A method for constructing complex network of truck mobility with fusion evolution

Through the combination of Lorentz curve method and the maximum entropy principle, the problem of POI misjudgment and insufficient dynamic evolution in the construction of truck mobility networks in the prior art is solved, and a more accurate and dynamic truck mobility network analysis is achieved.

CN118797873BInactive Publication Date: 2024-11-22JILIN UNIVERSITY
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411289916.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-14
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing technology failed to fully consider urban construction environmental factors and truck POI attributes when building truck mobility networks, resulting in one-sided results of mobility analysis, misjudged the temporary stop point as POI, and failed to dynamically evolve the network structure, neglecting the POI between POIs. breaking conditions.

Method used

The Lorentz curve method is used to accurately identify truck POI, combine the maximum entropy principle to reconstruct the POI attraction model, introduce a break point model, and dynamically generate a truck maneuverability network through trajectory data and POI attributes to realize the maneuverability evolution mechanism.

Benefits of technology

Accurately identify truck POI, reduce misjudgment, dynamically adjust network structure, reflect the complexity of truck travel services, and improve the accuracy and reality of mobility analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118797873B_ABST
    Figure CN118797873B_ABST
Patent Text Reader

Abstract

The present invention belongs to the field of traffic spatiotemporal networks, and specifically relates to a method for constructing a complex network of truck mobility that integrates evolution, including: step A: obtaining truck trajectory data on the road and capturing POIs related to each city freight; step B: extracting the vehicle stop destination from the truck trajectory; step C: obtaining the mobility network of a single truck; step D: obtaining the city freight transportation POI topological network; step E: according to the spatial interaction theory, using the attributes of the city freight transportation POI topological network to calculate the attraction and break conditions between each POI; step F: obtaining the evolution mechanism of the truck mobility network and constructing the truck mobility network. The advantages of the present invention are: by accurately identifying the truck POI set based on the Lorenz curve method, reconstructing the truck POI attraction model based on the maximum entropy principle, and considering the dynamic changes in the attraction between truck POIs, introducing a breakpoint model, and simulating and realizing the dynamic evolution mechanism of truck travel service mobility.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of traffic spatiotemporal networks; specifically, it relates to a method for constructing a fused and evolved truck mobility complex network. Background Art

[0002] As the most basic mobile unit of urban freight, trucks undertake large-scale transportation tasks between functional places such as industrial enterprises, logistics warehouses, and comprehensive markets. Their trajectory data reflect the spatial connection between truck travel service activities in different regions and between different enterprises. On this basis, it is urgently necessary to study the mobility evolution mechanism of truck travel services from the perspective of complex network theory from the perspective of truck group mobility, so as to give full play to the important supporting role of truck entities in urban economic activities.

[0003] The definition, construction and analysis of truck mobility networks are key technical contents to reveal the core structure and dynamic evolution of urban freight systems, among which the construction method of truck mobility networks is the primary research basis. The truck mobility network defined in this study refers to the truck mobile network composed of truck POIs, which is generated by the dynamic changes in the attraction relationship between truck POIs and is generated by the addition or removal of new truck POIs. That is, the essence of the truck mobility network is a dynamic network whose nodes are truck POIs; it not only represents the attraction relationship between POIs, but also reflects the actual movement of trucks.

[0004] At present, truck mobility analysis mostly uses truck spatiotemporal trajectory prediction analysis, such as the published Chinese patents CN202010631885.0 "A generative adversarial network model, and a vehicle trajectory prediction method using the generative adversarial network model", CN202211472334.X "A two-stage trajectory generation method based on a generative adversarial network", etc., which only focus on the trajectory data of trucks, and do not consider the impact of urban built environment factors (such as road conditions and spatial distribution of logistics facilities) on traffic travel, road vitality, truck distribution needs, and the movement trajectory of logistics vehicles. There is a lack of comprehensive consideration and judgment of road section correlation data, etc., which leads to the truck mobility analysis results obtained by this method being overly dependent on trajectory data and one-sided. There is also a reported Chinese patent CN202011414792.9 that proposes "a method for extracting spatial agglomeration features of inbound logistics clusters", which is limited to the mining of spatial agglomeration features of trucks on urban road networks and cannot reveal the various characteristics of the complex network composed of truck clusters.

[0005] After comparative analysis of the above studies, this patent adopts the principle of complex network and proposes a method for constructing a complex network of truck mobility that integrates evolution. This method takes into account the physical network of urban roads and the network of truck transportation routes, extracts a set of truck POIs based on trajectory data, integrates the attraction model and the breakpoint model, and forms a network growth model that integrates evolution, realizing the dynamic generation of urban truck mobility networks, the addition and removal of new POI points, etc. After consulting similar research literature and patents, there are few reports.

[0006] Existing literature such as DOI: 10.1016 / j.tranpol.2012.06.011 "Compliance with and influence of a new in-car navigation system for trucks: Results of a field test.", DOI: 10.1016 / j.sbspro.2012.09.776 "Spatial environmental analysis on the effects of a new navigation system for freight transport." In identifying truck POIs, trajectory features such as speed, acceleration, and truck direction are often used to infer the motion state (stationary or moving) of heavy trucks within a given time period, and then the truck stop is determined based on the GPS trajectory. However, only considering the trajectory characteristics of the truck, it is easy to misjudge the temporary stop of the truck as the POI of the truck; some studies DOI: 10.3141 / 2494-10: "Estimation of statewide Origin-Destination truck flows from large streams of GPS data application for Florida Statewide Model." use the empirical data of truck parking to distinguish whether the stop is the actual POI of the truck. However, this method relies too much on historical experience data and is more subjective.

[0007] When constructing truck POI attraction models, existing literature, such as DOI: 10.2307 / 3144521 "A probabilistic analysis of shopping center trade areas.", DOI: 10.2307 / 2084520 "Intervening opportunities: A theory relating mobility and distance.", DOI: 10.1111 / j.1435-5597.1959.tb01665.x " Gravity models and trip distribution theory.", only considers the distance between POIs and the flow of POIs, and fails to consider the impact of truck POI attributes on the attractiveness of truck POIs.

