Trajectory flow dynamic node importance prediction method based on high-order dependency relationship

By constructing a dynamic model based on high-order dependencies in the transportation network and using trajectory flow information to predict the importance of nodes, the shortcomings of traffic distribution prediction in the existing technology are solved, and dynamic management and efficient operation of the transportation system are achieved.

CN118917077AActive Publication Date: 2024-11-08NAT UNIV OF DEFENSE TECH

Patent Information

Application Number
CN202410984753.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-08
Estimated Expiration
2044-07-22

AI Technical Summary

Technical Problem

The prior art is difficult to accurately predict the true flow distribution of the topological layer in the transportation network, and the traffic prediction on the load layer requires a large amount of sample data, so it is impossible to effectively analyze the dynamic properties in small-scale trajectory flows.

Method used

Based on high-order dependencies, dynamically construct a dynamic model, and predict the importance of nodes in the transportation network through carrier particle trajectory flow information and maximum matching principle, and dynamically fit the flow distribution on the load layer.

Benefits of technology

Provides a new perspective on dynamic management of transportation systems, which can effectively predict the importance of nodes in non-Markov systems. It is suitable for trajectory flow analysis with short time scales and insufficient sample size, and improves the operational efficiency of transportation systems. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917077A_ABST
    Figure CN118917077A_ABST
Patent Text Reader

Abstract

The invention discloses a trajectory flow dynamic node importance prediction method based on a high-order dependency relationship. The method comprises the following steps: dynamically constructing a dynamic model based on the high-order dependency relationship in a trajectory flow; and on the basis of the dynamic model, predicting the importance of each node in the transportation network based on the trail flow information of the carrier particles and the maximum matching principle. The method is applied to the field of node importance prediction on a transportation network, and predicts the node importance on a particle scale based on a high-order dependency relation generally existing in a transportation system and a kinetic model for mining track flow connotation, thereby dynamically fitting flow distribution on an actual load layer. The method is suitable for analyzing and predicting the trajectory flow with short time scale and insufficient sample size, and can effectively solve the problems and current situations that an evaluation method on a topology layer in a transportation network is difficult to predict real flow distribution and large sample data is needed for flow prediction on a load layer.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of node importance prediction, and in particular to a trajectory flow dynamic node importance prediction method based on high-order dependency. Background Art

[0002] With the deepening development of globalization, transportation systems such as traffic and communications have become the blood and nerve center of modern society. Among them, ensuring the normal flow of entities such as pedestrians, ships, and mail is the most basic and important function in the transportation system. At the same time, contemporary transportation systems are becoming more and more complex and changeable, and the difficulty of managers' cognition and decision-making is constantly increasing. Correctly understanding the internal structure of the transportation system has important application value for design, management and other decisions. Researchers usually abstract the inherent connections of complex systems into complex networks, and then analyze their components or global properties and dynamic processes. Among them, key node identification and prediction are directions in the field of complex networks that have both theoretical and application value. Transportation systems are often modeled as transportation networks, that is, double-layer networks, such as Figure 1 As shown, it includes a static network topology layer representing actual locations and their pathways, and a temporal directed weighted network load layer representing real-time traffic distribution.

[0003] Existing methods for identifying key nodes in transportation networks can be divided into two categories: those targeting the topological layer or the load layer. At present, many studies focus on the topological layer, identifying key nodes based on graph theory or graph mining methods, which can be roughly divided into three categories: the first category is to quantify the importance of the node itself and its neighbors through topological information, such as degree centrality, clustering coefficient, etc.; the second category is to quantify the importance of the node based on its location information in the network, such as closeness centrality, betweenness centrality, etc. The third category is a quantitative method that simulates particles wandering on the network, such as PageRank, HITS, SALSA and other continuously improved algorithms. In addition, some studies comprehensively consider the above-mentioned multi-class attribute hybrid design indicators, or identify key nodes based on entropy theory and machine learning methods. The node importance calculated by the method targeting the topological layer can predict the importance of the node in the transportation network to a certain extent, but it has limitations such as static. Therefore, the topology-based method often deviates from the actual situation, especially the nodes with higher importance ranking have large deviations.

