used as a bridge, the metal material performance prediction model between the complex three-dimensional volume element structure and the mechanical performance data is established through the graph neural network, the model can rapidly and precisely predict the macroscopic performance of the metal material, the cost is low, and the method is easy to implement. And the requirements of rapid iteration and real-time feedback of the macroscopic performance of the material in engineering design can be met.

The invention relates to a multi-task-point cruise path planning method and device for an unmanned aerial vehicle. The method comprises the following steps: acquiring flight data of the unmanned aerial vehicle in multiple flight stages; the flight data comprises corner data when the flight stage is a corner stage; constructing a flight energy consumption model according to the undirected graph, the flight data and the rotation angle data; the energy consumption of each task point in the cruising process of the unmanned aerial vehicle is obtained; performing iterative optimization on the energy consumption of the plurality of task points based on a wild dog optimization algorithm to obtain a path planning result, and controlling the unmanned aerial vehicle to fly according to the path planning result; energy consumption of different flight stages is quantified by constructing a flight energy consumption model, including energy consumption of rotation angle data in a rotation angle stage, so that the complex environment adaptability is improved, iterative optimization is performed on a path in combination with a wild dog optimization algorithm, the problems of inaccurate energy consumption modeling and easy falling into local optimum are solved, and the energy consumption modeling efficiency is improved. The method has the advantages that the energy consumption optimization efficiency and accuracy of path planning are improved, and the local optimum problem is avoided.

The invention relates to the field of heliostat calibration, and discloses a scheduling method for calibrating heliostats through cooperation of artificial light sources and cameras in a large-scale heliostat field, and the method comprises the following steps: determining the heliostat range covered by each camera, determining the heliostat range covered by each artificial light source, and determining the heliostat range covered by each camera; the method comprises the following steps: determining the center and size of a light spot in an image shot by a camera when the heliostats are calibrated, determining the heliostats which can be calibrated in the coverage range of each artificial light source-camera combination, regarding each effective calibration combination as a node to construct an undirected graph, dividing the nodes into a plurality of groups through a graph grouping algorithm, and ensuring that the nodes in the same group are not connected with each other, therefore, all combinations in the same group can execute calibration in parallel. According to the method disclosed by the invention, the problem of affiliation misjudgment caused by light spot interference between heliostats is solved, possible shadow and shielding problems are avoided, dynamic scheduling of the artificial light source and the camera is realized, and invalid operation time is shortened.

The invention discloses a GCN-LSTM-based tailing dam multi-point settlement deformation prediction method. The method comprises the steps of collecting and preprocessing original settlement monitoring data of monitoring points in a research area; constructing a graph structure for settlement data between all monitoring point pairs, setting a threshold value, connecting node pairs with significant correlation coefficient relationships by using edges, constructing a weighted undirected graph, and converting the weighted undirected graph into a normalized adjacent matrix as the input of a graph convolutional network GCN; extracting the spatial topology of each monitoring point by using a GCN; capturing long-term time dependence in the settlement process by using a gating mechanism of the LSTM network; and fusing the GCN and the LSTM network, and finally outputting a predicted value through a full connection layer. According to the method, the spatial topological graph among the monitoring points of the tailing dam is constructed, and the GCN and LSTM networks are fused, so that the spatial correlation among the monitoring points and the time dynamic characteristics of the settlement data are effectively captured, and high-precision tailing dam multipoint settlement prediction is realized.