[0008] Existing literature such as DOI: 10.1016 / j.trc.2023.104442 “Structure and dynamics of urban freight truck movements: A complex network theory perspective” adds truck POIs to the truck mobility network one by one according to the attraction relationship when constructing the dynamic evolution mechanism of the truck mobility network, but fails to consider that as nodes are continuously introduced, the fracture conditions between POIs exceed the fracture threshold, and the connection between POIs will be broken. Summary of the invention

[0009] In view of the above problems, the purpose of the present invention is to provide a method for constructing a fused and evolving complex network of truck mobility, by accurately identifying the truck POI set based on the Lorenz curve method, considering the differences in truck POI attributes, reconstructing the truck POI attraction model based on the maximum entropy principle, and considering the dynamic changes in the attraction between truck POIs. A breakpoint model is introduced to simulate and realize the dynamic evolution mechanism of truck travel service mobility, so as to overcome the shortcomings of the above-mentioned prior art.

[0010] To achieve the above object, the present invention adopts the following technical solution:

[0011] A method for constructing a complex network of truck mobility that integrates evolution, characterized by comprising the following steps:

[0012] Step A: Obtain truck trajectory data on the road and capture POIs related to freight in each city, where the truck trajectory data includes the vehicle ID, longitude, latitude, GPS time, and speed, and the POI data related to freight in the city includes the POI ID, name, longitude, latitude, and attributes;

[0013] Step B: Extract vehicle stop destinations from truck trajectories;

[0014] Step C: Match each vehicle’s stop destination with the POI to obtain the mobility network of a single truck;

[0015] Step D: Aggregate the trajectory network of a single truck and cluster the POIs of the trajectory network using the adaptive DBSCAN method to obtain the POI topology network of urban freight transportation;

[0016] Step E: According to the spatial interaction theory, the properties of the urban freight transportation POI topological network are used to calculate the attraction and break conditions between POIs;

[0017] Step F: Using the attraction and fracture conditions, the evolution mechanism of the truck mobility network is obtained, and the truck mobility network is constructed according to this mechanism.

[0018] As the preferred embodiment of the present invention, step B also includes the following steps:

[0019] Step B1: extract the position when the vehicle speed is 0 from the truck trajectory, that is, the truck stop position;

[0020] Step B2: Determine the parking time threshold;

[0021] Step B2.1: Arrange the truck stop locations in descending order of parking time, calculate the cumulative parking time of heavy trucks at the ordered stops, normalize the ordinal numbers of the ordered truck stops and the corresponding cumulative parking time, and draw the Lorenz curve L1:

[0022] Step B2.2: Calculate the intersection point F* of the tangent line of the Lorenz curve at (1, 1) and the horizontal axis. The long-term truck stop with a standardized ordinal greater than F* is classified as the first type of stop. Remove the first type of stop and repeat step B2.1 to obtain the Lorenz curve L2. The intersection point F*' of the tangent line of L2 at (1, 1) and the horizontal axis. The long-term truck stop with a standardized ordinal greater than F*' is classified as the second type of stop. Remove the second type of stop again and repeat step B2.1 to obtain the Lorenz curve L3. The intersection point F*'' of the tangent line of L3 at (1, 1) and the horizontal axis. The long-term truck stop with a standardized ordinal greater than F*'' is classified as the third type of stop. The parking time corresponding to the third type of stop is the parking time threshold.

[0023] Step B3: Compare the parking time of all parking spots with the parking time threshold, and regard the parking spots whose parking time is less than the time threshold as temporary parking spots and delete them, thereby obtaining the vehicle parking destination.

[0024] As the preferred embodiment of the present invention, step C also includes the following steps:

[0025] Step C1: Match the truck stop destination obtained in step B with the freight-related POI;

[0026] Step C1.1: The distance formula between two points in WGS84 coordinate system;

[0027] In the coordinate system, let the actual distance between points A and B be ,but

[0028] ;

[0029] In the formula, The radius of the Earth is 6371km; The unit is km; is the latitude A; is the B latitude; is the longitude of A; is the longitude of B;

[0030] Step C1.2: Calculate the truck stop points using the formula described in Step C1.1 The distance between each POI is recorded as ;

[0031] Stops The Haversine distance between any POI is: ;

[0032] The meanings of the parameters in the formula are the same as those in step C1.1. The radius of the Earth is 6371km; The unit is km; For Node i latitude; For Node j latitude; For Node i longitude; for j longitude;

[0033] Step C1.3: Utilization Construct the distance matrix between vehicle stops and POIs;

[0034] ;

[0035] The meanings of the parameters in the formula are as follows: Represents the distance matrix between the vehicle stop and the POI, where Represents when the vehicle is parked at position 1 The distance between Represents when the vehicle is parked at position 1 The distance between Represents the vehicle parked at position m and The distance between Represents the vehicle parked at position m and The distance between; Step C1.4: Take the matrix Minimum value The corresponding subscript ,but The truck is at the stop The actual destination of the truck can be recorded as ,but

[0036] ;

[0037] In the formula, L represents the actual trajectory of the vehicle, that is, the actual trajectory of the vehicle is: ,in, When the actual stop point of the vehicle is 1, its actual destination stop point is POI points, When the actual stop point of the vehicle is 2, its actual destination stop point is POI points, and so on. When the actual stop point of the vehicle is m, its actual destination stop point is POI points.