[0004] Some studies predict the flow of edges or nodes on the load layer to evaluate their importance, which can be divided into three categories: the first category is statistical methods, such as Kalman filters and autoregressive moving average models (ARIMA); the second category is machine learning methods, such as K-nearest neighbor algorithms and linear regression vector machines; the third category is deep learning methods, such as convolutional neural networks with residual layers, long short-term memory recurrent neural networks, and gated neural networks. At present, the accuracy of traffic prediction methods on the load layer is relatively high. However, in order to train an effective model, researchers need to train data with a longer time scale and a large sample size, and rarely analyze the dynamic properties of the vehicle on a small scale to predict key nodes. Summary of the invention

[0005] In view of the deficiencies in the above-mentioned prior art, the present invention provides a method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies, which can effectively solve the problems and current situation that the evaluation method at the topological layer in the transportation network is difficult to predict the actual flow distribution and the flow prediction at the load layer requires large sample data.

[0006] To achieve the above object, the present invention provides a method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies, comprising the following steps:

[0007] Dynamically construct a kinetic model based on high-order dependencies in the trajectory flow;

[0008] On the basis of the dynamic model, the importance of each node in the transportation network is predicted based on the trajectory flow information of the carrier particles in the past and the maximum matching principle.

[0009] In one embodiment, the process of constructing the kinetic model is:

[0010] Step 101, the trajectory flow of the carrier particle is represented as Among them, Node j represents the node where the carrier particle is located at time j, l max is the length of the trajectory flow, j = 1 ~ l max ;

[0011] Step 102, split the trajectory flow of the carrier particle into multiple segments with lengths of 2, 3, ..., 1 max A continuous segment of length l max The continuous segment is the trajectory flow itself, with a length of l max There are 2 consecutive segments of length -1, and so on, there are l consecutive segments of length 2. max -1 item;

[0012] Step 103, calculate the number of times each continuous segment appears in the continuous segment set after all trajectory streams are split, and record the continuous segment as (Node t-K ,Node t-K+1 ,…,Node t Node t+1 ) with count, indicating that a carrier particle is from Node t-K Walk to Node one by one t+1 The trajectory of , where K represents the order of the rule, and 1<K<1 max -1, t+K+1≤l max ;

[0013] Step 104: assemble the continuous segments according to the node t Classify and obtain a collection Represents the set of high-order dependency rules corresponding to the node. For a transportation network with N nodes, there are N sets S Nodei (i=1,2,3,...,N), all sets are the dynamic model of the transportation network

[0014] In one embodiment, step 2 specifically includes:

[0015] Step 201, obtain the trajectory flow information of M carrier particles used for node importance prediction, and record the importance of each node in the transportation network as and initialized to 0, where N is the number of nodes in the transportation network;

[0016] Step 202: The trajectory flow information of the mth carrier particle is passed to the trajectory flow information of the past movement. Split into nodes And the length is 1 to l max -1 continuous sequence data, get l max -1 sequence and record it as a sequence set

[0017] Step 203: Set the sequence The longest sequence in At the node The corresponding high-order dependency rule set Match and judge the node Does the corresponding high-order dependency rule set contain the same sequence? High-order dependency rules for matching points:

[0018] If yes, proceed to step 204;

[0019] Otherwise, the sequence From the sequence set S seqm , and perform step 203 again;

[0020] Step 204, obtain sequence In the high-level dependency rule set All matching sequences in , and the importance of the prediction nodes corresponding to each matching sequence is updated;

[0021] Step 205, repeat steps 202 to 204 until the trajectory flow analysis of M carrier particles is completed and the node importance is obtained.

[0022] In one embodiment, step 204 is specifically:

[0023] Get Sequence In the high-level dependency rule set All matching sequences in are represented as:

[0024]

[0025] For the matched nodes Add its importance Among them, g=1~K.