The invention relates to a capacity expansion optimization method, system and equipment based on network performance marginal gain, and belongs to the technical field of network communication. The method comprises the following steps: constructing an undirected graph of communication network topology; defining a service requirement for each pair of nodes in the undirected graph, endowing each edge with a capacity value, and constructing to obtain an MCFP model; on the basis of the Lagrange duality theory, a duality model of the MCFP model is constructed, and a column generation algorithm is adopted to carry out double-model synchronous solution to obtain a maximum concurrency factor, optimal path flow and capacity constraint optimal duality variable; performing network key bottleneck identification according to the capacity constraint optimal dual variable and generating a plurality of capacity expansion candidate schemes; and comprehensively evaluating the capacity expansion candidate scheme by combining the network performance marginal gain, the influence degree of the link capacity change and the implementation cost, and selecting the optimal capacity expansion scheme for implementation. According to the invention, accurate load evaluation, bottleneck quantitative identification and efficient capacity expansion optimization of the communication network can be realized.

The invention provides a flower bud and flower recognition and positioning method for a picking robot to perform zoned picking, and belongs to the technical field of picking robots. Inputting the segmented mask into a multi-scale feature fusion semantic recognition model based on hierarchical attention aggregation to perform preliminary classification of foreground flowers and background vegetation, performing a graph cut energy minimization segmentation algorithm on the preliminary segmentation mask to construct an undirected graph, and calculating a minimum cut set to obtain an accurate flower segmentation mask; key point heat map prediction is carried out on separated flower individual masks to extract flower center point coordinates and petal tip coordinates, depth image channel information is combined to calculate three-dimensional space position coordinates of each flower, and a flower three-dimensional positioning database is established for picking robot path planning; the problem of insufficient segmentation precision caused by serious overlapping of foreground flowers and background vegetation in a color space in a flower picking scene is solved.

The invention relates to the technical field of data processing, and discloses a video monitoring intelligent analysis method and system based on edge computing, and the method comprises the steps: generating a sub-region set for each camera in a monitoring region, a time sequence people flow feature vector of each camera, and a people flow abnormal sign of the current real-time people flow; a dynamic weighted undirected graph representing a topological relation of a plurality of sub-regions in a monitoring region is constructed, and then a feature matrix fusing time sequence people flow feature vectors and people flow abnormal marks is input into a sub-region key level probability prediction model based on a space-time diagram convolutional network and the dynamic weighted undirected graph. And the final key level of each sub-region is obtained after prediction and correction, so that edge equipment is controlled to execute frame extraction and video picture analysis of a corresponding frame rate. Therefore, the video analysis precision and the computing power load of edge equipment can be balanced by acquiring the real-time and historical people flow characteristics and the abnormal marks and fusing the physical distance and the topological connectivity between the sub-regions.

The invention relates to a heat supply prediction method and system based on spatial-temporal feature fusion deep learning, and the method comprises the following steps: S1, carrying out the collection and fusion of multi-source heterogeneous data, and constructing an integrated data set; s2, preprocessing the data; s3, constructing a graph structure model of the heat supply system, and constructing a weighted undirected graph; s4, constructing and executing forward calculation of the space-time double-flow deep network; s5, designing a composite loss function including mean square error loss and physical constraint loss, and performing joint optimization training on the space-time double-flow deep network; and S6, performing multi-step heat supply load prediction by using the trained model, outputting a heat supply load curve of each heat exchange station in a specified time period in the future, and integrating a prediction result with a heat supply scheduling system. The method has the advantages that the prediction precision is improved compared with that of a traditional machine learning model by capturing the spatial-temporal characteristics at the same time, and the advantages are more remarkable in extreme weather.

The embodiment of the application provides a kind of welding path planning method and device, method includes: by machine vision algorithm to the object to be processed is three-dimensional scanning, obtains three-dimensional point cloud data, by the segmentation algorithm based on deep learning to three-dimensional point cloud data is data analysis, and the geometry feature is identified;With the geometry feature is constructed into a undirected graph, and the optimal connected path of the weld center line is found on the graph by minimum spanning tree algorithm, and the optimal connected path is constructed to obtain the fine three-dimensional model of processing object;According to the three-dimensional coordinates of weld search site, current welding gun attitude and target weld length, determine the optimal search height, search depth, inclination angle and offset of welding gun, generate corresponding welding path planning based on search height, search depth, inclination angle and offset by welding process simulation algorithm;The application can effectively improve the precision and efficiency of welding path planning.