[0038] As the preferred embodiment of the present invention, step D also includes the following steps:

[0039] Step D1: Aggregate the trajectory network of all trucks;

[0040] Step D2: Cluster POIs with similar distances according to their geographical location using the DBSCAN method;

[0041] Step D2.1: Calculate the distance between any two POIs in the network graph obtained in step D1 and obtain the distance distribution matrix ;

[0042] ;

[0043] In the formula, is the distance distribution matrix between each POI, yes and The distance between yes and The distance between yes and The distance between yes and distance;

[0044] Step D2.2: Sort the elements of each row in ascending order, and get the matrix , then the distance vector composed of the elements in the first column is Indicates the distance from the object to itself, all are 0, The elements of the column constitute the -Nearest neighbor distance vector ;

[0045] Step D2.3: Vector Taking the average of the elements in , we can get the vector of -Average nearest neighbor distance , and use it as the candidate neighborhood radius Parameters, for all Calculate the value, and you get Parameter List , expressed as:

[0046] ;

[0047] Step D2.4: Parameter list The intersection of the actual buffer radius of the POI is taken to obtain the final neighborhood radius. Parameters, denoted as ,but

[0048] ;

[0049] In the formula, is the actual buffer radius of the POI;

[0050] Step D2.5: Generate neighborhood density parameter;

[0051] Generated using mathematical expectation method Parameters, then

[0052] ;

[0053] In the formula, For the i object The number of neighboring objects,n is the total number of POIs matching the truck trajectory;

[0054] Step D2.6: The parameters obtained in step D2.4 , the parameters obtained in step D2.5 As a DBSCAN parameter, the “Minkowski distance” in the classic DBSCAN algorithm is replaced with the Haversine distance in step C1.1, and clustering and topology are performed to obtain the urban freight transportation POI topology network.

[0055] As a preferred embodiment of the present invention, step E also includes the following steps:

[0056] Step E1: using the maximum entropy model method to obtain the attraction between each POI;

[0057] Step E1.1: Feature extraction,

[0058] Extract various attributes of POI, specifically attributes, geographical location, and distance between POIs, as feature attributes, denoted as

[0059] ;

[0060] In the formula, is a collection of various characteristic attribute functions; Characteristic attributes The corresponding function is used to describe the characteristic attributes The attraction of interconnection between POIs; This means that in the feature attribute Under the effect of right The attractiveness of Characteristic attributes Under the effect of right The attractiveness of Characteristic attributes Under the effect of right The attractiveness of Characteristic attributes Under the effect of right The attraction effect; Step E1.2: Construct a constraint set;

[0061] Step E1.2.1: Using the topological network obtained in step D, obtain the statistics based on the existing data and The joint probability distribution of ,as well as The probability distribution of ,but

[0062] ;

[0063] ;

[0064] In the formula, express and The number of mutual attractions, represents the total number of attraction of the node, Indicates the total number of nodes; According to the existing data statistics and The joint probability distribution of According to the existing data statistics The probability distribution of ; Representation Node The number of nodes is brought into this problem. Since there is only one node per node, the value is 1.

[0065] Step E1.2.2: Determine the constraints;

[0066] ;

[0067] ;

[0068] In the formula, is the expected impact of the characteristic attribute on the attractiveness of the POI, which should be as close as possible to , The impact of various characteristic attributes on the attractiveness between POIs in reality; Represents the probability of node x attracting node y. The sum of the attraction probabilities of node x to all nodes should be 1.

[0069] Step E1.2.3: Establish a connection between the constraints and the attraction between POIs.

[0070] ;

[0071] ;

[0072] ;

[0073] ;

[0074] In the formula, For theory and The joint probability distribution of Based on the existing data statistics and The joint probability distribution of for attract The probability distribution of For theory The probability distribution of , According to the existing data statistics and The joint probability distribution of According to the existing data statistics The probability distribution of ;

[0075] Step E1.3: Construct the target entropy function,

[0076] Maximizing entropy means allocating the most uncertainty and not making any unnecessary assumptions about unknown information. It is usually regarded as the most fair and least biased choice. Let the target entropy function be ,but

[0077] ;

[0078] In the formula, is the total entropy generated by the mutual attraction of POIs; for right attractiveness; for Logarithm to base e; In theory Probability distribution in the entire POI set;

[0079] Step E1.4: Model preparation,

[0080] Step E1.4.1: Objective function,

[0081] ;

[0082] Step E1.4.2: Constraints,

[0083] ;

[0084] ;

[0085] ;

[0086] ;

[0087] ;

[0088] ;

[0089] ;

[0090] ;

[0091] ;

[0092] Step E1.4.3: Existing training dataset,

[0093] ;

[0094] In the formula, Represents the set of all POI nodes in the area. Represents the first POI, Represents the second POI, and so on. represents the nth POI;

[0095] Solve the objective function of step E1.4.1, the constraints of step E1.4.2, and the model composed of the data set of step E1.4.3 to obtain the attraction between each POI;

[0096] Step E2: Break conditions between POIs;

[0097] Step E2.1: Determine the attributes of each POI;

[0098] Step E2.2: Calculate the break conditions between each POI and the logistics company;

[0099] ;

[0100] In the formula, is the distance between two competing logistics parks, and The influencing factors are The weight of impact on the service scope of two competing logistics parks; and The influencing factors are Quantify the impact on two competing logistics parks; is the number of influencing factors, where the influencing factors include the Haversine distance between the POI and the logistics company, the actual cargo demand of each POI, and the park scale of the logistics company;

[0101] Step E2.3: Obtain the fracture matrix of each logistics company;

[0102] Arranging the break conditions of any two points obtained in step E2.2, we can obtain the break matrix in the entire network,

[0103] ;

[0104] This matrix is ​​a symmetric matrix, and all elements on the main diagonal are 0;

[0105] In the formula, is the fracture matrix of the network, for and The breaking value, for and The breaking value, for and The breaking value, for and The breaking value of

[0106] Step E2.4: Determine the fracture threshold , the matrix Each element and For comparison, Less than , then the node i With Node j Perform fracture treatment.

[0107] As the preferred embodiment of the present invention, step F also includes the following steps:

[0108] Step F1: Determine the initialization network node;

[0109] Step F2: randomly connect the edges of the initialized network; by initializing the edges of the network, simulate the existing connection mode of each freight-related POI;

[0110] Step F3: randomly introduce a freight POI into the network;

[0111] Step F3.1: Introduce a POI into the network and treat it as a new node;

[0112] Step F3.2: Based on the properties of this new node, calculate the attraction and break conditions between this node and all nodes in the network;

[0113] Step F3.3: Determine how many nodes in the network this new node should be connected to based on the attraction obtained in step F3.2;

[0114] Step F3.4: Determine the connection edges between the existing node and several nodes that should be broken according to the breaking conditions obtained in step F3.2:

[0115] Step F3.4: Based on the attraction obtained in step F3.2, use the wheel method to determine the connection and break conditions;

[0116] Step F4: Repeat the process of step F3 until the truck mobility network is constructed.