[0026] Compared with the prior art, the present invention has the following beneficial technical effects:

[0027] 1. Based on the high-order dependencies that are prevalent in transportation systems, the present invention mines the dynamic model of trajectory flow connotation, predicts the importance of nodes at the particle scale, and can dynamically fit the flow distribution on the actual load layer, thereby providing managers with a new perspective to help them operate the basic functions of the transportation system more effectively;

[0028] 2. The present invention is applicable to transportation systems with strong non-Markov properties in the real world. It can dynamically and effectively mine high-order dependencies in trajectory flows and predict node importance based on this. It is also applicable to analyzing and predicting trajectory flows with short time scales and insufficient sample sizes. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0030] Figure 1 It is a schematic diagram of the double-layer structure of the transportation network;

[0031] Figure 2 A schematic diagram of a high-order dependency relationship in an embodiment of the present invention;

[0032] Figure 3 Flow chart of a method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies in an embodiment of the present invention;

[0033] Figure 4 is a flow chart of the process of constructing a kinetic model in an embodiment of the present invention;

[0034] Figure 5 A schematic diagram of modeling high-order dependency rules in a trajectory flow in an embodiment of the present invention;

[0035] Figure 6 A flowchart of node importance prediction in an embodiment of the present invention;

[0036] Figure 7 It is a line graph of average traffic proportion under three sortings in the embodiment of the present invention;

[0037] Figure 8 In the embodiment of the present invention, Sort HOD,j Accuracy matrix heat map.

[0038] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0041] In addition, the technical solutions between the various embodiments of the present invention can be combined with each other, but it must be based on the fact that ordinary technicians in the field can implement it. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0042] The movement of carriers in a transportation system, such as a ship sailing from its port to its destination port, can be abstracted as the movement of particles in a network, where the temporal position data recording the movement of particles is the trajectory flow. The topology layer is a static network inherent in the transportation system, and the load layer is a temporal directed weighted network established using the trajectory flow. The flow of a node is the weighted in-degree of the corresponding node on the load layer. In practice, particle movement largely follows high-order dependency rules, such as Figure 2 As shown in the figure, under the first-order dependency, the trajectory of the particle is only related to the node at the current moment; under the high-order dependency, the trajectory of the particle is related to the two nodes it passes through, such as the particle passing through the "sun" node will go to the "smiley face" node. Generally speaking, the node that depends on the first K paths is called the K-order dependency, and when K>1, it is a high-order dependency. Based on this, this embodiment discloses a trajectory flow dynamic node importance prediction method based on high-order dependency, referring to Figure 3 , including the following steps:

[0043] In the first step, a dynamic model of the transportation network is dynamically constructed based on the high-order dependencies in the trajectory flow;

[0044] In the second step, based on the dynamic model, based on the trajectory flow information of the carrier particles in the transportation network and the maximum matching principle, the target node of the carrier particles at the next moment in the current transportation network is predicted, so as to fit the actual flow distribution on the load layer at the next moment, so as to evaluate the importance of the node and obtain the importance of each node in the transportation network.

[0045] refer to Figure 4 , the construction process of the kinetic model in this embodiment is:

[0046] Step 101, obtain the known historical trajectory flows of several carrier particles on the current transportation network, and express the trajectory flows of the carrier particles as Among them, Node j represents the node where the carrier particle is located at time j, l max is the length of the trajectory flow, j = 1 ~ l max ;

[0047] Step 102, split the trajectory flow of the carrier particle into multiple segments with lengths of 2, 3, ..., 1 max A continuous segment of length l max The continuous segment is the trajectory flow itself, with a length of l max There are 2 consecutive segments of length -1, and so on, there are l consecutive segments of length 2. max -1 item;