The invention relates to the technical field of computer graphics, in particular to a three-dimensional geological section real-time cutting method and system and a storage medium, and the method comprises the steps: constructing a BVH tree to accelerate the cutting of a three-dimensional geological model, and calculating the projection of intersecting line segments on a cutting surface two-dimensional coordinate system to obtain a projection line segment set; constructing an undirected graph structure based on the projection line segment set; and extracting a closed profile contour based on the undirected graph structure, subdividing a triangular grid, and rendering and displaying. The implementation of the scheme can break through the traditional preprocessing dependence, and improve the profile generation efficiency and the visualization integrity.

The present disclosure relates to a method, a device, and a program product for training a model. The method includes: receiving at least one unlabeled sample and at least one labeled sample for training a pre-training model, the pre-training model being used to extract features of the samples; creating an undirected graph associated with the pre-training model using the at least one unlabeled sample and a set of training samples associated with the pre-training model; dividing the undirected graph to form a plurality of sub-graphs based on corresponding features of the unlabeled sample and the set of training samples, the plurality of sub-graphs corresponding to a plurality of classifications of the samples, respectively; and training, based on the plurality of sub-graphs and the at least one labeled sample, the pre-training model to generate a training model. A corresponding device and a corresponding computer program product are provided.

The invention discloses a motor temperature prediction transfer learning method based on a physical constraint graph neural network, and the method comprises the following steps: S1, constructing an undirected graph with a label based on a physical structure of a source domain motor; s2, on the basis of the undirected graph of the label, constructing a physical constraint graph neural network; s3, training the physical constraint graph neural network by using operation data and temperature data of a source domain motor to obtain a pre-training model; s4, on the basis of the physical structure of the target motor, constructing an undirected graph of the target motor, and multiplexing the heat capacity parameter and the heat conductivity function in the pre-training model so as to obtain an initialized physical constraint graph neural network for the target motor; and S5, obtaining a temperature prediction model of the target motor. According to the method, the physical heat transfer rule is embedded into the transferable graph neural network structure, so that a high-precision temperature prediction model with physical interpretability can be quickly constructed for a new motor only by using small-batch data.

The invention belongs to the technical field of network key node identification, and particularly relates to a social information network key node identification method based on reinforcement learning. Comprising the following steps: acquiring original graph data of a social information network to be identified, introducing a virtual node to be connected with all nodes and representing the virtual node as an undirected graph; generating an initial node feature matrix based on the adjacent matrix of the undirected graph; carrying out topological relation enhancement processing according to the adjacent matrix and the initial node feature matrix to obtain a weighted topological incidence matrix; performing channel-level dynamic calibration on the initial node feature matrix to obtain a weighted node feature matrix; performing hierarchical topological aggregation according to the weighted topological incidence matrix and the weighted node feature matrix to obtain three aggregation features; fusing the three aggregation features to obtain weighted fusion features; the weighted fusion features serve as agent input, a reinforcement learning method is adopted to train a key node recognition network, and a trained key node recognition network is obtained; according to the invention, the node identification accuracy is improved.

The invention discloses an unmanned aerial vehicle ad hoc network synchronization method, which comprises the steps of establishing an undirected graph according to all nodes in an unmanned aerial vehicle ad hoc network and a communication link set among the nodes; in each round of iteration, each node obtains a synchronization state of a neighbor node and calculates a local synchronization error; in each round of iteration, abnormal nodes are eliminated, the fitness function of each node is minimized, each node evaluates the current state according to the fitness function of the node, meanwhile, an individual optimal solution of the node and a neighborhood global optimal solution obtained by exchanging information with neighbor nodes are updated respectively, and the node obtains a neighborhood global optimal solution according to the individual optimal solution and the neighborhood global optimal solution; and the time-frequency state vector of the node is adjusted, so that the node iteratively evolves towards the directions of a historical optimal state and a neighborhood optimal state. According to the invention, the consistency of time and frequency of the whole network can be quickly achieved, the synchronization precision and robustness are remarkably improved, and the distributed time-frequency synchronization capability with intelligent anti-interference is realized.