[0117] The advantages and positive effects of the present invention are:

[0118] 1. The present invention aims to accurately identify the truck POI problem before constructing a complex truck mobility network. If the temporary stop of the truck is mistakenly regarded as a truck POI, a large number of nodes that are irrelevant to the actual operation of the truck will appear in the complex truck mobility network, which will cause the established complex truck mobility network to obtain incorrect truck trajectory operation rules and the inherent movement rules of the urban freight system. By accurately identifying the truck POI set based on the Lorenz curve method, in the process of the trajectory topology being a network structure, the Lorenz curve method is used to determine the parking time threshold, and by comparing the parking time with the parking time threshold, the temporary stop points in the truck parking points are screened and eliminated to ensure that there are no temporary stop points in the truck stop positions screened by the truck. Thereby, the phenomenon of POIs that are not related to freight being misjudged as POIs related to freight is eliminated.

[0119] 2. The present invention aims at the mutual attraction relationship between truck POIs, which determines which POIs will cooperate with each other, and then which POIs will have truck driving tracks. The attribute category, distance, and flow of POIs will affect the attraction between POIs. By considering the attribute differences of truck POIs, the truck POI attraction model is reconstructed based on the maximum entropy principle. When constructing network nodes, the attributes of freight-related POIs are randomly generated according to the actual proportion, and the maximum entropy model of truck POI attraction is established. By solving the model, the attraction between POIs with various attributes is obtained, thereby accurately describing the attraction relationship between truck POIs with various attributes.

[0120] 3. The present invention aims to solve the problem that as the network evolves, the attraction relationship between truck POIs will also change. When the attraction between POIs is too low, the "cooperation" between POIs may be "temporarily interrupted". By considering the dynamic changes in the attraction between truck POIs, a breakpoint model is introduced to simulate and realize the dynamic evolution mechanism of truck travel service mobility. The breakpoint principle is introduced to analyze the changes in the mutual attraction between POIs during the evolution of the truck mobility network. If the attraction between two POIs decreases and is lower than the break threshold, the two POIs temporarily interrupt their cooperation until their attraction meets the connection requirements again and they resume cooperation. This process reflects the process of each node cooperating, terminating and re-cooperating with freight companies in reality. BRIEF DESCRIPTION OF THE DRAWINGS

[0121] Figure 1 It is the technical roadmap of the present invention;

[0122] Figure 2 It is a technical flow chart of extracting a vehicle stop destination according to the present invention;

[0123] Figure 3 It is a technical flow chart of the trajectory network POI clustering of the present invention;

[0124] Figure 4 It is a schematic diagram of calculating the attraction between different POIs, obtaining a network evolution mechanism and establishing a truck mobility network according to the present invention;

[0125] Figure 5 It is a distance calculation diagram under the WGS84 coordinate system of the present invention;

[0126] Figure 6 is a schematic diagram of the relationship between the vehicle stop points and POIs of the present invention, wherein the marks 1-8 in the figure are POIs;

[0127] Figure 7 It is a POI network diagram obtained from two vehicle trajectory data of the present invention;

[0128] Figure 8 is the POI network topology diagram before clustering of the present invention;

[0129] Fig. 9 is the POI network topology diagram after clustering of the present invention;

[0130] Fig.10 is a degree distribution image of the network structure of the present invention;

[0131] Fig.11 is a schematic diagram of a truck parking node network obtained without removing temporary vehicle parking points of the present invention;

[0132] Fig.12 is a schematic diagram of a truck node network obtained after screening by a parking time threshold value of the present invention;

[0133] Fig.13 is a schematic diagram of a truck mobility network obtained by directly topologically performing truck trajectories of the present invention;

[0134] Fig.14 is a schematic diagram of a truck mobility network obtained by considering only the mutual attraction between logistics companies of the present invention;

[0135] Fig.15 It is the truck mobility network generated by the method of the present invention. DETAILED DESCRIPTION

[0136] Example 1

[0137] See also Figure 1-9 This embodiment provides a method for constructing a complex network of truck mobility that integrates evolution, including the following steps:

[0138] Step A: Obtain truck trajectory data on the road and capture POIs (areas of interest) related to freight in each city. Truck trajectory data includes vehicle ID, longitude, latitude, GPS time, and speed. POI data related to freight in the city includes POI ID, name, longitude, latitude, and attributes.

[0139] Step B: Extract vehicle stop destinations from truck trajectories;

[0140] Step B1: extract the position when the vehicle speed is 0 from the truck trajectory, that is, the truck stop position;

[0141] Step B2: Determine the parking time threshold;

[0142] Step B2.1: Arrange the truck stop locations in descending order of parking time, calculate the cumulative parking time of heavy trucks at the ordered stops, normalize the ordinal numbers of the ordered truck stops and the corresponding cumulative parking time, and draw the Lorenz curve L1:

[0143] Step B2.2: Calculate the intersection point F* of the tangent line of the Lorenz curve at (1, 1) and the horizontal axis. The long-term truck stop with a standardized ordinal greater than F* is classified as the first type of stop. Remove the first type of stop and repeat step B2.1 to obtain the Lorenz curve L2. The intersection point F*' of the tangent line of L2 at (1, 1) and the horizontal axis. The long-term truck stop with a standardized ordinal greater than F*' is classified as the second type of stop. Remove the second type of stop again and repeat step B2.1 to obtain the Lorenz curve L3. The intersection point F*'' of the tangent line of L3 at (1, 1) and the horizontal axis. The long-term truck stop with a standardized ordinal greater than F*'' is classified as the third type of stop. The parking time corresponding to the third type of stop is the parking time threshold.