[0048] Step 103, calculate the number of times each continuous segment appears in the continuous segment set after all trajectory streams are split, and record the continuous segment as (Nodet-K ,Node t-K+1 ,…,Node t Node t+1 ) with count, indicating that a carrier particle is from Node t-K Walk to Node one by one t+1 The trajectory of , where K represents the order of the rule, and 1<K<1 max -1, t+K+1≤l max , the node before the semicolon t Indicates the end node of the rule, the node after the semicolon t+1 Indicates the true value of the prediction node of this rule;

[0049] Step 104: assemble the continuous segments according to the node t Classify and obtain a collection Represents the set of high-order dependency rules corresponding to the node, that is, the set Contains all terminal nodes as Node t For a transportation network with N nodes, there are N sets of All the sets are the dynamic model of the transportation network

[0050] For example Figure 5 As shown, for the trajectory flows A, C, D, B, C, E, A, C, D, B, C, E, they have first-order dependency rules of (A; C) 2, (B; C) 2, (C; D) 2, (C; E) 2, (D; B) 2, and (E; A) 1; their second-order dependency rules are 1 (E, A; C), 2 (D, B; C), 2 (A, C; D), 2 (B, C; E), 2 (C, D; B), 2, (C, E; A) 1; their third-order dependency rules are 1 (C, E, A; C), 1 (C, D, B; C), 2 (E, A, C; D), 1 (D, B, C; E), 2 (A, C, D; B), 2 (B, C, E; A), 1 (B, C, E; A).

[0051] The second step of the prediction method in this embodiment is mainly aimed at the basic functions of the transportation network, predicting the flow distribution of the load layer at the next moment, that is, when it is predicted that a certain node will be the wandering target of more particles, the node importance should be higher. This embodiment is based on the maximum matching principle. This step uses the dynamic model constructed in the first step to analyze the trajectory flow of M carrier particles, where each carrier particle has equal importance. Figure 7 , the specific implementation process of the second step is:

[0052] Step 201, obtain the trajectory flow information of M carrier particles used for node importance prediction, and record the importance of each node in the transportation network as and initialized to 0, where N is the number of nodes in the transportation network;

[0053] Step 202: The trajectory flow information of the mth carrier particle is passed to the trajectory flow information of the past movement. Split into nodes And the length is 1 to l max -1 continuous sequence data, we get:

[0054]

[0055] , a total of l max -1 sequence and record it as a sequence set

[0056] Step 203: Set the sequence The longest sequence in In the dynamics model, nodes The corresponding high-order dependency rule set Match and judge the node Is there a sequence in the corresponding high-order dependency rule set? Matching higher-order dependency rules, such as and The match is successful:

[0057] If yes, proceed to step 204;

[0058] Otherwise, the sequence From the sequence collection and perform step 203 again until a The longest sequence successfully matched in represents the particle The highest-order dependency rules inherent in the walk;

[0059] Step 204, obtain sequence In the high-level dependency rule set All matching sequences in , and update the importance of the prediction nodes corresponding to each matching sequence. Assuming there are k (k = 1, 2, 3, ..., K) matching sequences, it can be expressed as:

[0060]

[0061] For the matched nodes Add its importance Wherein, g = 1 ~ K;

[0062] Step 205, repeating steps 202 to 204 until the trajectory flow analysis of M carrier particles is completed, the importance of each node in the transportation network can be obtained.

[0063] This embodiment regards each carrier particle as having an equal vote. By analyzing the high-order dependencies contained therein, the votes will be distributed to multiple possible target nodes obtained using the highest-order rule information. This distribution is unequal and depends on the probability of each possible target node becoming a real target. Ultimately, the votes obtained by each node in the transportation network represent the predicted importance of the node in the next load layer.

[0064] According to node importance Sort HOD Represents the predicted traffic distribution, by analyzing the Sort HOD That is, the prediction method in this embodiment is evaluated. The following further describes the trajectory flow dynamic node importance prediction method based on high-order dependency in this embodiment in combination with specific examples.