The invention relates to a modeling method for accurate traffic flow prediction, which comprises the following steps of S1, integrally modeling all intersections and roads into a road network undirected graph, and defining an adjacent matrix and a distance matrix, S2, defining propagation time delay of traffic flow at the intersections, and S3, calculating the traffic flow of the intersections according to the propagation time delay. S3, combining propagation time delay and traffic flow abrupt change influence of adjacent intersections to confirm a comprehensive effect of each intersection at the moment t, and obtaining traffic flow representation of the intersection at the moment t, S3, aggregating features of each intersection through an adjacent matrix and the propagation time delay by a space graph convolutional layer, and capturing a dynamic rule through a time graph convolutional layer, the method comprises the steps of (S1) obtaining a graph convolution layer, combining the graph convolution layer and a time convolution layer to form a space-time graph convolution network, and outputting final traffic flow prediction, (S2) giving the final traffic flow prediction and a kernel density estimation matrix of GKDE and setting three learnable matrixes, and (S5) outputting through a feedforward neural network. The method has the advantage of accurate traffic flow prediction.

The invention discloses a video abstract generation method and system based on graph-guided adaptive key frame sampling, and relates to the technical field of video abstract generation, and the method comprises the steps: obtaining and extracting different scale features of each video frame, and generating multi-scale fusion features through the weighted fusion of scale attention weights; calculating inter-frame similarity to construct a weighted undirected graph, obtaining global structured frame features of each video frame through multi-layer feature propagation, and calculating graph guide weight of each frame; introducing a learnable distribution focusing parameter to perform power scaling on the graph guide weight to obtain an adaptive sampling weight, and calculating a final comprehensive weight; and according to the video frame length of the original video, calculating the number of sampling frames through a video length adaptive function, and according to the comprehensive weight and the number of sampling frames, sampling key frames by adopting a frame screening strategy based on sorting to generate a video abstract. According to the method, the video abstract with better coverage, diversity and stability balance can be generated.

The invention discloses an unsupervised point cloud segmentation method and system based on multi-feature fusion and graph optimization. The method comprises the following steps: training a point cloud instance segmentation model by using a 3D indoor scene data set; firstly, scene point cloud in a data set is extracted to segment a foreground and a background, downsampling is carried out on the foreground, and features are extracted; then constructing an undirected graph segmentation point cloud distribution pseudo tag, and performing up-sampling to a complete point cloud; inputting a Mask3D model, and training a segmentation model in combination with a weak supervision loss function; and finally, deploying a depth camera to capture a target scene point cloud, and inputting the trained model to output an instance segmentation result. According to the method, the traditional point features and the pre-training features are considered, the performance dependence on the pre-training model is reduced, and the reliability of the method is improved. Meanwhile, aiming at the characteristics that foreground objects in an indoor scene are diverse in distribution and complex in structure, a foreground separation method is innovatively adopted, so that the model can be adjusted and optimized aiming at a foreground effect, and a better segmentation effect is achieved.

The invention discloses a drainage pipe network overflow prediction method based on a Pry-STNN model, and the method comprises the steps: carrying out the calibration of a checked SWMM model, and generating multiple rainfall data based on a Chicago rain pattern; modeling the drainage pipe network into an undirected graph with node features and edge features, and constructing a data set in a PyG format; a GNN and LSTM series space-time neural network architecture is adopted, space features are extracted, and time correlation is captured; a differentiable physical layer is embedded into the output end of the neural network, and the physical consistency of prediction results is ensured through physical constraints such as a continuity equation and a mixed loss function; finally, the model is optimized through staged training and a dynamic regularization strategy, evaluation is carried out according to indexes such as R, MAPE and RMSE, high-precision and robust drainage pipe network overflow early warning can be achieved under the scenes such as conventional rainfall, extreme rainstorm and pipe network faults, and powerful technical support can be provided for intelligent drainage and waterlogging prevention scheduling.