[0144] Step B3: Compare the parking time of all parking spots with the parking time threshold, and regard the parking spots with parking time less than the time threshold as temporary parking spots and delete them, thereby obtaining the vehicle parking destination;

[0145] Step C: Match each vehicle’s stop destination with the POI to obtain the mobility network of a single truck;

[0146] Step C1: Match the truck stop destination obtained in step B with the freight-related POI;

[0147] Step C1.1: The distance formula between two points in WGS84 coordinate system;

[0148] In the coordinate system, let the actual distance between points A and B (also known as Haversine distance) be ,but

[0149] ;

[0150] In the formula, The radius of the Earth is 6371km; The unit is km; is the latitude A; is the B latitude; is the longitude of A; is the longitude of B;

[0151] Step C1.2: Calculate the truck stop points using the formula described in Step C1.1 The distance between each POI is recorded as

[0152] Stops The Haversine distance between any POI is:

[0153] ;

[0154] The meanings of the parameters in the formula are the same as those in step C1.1. The radius of the Earth is 6371km; The unit is km; For Node i latitude; For Node j latitude; For Node i longitude; for j longitude;

[0155] Step C1.3: Utilization Construct the distance matrix between vehicle stops and POIs;

[0156] ;

[0157] The meanings of the parameters in the formula are as follows: Represents the distance matrix between the vehicle stop and the POI, where Represents when the vehicle is parked at position 1 The distance between Represents when the vehicle is parked at position 1 The distance between Represents the vehicle parked at position m and The distance between Represents the vehicle parked at position m and The distance between; Step C1.4: Take the matrix Minimum value The corresponding subscript ,but The truck is at the stop The actual destination of the truck can be recorded as ,but

[0158] ;

[0159] In the formula, L represents the actual trajectory of the vehicle, that is, the actual trajectory of the vehicle is: ,in, When the actual stop point of the vehicle is 1, its actual destination stop point is POI points, When the actual stop point of the vehicle is 2, its actual destination stop point is POI points, and so on. When the actual stop point of the vehicle is m, its actual destination stop point is POI points;

[0160] Step D: Aggregate the trajectory network of a single truck and cluster the POIs of the trajectory network using the adaptive DBSCAN method to obtain the POI topology network of urban freight transportation;

[0161] Step D1: Aggregate the trajectory network of all trucks.

[0162] Step D2: Cluster POIs with similar distances according to their geographical location using the DBSCAN method;

[0163] Step D2.1: Calculate the distance between any two POIs in the network graph obtained in step D1 and obtain the distance distribution matrix ;

[0164] ;

[0165] In the formula, is the distance distribution matrix between each POI, yes and The distance between yes and The distance between yes and The distance between yes and distance;

[0166] Step D2.2: Sort the elements of each row in ascending order, and get the matrix , then the distance vector composed of the elements in the first column is Indicates the distance from the object to itself, all are 0, The elements of the column constitute the -Nearest neighbor distance vector ;

[0167] Step D2.3: Vector Taking the average of the elements in , we can get the vector of -Average nearest neighbor distance , and use it as the candidate neighborhood radius Parameters, for all Calculate the value, and you get Parameter List , expressed as:

[0168] ;

[0169] Step D2.4: Parameter list The intersection of the actual buffer radius of the POI is taken to obtain the final neighborhood radius. Parameters, denoted as ,but

[0170] ;

[0171] In the formula, is the actual buffer radius of the POI;

[0172] Step D2.5: Generate neighborhood density parameter;

[0173] Generated using mathematical expectation method Parameters, then

[0174] ;

[0175] In the formula, For the i object The number of neighboring objects, n is the total number of POIs matching the truck trajectory;

[0176] Step D2.6: The parameters obtained in step D2.4 , the parameters obtained in step D2.5 As a DBSCAN parameter, the "Minkowski distance" in the classic DBSCAN algorithm is replaced with the Haversine distance in step C1.1, and clustering and topology are performed to obtain the urban freight transportation POI topology network;

[0177] Step E: According to the spatial interaction theory, the properties of the urban freight transportation POI topological network are used to calculate the attraction and break conditions between POIs;

[0178] Step E1: using the maximum entropy model method to obtain the attraction between each POI;

[0179] Step E1.1: Feature extraction,

[0180] Extract various attributes of POI, specifically attributes, geographical location, and distance between POIs, as feature attributes, denoted as

[0181] ;

[0182] In the formula, is a collection of various characteristic attribute functions, Characteristic attributes The corresponding function is used to describe the characteristic attributes The attraction of interconnection between POIs, e.g. This means that in the feature attribute Under the effect of right The attractiveness of Characteristic attributes Under the effect of right The attractiveness of Characteristic attributes Under the effect of right The attractiveness of Characteristic attributes Under the effect of right The attraction effect; Step E1.2: Construct a constraint set;

[0183] Step E1.2.1: Using the topological network obtained in step D, obtain the statistics based on the existing data and The joint probability distribution of ,as well as The probability distribution of ,but

[0184] ;

[0185] ;

[0186] In the formula, express and The number of mutual attractions, represents the total number of attraction of the node, Indicates the total number of nodes; According to the existing data statistics and The joint probability distribution of According to the existing data statistics The probability distribution of ; Representation Node The number of nodes is brought into this problem. Since there is only one node per node, the value is 1.

[0187] Step E1.2.2: Determine the constraints;

[0188] ;

[0189] ;

[0190] In the formula, is the expected impact of the characteristic attribute on the attractiveness of the POI, which should be as close as possible to , The impact of various characteristic attributes on the attractiveness between POIs in reality; Represents the probability of node x attracting node y. The sum of the attraction probabilities of node x to all nodes should be 1.

[0191] Step E1.2.3: Establish a connection between the constraints and the attraction between POIs.