[0065] In this example, simulation trajectory data is generated by mixing first-order, second-order, third-order, and fourth-order dependency rules of similar proportions. Simulation experiments are carried out in the Python environment to verify the effectiveness of the proposed dynamic node importance ranking method. In the context of shipping, the data uses 100 ports as nodes, each of which has edges with 4 adjacent ports, and then generates trajectories of 100,000 ships moving between these ports according to certain high-order dependency rules. Each ship moves 100 steps, resulting in a total of 10,000,000 moves. This example takes the first 20 steps of the first 10,000 ships, a total of 200,000 moves corresponding to the trajectory flow for the experiment, and the total flow at each moment is 10,000 ships.

[0066] On the trajectory flow, a time window of length 10 is continuously slid to intercept the trajectory flow, such as 1 to 10, 2 to 11, ..., 10 to 19. The dynamic modeling in the first step of this embodiment is first used to calculate S Rule,j (j=1,2,3,...,10); then use S Rule,j The importance of each node is obtained by evaluating the second step of the method of this embodiment. And sort them from most important to least important to get Sort HOD,j (j=1,2,3,...,10). In addition, the comparison method is set up. First, the nodes are sorted according to the actual traffic of each node on the load layer in the next step, and the Sort Real,j(j=1,2,3,...,10); second, sorting by degree, PageRank and other indicators. At this time, since the topology of the simulation data is highly homogeneous, that is, each node is equally important, the virtual sorting Sort homo ,There are 100 virtual nodes, each of which corresponds to 1% of the actual traffic.,The experiment obtained a total of 21 rankings under different importance ranking methods.

[0067] Evaluation indicators for node importance ranking:

[0068] First, in order to quantify the traffic distribution represented by the node importance ranking, the indicator traffic ratio is calculated in, Indicates sorting ·,j The actual flow corresponding to the node selected by the previous ratio, flow all Indicates the total flow, which is a fixed value. For example, Sort = {Node1, Node2, ..., Node 10}, where each node corresponds to the flow {10,9,8,...,1}, flow all =55, select the nodes with ratio = 20% before Sort. In addition, in order to analyze more clearly and macroscopically, the average traffic share is calculated Indicates multiple groups of sorting obtained by a certain method ·,j The average of the actual traffic corresponding to the nodes selected by the previous ratio;

[0069] Secondly, in order to evaluate the effectiveness of the node importance prediction method, the accuracy index is calculated Indicates the sorting obtained by a certain method ·,j Sort the actual traffic distribution at the ratio Real,j The degree of fit is Sort Real,j The traffic ratio under the ratio. In addition, the average accuracy is calculated similarly It represents the mean value of the accuracy of the method under a certain ratio, and evaluates the effectiveness of the method in a more macroscopic way. or The closer it is to 1, the better the method fits the actual traffic distribution. Conversely, the lower it is, the worse the performance.

[0070] For experimental results and analysis:

[0071] First, use the average traffic share Analyze the sorting obtained by three node importance quantification methods HOD,j、Sort Real,j and Sort homo The corresponding traffic distribution. The average traffic ratio is Figure 8 As shown, it can be seen that compared to Sort homo , Sort HOD,j Can fit Sort better Real,j , which is the actual traffic distribution. At the same time, since the actual traffic distribution of each node is relatively uniform, the curve curvature is small, and the difference between the rankings is not obvious.

[0072] Secondly, the accuracy Analysis of the method Sort proposed in this embodiment HOD,j The effectiveness at different times provides a more comprehensive evaluation of the effectiveness of the method. Figure 8 ,It can be seen from the heat map drawn by the accuracy matrix that there are slight fluctuations at different times. ,The prediction effect of the method in this embodiment is relatively stable, and the accuracy is generally high.

[0073] Finally, calculate the sorting ratio under a certain ratio HOD,j The corresponding average accuracy Select ratio = 5%, 10%, 20%, 50%, 70% to observe how well the traffic distribution corresponding to the top ranking can fit the actual traffic distribution, and calculate They are 91.43%, 93.35%, 94.62%, 96.24% and 97.50% respectively, which quantifies the accuracy of the method of this embodiment under a certain proportion, and finds that a smaller proportion of nodes can better fit the corresponding actual traffic distribution.