The invention provides a full-life-cycle water and fertilizer management method and system for blueberry planting, and relates to the field of crop fertilization, and the method comprises the steps: obtaining water and fertilizer management experiment data of a plurality of blueberry growth stages; based on the water and fertilizer management experimental data of the plurality of blueberry growth stages, establishing a growth-related weighted undirected graph of the plurality of blueberry growth stages; determining a key growth factor and a key water and fertilizer management factor of each blueberry growth stage based on the water and fertilizer management experimental data of the plurality of blueberry growth stages and the growth-related weighted undirected graph of the plurality of blueberry growth stages; based on the water and fertilizer management experiment data, the key growth factors and the key water and fertilizer management factors of the blueberry growth stage, determining candidate water and fertilizer management schemes of the blueberry growth stage; based on the candidate water and fertilizer management scheme of each blueberry growth stage, a water and fertilizer management scheme for the whole life cycle of blueberry planting is generated, and the method has the advantages that accurate supply and intelligent regulation of water and fertilizer are achieved, and the water and fertilizer utilization rate is increased.

The invention discloses a graph enhancement-based rolling bearing small sample fault diagnosis method, which comprises the following steps of: firstly, mapping a time-frequency image into a weighted undirected graph capable of accurately describing local texture and global semantic association through a self-adaptive dynamic graph construction strategy by calculating a pixel neighborhood variance and a global expectation thereof and dynamically generating a connection threshold value; the space-frequency dependency relationship between the pixels is explicitly coded; a graph feature enhancement residual block is designed in the diffusion denoising process, and the feature expression of a diffusion model in the denoising process is enhanced by cooperating with the local perception of the convolutional neural network and the global reasoning ability of the graph neural network by using the gating fusion and attention mechanism. And finally, the modules are integrated in the graph enhanced U-shaped network, so that high-quality and high-diversity fault sample generation is realized. Experiments show that the method significantly improves sample diversity and diagnosis accuracy on two bearing data sets, and provides a new approach for intelligent fault diagnosis under small sample conditions.

The invention relates to the technical field of path planning, in particular to a priority-based multi-agent path planning method and system, and the method comprises the steps: abstracting a target scene into an undirected graph; setting a set of agents participating in path planning in the target scene; setting a starting point and an ending point of each agent; defining a priority, a calculation formula of the priority, an initialization rule of a definition and an updating rule in a subsequent path planning step for each agent, wherein the corridor priority represents the defined agent, and the movement priority represents the defined agent; and performing path planning on the agents in the set based on the graph, the starting point and the ending point of each agent, the priority and the calculation formula of the priority. The method is used for improving the quality of the solution generated in the map with the corridor structure.

The invention discloses a vehicle path problem solving method, system and equipment and a medium, and particularly relates to a vehicle path problem solving method based on clustering decomposition and graph matching, which comprises the following steps: S1, receiving a large-scale vehicle path problem instance to be solved, modeling into a weighted undirected graph, and combining a series of constraint conditions and optimization targets; s2, constructing an offline knowledge base with a diversified structure, inputting a group of preset parameters in the step, and outputting a knowledge base stored in the memory; s3, aiming at the large-scale vehicle path problem instance to be solved, executing initialization operation to generate a global initial solution; and S4, by taking the global initial solution as a starting point, executing an iterative local search framework to carry out deep optimization on the solution until a preset termination bar is met. The technical problems that in the prior art, when a large-scale complex logistics network is processed, the calculation time consumption is long, the planning cost is high, and the result is unstable are solved.