[0192] ;

[0193] ;

[0194] ;

[0195] ;

[0196] In the formula, For theory and The joint probability distribution of Based on the existing data statistics and The joint probability distribution of for attract The probability distribution of For theory The probability distribution of , According to the existing data statistics and The joint probability distribution of According to the existing data statistics The probability distribution of ;

[0197] Step E1.3: Construct the target entropy function,

[0198] Maximizing entropy means allocating the most uncertainty and not making any unnecessary assumptions about unknown information. It is usually regarded as the most fair and least biased choice. Let the target entropy function be ,but

[0199] ;

[0200] In the formula, is the total entropy generated by the mutual attraction of POIs; for right attractiveness; for Logarithm to base e; In theory Probability distribution in the entire POI set;

[0201] Step E1.4: Model preparation,

[0202] Step E1.4.1: Objective function,

[0203] ;

[0204] Step E1.4.2: Constraints,

[0205] ;

[0206] ;

[0207] ;

[0208] ;

[0209] ;

[0210] ;

[0211] ;

[0212] ;

[0213] ;

[0214] Step E1.4.3: Existing training dataset,

[0215] ;

[0216] In the formula, Represents the set of all POI nodes in the area. Represents the first POI, Represents the second POI, and so on. represents the nth POI;

[0217] Solve the objective function of step E1.4.1, the constraints of step E1.4.2, and the model composed of the data set of step E1.4.3 to obtain the attraction between each POI;

[0218] Step E2: Break conditions between POIs;

[0219] Step E2.1: Determine the attributes of each POI;

[0220] Step E2.2: Calculate the break conditions between each POI and the logistics company;

[0221] ;

[0222] In the formula, is the distance between two competing logistics parks, and The influencing factors are The weight of impact on the service scope of two competing logistics parks; and The influencing factors are Quantify the impact on two competing logistics parks; is the number of influencing factors, where the influencing factors include the Haversine distance between the POI and the logistics company, the actual cargo demand of each POI, and the park scale of the logistics company;

[0223] Step E2.3: Obtain the fracture matrix of each logistics company;

[0224] Arranging the break conditions of any two points obtained in step E2.2, we can obtain the break matrix in the entire network,

[0225] ;

[0226] This matrix is ​​a symmetric matrix, and all elements on the main diagonal are 0;

[0227] In the formula, is the fracture matrix of the network, for and The breaking value, for and The breaking value, for and The breaking value, for and The breaking value of

[0228] Step E2.4: Determine the fracture threshold , the matrix Each element and For comparison, Less than , then the node i With Node j Perform fracture treatment.

[0229] Step F: Using the attraction and fracture conditions, the evolution mechanism of the truck mobility network is obtained, and the truck mobility network is constructed according to this mechanism;

[0230] Step F1: Determine the initialization network node;

[0231] Step F2: randomly connect the edges of the initialized network; by initializing the edges of the network, simulate the existing connection mode of each freight-related POI;

[0232] Step F3: randomly introduce a freight POI into the network;

[0233] Step F3.1: Introduce a POI into the network and treat it as a new node;

[0234] Step F3.2: Based on the properties of this new node, calculate the attraction and break conditions between this node and all nodes in the network;

[0235] Step F3.3: Determine how many nodes in the network this new node should be connected to based on the attraction obtained in step F3.2;

[0236] Step F3.4: Determine the connection edges between the existing node and several nodes that should be broken according to the breaking conditions obtained in step F3.2:

[0237] Step F3.4: Based on the attraction obtained in step F3.2, use the wheel method to determine the connection and break conditions;

[0238] Step F4: Repeat the process of step F3 until the truck mobility network is constructed.

[0239] Example 2

[0240] The following is a truck mobility network constructed using truck trajectory data from a certain city in August 2022 to illustrate the contribution of this method in constructing a truck mobility network.

[0241] First, the model fully explains the real truck trajectories.

[0242] The structure of complex network models, such as nodes, edges, and weights, can intuitively represent different entities (such as vehicles, roads) in the transportation system and the relationships between them. The weight of the network model reflects the Haversine clustering of POIs in the truck network, and the degree distribution of the network model represents the structural characteristics of the truck mobility network.

[0243] For example, Fig.10 It shows that the degree distribution of a city's truck mobility network conforms to the power-law distribution characteristics. A network with a power-law degree distribution is called a scale-free network. There are a small number of nodes connected to a large number of nodes in the network, which indicates that a small number of POIs in the truck mobility network are hub nodes, and a large number of trucks need to pass through these "hub nodes" to transport goods, which corresponds to the layout of the logistics distribution center in Changchun. By analyzing the various attributes of the truck mobility network, the complex interactions and dynamic changes of trucks during the transportation process can be captured.

[0244] 2. When matching truck stops with POIs, a large number of temporary stops were removed.

[0245] The construction method in Example 1 is compared with the matching method that fails to remove temporary vehicle stops. Fig.11 The truck parking node network diagram obtained by removing the temporary vehicle stops in the reference Fig.12 The truck node network obtained after parking time threshold screening (this method).

[0246] Comparison Fig.11 and Fig.12 ,It can be found that by removing a large number of temporary parking ,location nodes during the driving process of trucks through this ,method, the obtained network nodes can more realistically ,reflect the distribution status of freight related POIs.

[0247] contrast Fig.12 and Fig.13 , see Fig.13 The purpose is to directly topology the truck trajectories and obtain a schematic diagram of the truck mobility network; Fig.12Through the operation of step B in the present invention, all vehicle stops after removing temporary vehicle stops are obtained, that is, the truck node network diagram obtained after screening by the parking time threshold; in reality, the vehicle stops remaining after removing all temporary stops are truck points of interest (POIs), and the network formed by the truck's driving trajectory passing through these POI points in reality is the truck mobility network; therefore, it can be said that Fig.12 The resulting schematic diagram of all vehicle stops is a truck mobility network diagram obtained by directly topologically analyzing truck trajectories.

[0248] 3. In terms of network generation effect.

[0249] See also Fig.13 The truck mobility network diagram obtained by directly topologically analyzing the truck trajectories in the figure. Fig.14 Schematic diagram of the truck mobility network obtained by considering only the mutual attraction between logistics companies. Fig.15 Schematic diagram of the truck mobility network generated by this method.