[0074] According to the above experimental results, it can be seen that the effect difference at different times is small, the prediction effect is relatively stable, and a small proportion of sorting can better predict the actual traffic distribution, that is, the method of this embodiment can better predict the importance of nodes on the transportation network at the next moment. At the same time, the high-order rule extraction and importance sorting at each moment takes less than 1 minute, and the time cost is low. The method of this embodiment can quickly analyze the time series data in each time window.

[0075] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies, characterized in that: The steps include: Dynamically construct a kinetic model based on high-order dependencies in the trajectory flow; On the basis of the dynamic model, the importance of each node in the transportation network is predicted based on the trajectory flow information of the carrier particles in the past and the maximum matching principle.

2. According to claim 1, the method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies is characterized in that: The construction process of the kinetic model is as follows: Step 101, the trajectory flow of the carrier particle is represented as Among them, Node j represents the node where the carrier particle is located at time j, l max is the length of the trajectory flow, j = 1 ~ l max ; Step 102, split the trajectory flow of the carrier particle into multiple segments with lengths of 2, 3, ..., 1 max A continuous segment of length l max The continuous segment is the trajectory flow itself, with a length of l max There are 2 consecutive segments of length -1, and so on, there are l consecutive segments of length 2. max -1 item; Step 103, calculate the number of times each continuous segment appears in the continuous segment set after all trajectory streams are split, and record the continuous segment as (Node t-K ,Node t-K+1 ,…,Node t Node t+1 ) with count, indicating that a carrier particle is from Node t-K Walk to Node one by one t+1 The trajectory of , where K represents the order of the rule, and 1<K<1 max -1, t+K+1≤l max ; Step 104: group the continuous segments into nodes t Classify and obtain the set S Nodet , represents the set of high-order dependency rules corresponding to the node. For a transportation network with N nodes, there are N sets corresponding to it. All the sets are the dynamic model of the transportation network 3. The method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies according to claim 2 is characterized in that: Step 2 specifically includes: Step 201, obtain the trajectory flow information of M carrier particles used for node importance prediction, and record the importance of each node in the transportation network as and initialized to 0, where N is the number of nodes in the transportation network; Step 202: The trajectory flow information of the mth carrier particle is passed to the trajectory flow information of the past movement. Split into nodes And the length is 1 to l max -1 continuous sequence data, get l max -1 sequence and record it as a sequence set Step 203: Set the sequence The longest sequence in At the node The corresponding high-order dependency rule set Match and judge the node Does the corresponding high-order dependency rule set contain the same sequence? High-order dependency rules for matching points: If yes, proceed to step 204; Otherwise, the sequence From the sequence collection , and perform step 203 again; Step 204, obtain sequence In the high-level dependency rule set All matching sequences in , and the importance of the prediction nodes corresponding to each matching sequence is updated; Step 205, repeat steps 202 to 204 until the trajectory flow analysis of M carrier particles is completed and the node importance is obtained.

4. The method for predicting the importance of dynamic nodes in trajectory flow based on high-order dependencies according to claim 3 is characterized in that: Step 204 is specifically as follows: Get Sequence In the high-level dependency rule set All matching sequences in are represented as: For the matched nodes Add its importance Among them, g=1~K.

Citation Information

Patent Citations

  • Station importance evaluation method and device in rail transit

    CN112002127A

  • High-order analysis method for flight delay characteristics

    CN112182059A

  • Traffic OD passenger flow prediction method and system based on multi-channel hypergraph convolutional network

    CN117475638A

  • Near Online Multi-Target Tracking with Aggregated Local Flow Descriptor (ALFD)

    US20160132728A1

  • Interactive conflict detection and resolution for air and air-ground traffic control

    US8831864B1

Cited By

  • Trajectory flow node community partitioning and role recognition method and device based on dependency hypergraph

    CN122925948B