[0250] contrast Fig.13 , Fig.14 , Fig.15 It can be seen that only considering the attraction between logistics companies and ignoring freight-related POIs with other attributes will cause the logistics companies in the network to attract each other, and other nodes will spontaneously move closer to the logistics companies, which will lead to the generated truck mobility network as a whole. This is inconsistent with the situation in which the truck mobility network in reality is composed of several disconnected sub-networks (in reality, some freight-related POIs will form a small community according to their own freight needs. When the community becomes larger, it will break and decompose into several small communities, which will aggregate and break again, and repeat the cycle); the network generated by this method calculates the attraction between POIs of various attributes based on the maximum entropy model of POI attraction relationship, and comprehensively considers the spatial attraction of freight-related POIs with various attributes. Therefore, the generated truck mobility network is more consistent with the existing truck trajectory network.

[0251] The above description is only a preferred example of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and variations. Any modification, equivalent substitution, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a complex network of truck mobility with fusion evolution, It is characterized in that The following steps are involved: Step A: Obtain truck trajectory data on the road and capture POIs related to freight in each city, where the truck trajectory data includes the vehicle ID, longitude, latitude, GPS time, and speed, and the POI data related to freight in the city includes the POI ID, name, longitude, latitude, and attributes; Step B: Extract vehicle stop destinations from truck trajectories; Step C: Match each vehicle’s stop destination with the POI to obtain the mobility network of a single truck; Step D: Aggregate the trajectory network of a single truck and cluster the POIs of the trajectory network using the adaptive DBSCAN method to obtain the POI topology network of urban freight transportation; Step E: According to the spatial interaction theory, the properties of the urban freight transportation POI topological network are used to calculate the attraction and break conditions between POIs; Step E1: using the maximum entropy model method to obtain the attraction between each POI; Step E1.1: Feature extraction, Extract various attributes of POI, specifically attributes, geographical location, and distance between POIs, as feature attributes, denoted by F = {f 1 (x,y),f 2 (x,y),…,f i (x,y)…f n (x,y)} Where F is the set of various characteristic attribute functions; f i is the function corresponding to the feature attribute i, which is used to describe the attraction of the feature attribute i to the connection between POIs; 1 (x,y) means that in the feature attribute f 1 Under the effect of x POI y The attraction effect; f 2 (x,y) is the feature attribute f 1 Under the effect of x POI y The attraction effect; f i (x,y) is the feature attribute f 1 Under the effect of x POI y The attraction effect; f n (x,y) is the feature attribute f 1 Under the effect of x POI y The attraction effect; Step E1.2: Construct a constraint set; Step E1.2.1: Using the topological network obtained in step D, obtain the POI statistics based on the existing data x With POI y The joint probability distribution of and POI x The probability distribution of but Where, v(x=POI x , y = POI y ) indicates POI x With POI y The number of mutually attractive points, N(POI x , POI y ) represents the total number of nodes attracted, N(POI) represents the total number of nodes; To count POIs based on existing data x With POI y The joint probability distribution of To count POIs based on existing data x The probability distribution of v(x=POI x ) represents the node POI x The number of nodes is brought into this problem. Since there is only one node per node, the value is 1. Step E1.2.2: Determine the constraints; In the formula, E P (fi) is the expected impact of the characteristic attribute on the attractiveness of the POI, which should be as close as possible to The influence of various characteristic attributes on the attraction between POIs in reality; P(y|x) represents the probability of node x attracting node y, and the sum of the attraction probabilities of node x to all nodes should be 1; Step E1.2.3: Establish a connection between the constraints and the attraction between POIs. P(x,y)=P(x)P(y|x) Where P(x,y) is the theoretical POI x With POI y The joint probability distribution of f(x,y) is the POI obtained based on the existing data statistics. x With POI y The joint probability distribution of , P(y|x) is the POI x Attract POI y The probability distribution of P(x) is the theoretical POI x The probability distribution of To count POIs based on existing data x With POI y The joint probability distribution of To count POIs based on existing data x The probability distribution of Step E1.3: Construct the target entropy function, Let the target entropy function be H(P), then Where H(P) is the total entropy generated by the mutual attraction of POIs; P(y j |x i ) is POI x POI y The attraction of logP(y j |x i ) is P(y j |x i ) logarithm to base e; P(x i ) is the theoretical POI x Probability distribution in the entire POI set; Step E1.4: Model preparation, Step E1.4.1: Objective function, Step E1.4.2: Constraints, P(x,y)=P(x)P(y|x) f i ∈F={f 1 ,f 2 …f n } Step E1.4.3: Existing training dataset, T={THEN 1 ,THEN 2 …THEN n } In the formula, T represents the set of all POI nodes in the region, POI 1 Represents the first POI, POI 2 Represents the second POI, and so on. n represents the nth POI; Solve the objective function of step E1.4.1, the constraints of step E1.4.2, and the model composed of the data set of step E1.4.3 to obtain the attraction between each POI; Step E2: Break conditions between POIs; Step E2.1: Determine the attributes of each POI; Step E2.2: Calculate the break conditions between each POI and the logistics company; Where, d 1,2 is the distance between two competing logistics parks, α i With β i are the influence weights of factor i on the service scope of the two competing logistics parks; X i With Y i are the quantification of the impact of influencing factor i on the two competing logistics parks; m is the number of influencing factors, where the influencing factors include the Haversine distance between POI and logistics company, the actual cargo demand of each POI, and the park scale of the logistics company; Step E2.3: Obtain the fracture matrix of each logistics company; Arranging the break conditions of any two points obtained in step E2.2, we can obtain the break matrix in the entire network, This matrix is ​​a symmetric matrix, and all elements on the main diagonal are 0; Where D n×n is the fracture matrix of the network, d 1-1 For POI 1 With POI 1 The fracture value, d 1-n For POI 1 With POI n The fracture value, d n-1 For POI n With POI 1 The fracture value, d n-n For POI n With POI n The breaking value of Step E2.4: Determine the fracture threshold d * , the matrix D n×n Each element (d ij ) and d * For comparison, such as (d ij ) is less than d * , then node i and node j are fractured; Step F: Using the attraction and fracture conditions, the evolution mechanism of the truck mobility network is obtained, and the truck mobility network is constructed according to this mechanism.

2. According to claim 1, a method for constructing a complex network of truck mobility that integrates evolution, It is characterized in that Step B also includes the following steps: Step B1: extract the position when the vehicle speed is 0 from the truck trajectory, that is, the truck stop position; Step B2: Determine the parking time threshold; Step B2.1: Arrange the truck stop locations in descending order of parking time, calculate the cumulative parking time of heavy trucks at the ordered stops, normalize the ordinal numbers of the ordered truck stops and the corresponding cumulative parking time, and draw the Lorenz curve L1: Step B2.2: Calculate the intersection F* of the tangent of the Lorenz curve at (1,1) and the horizontal axis. Long-term truck stop sites with standardized ordinal numbers greater than F* are classified as first-class stop sites. Remove the first-class stop sites and repeat step B2.1 to obtain the Lorenz curve L2. The intersection F*' of the tangent of L2 at (1,1) and the horizontal axis is obtained. Long-term truck stop sites with standardized ordinal numbers greater than F*' are classified as second-class stop sites. Remove the second-class stop sites again and repeat step B2.1 to obtain the Lorenz curve L3. The intersection F*' of the tangent of L3 at (1,1) and the horizontal axis is obtained. Long-term truck stop sites with standardized ordinal numbers greater than F*' are classified as third-class stop sites. The parking time corresponding to the third-class stop sites is the parking time threshold. Step B3: Compare the parking time of all parking spots with the parking time threshold, and regard the parking spots whose parking time is less than the time threshold as temporary parking spots and delete them, thereby obtaining the parking purpose of the vehicle.

3. According to claim 1, a method for constructing a fusion-evolved truck mobility complex network, It is characterized in that Step C also includes the following steps: Step C1: Match the truck stop destination obtained in step B with the freight-related POI; Step C1.1: The distance formula between two points in WGS84 coordinate system; In the coordinate system, let the actual distance between points A and B be Dist AB ,but Where, R is the radius of the earth 6371km; Dist AB The unit is km; lat A A is the latitude; lat B is the B latitude; lon A A is longitude; lon B is the longitude of B; Step C1.2: Calculate the distance between the truck stop i and each POI using the formula described in step C1.1, denoted as d ij , The Haversine distance between stop point i and any POI is: The meanings of the parameters in the formula are the same as those in step C1.1, R is the radius of the earth 6371km; d ij The unit is km; lat i is the latitude of node i; lat j is the latitude of node j; lon i is the longitude of node i; lon j is j longitude; Step C1.3: Using d ij Construct the distance matrix between vehicle stops and POIs; The meanings of the parameters in the formula are as follows: m×n Represents the distance matrix between the vehicle stop and the POI, where d 11 Indicates that the vehicle is parked at position 1 and POI 1 The distance between 1n Indicates that the vehicle is parked at position 1 and POI n The distance between m1 Represents the vehicle parked at position m and POI 1 The distance between mn Represents the vehicle parked at position m and POI n The distance between them; Step C1.4: Take the matrix D m×n The minimum value min(d ij ) corresponds to the subscripts i and j, then POI j is the actual destination of the truck at stop i. The driving trajectory of the truck can be recorded as L. Then In the formula, L represents the actual trajectory of the vehicle, that is, the actual trajectory of the vehicle is: in, When the actual stop point of the vehicle is 1, its actual destination stop point is the jth 1 POI points, When the actual stop point of the vehicle is 2, its actual destination stop point is the jth 2 POI points, and so on. When the actual stop point of the vehicle is m, its actual destination stop point is the jth m POI points.

4. According to claim 1, a method for constructing a fusion-evolved truck mobility complex network, It is characterized in that Step D also includes the following steps: Step D1: Aggregate the trajectory network of all trucks; Step D2: Cluster POIs with similar distances according to their geographical location using the DBSCAN method; Step D2.1: Calculate the distance between any two POIs in the network graph obtained in step D1 and obtain the distance distribution matrix DPOI n×n ; Where DPOI n×n is the distance distribution matrix between each POI, dpoi 1-1 It is a POI 1 With POI 1 The distance between 1-n It is a POI 1 With POI n The distance between n-1 It is a POI n With POI 1 The distance between, and so on, dpoi n-n It is a POI n With POI n distance; Step D2.2: DPOI n×n Sort the elements of each row in ascending order, and get the matrix Then the distance vector D composed of the elements in the first column is 1 Represents the distance from the object to itself, all 0, and the elements of the Kth column constitute the K-nearest neighbor distance vector D of all data points K ; Step D2.3: Vector D K Taking the average of the elements in, we can get the vector D K The K-average nearest neighbor distance D K , and use it as the candidate neighborhood radius ε parameter, calculate all K values, and get the ε parameter list D ε , expressed as: D ε ={D K |1≤K≤n} Step D2.4: For parameter list D ε The intersection of the actual buffer radius of the POI is taken to obtain the final neighborhood radius parameter ε, which is recorded as but Where R is the actual buffer radius of the POI; Step D2.5: Generate neighborhood density MinPts parameters; The MinPts parameter is generated by the mathematical expectation method. Where P i is the number of ε-neighborhood objects of the i-th object, and n is the total number of POIs matching the truck trajectory; Step D2.6: The parameters obtained in step D2.4 The parameter MinPts obtained in step D2.5 is used as the DBSCAN parameter. The "Minkowski distance" in the classic DBSCAN algorithm is replaced with the Haversine distance in step C1.

1. Clustering and topology are performed to obtain the urban freight transportation POI topology network.

5. According to claim 1, a method for constructing a fusion-evolved truck mobility complex network, It is characterized in that Step F also includes the following steps: Step F1: Determine the initialization network node; Step F2: randomly connect the edges of the initialized network; by initializing the edges of the network, simulate the existing connection mode of each freight-related POI; Step F3: randomly introduce a freight POI into the network; Step F3.1: Introduce a POI into the network and treat it as a new node; Step F3.2: Based on the properties of this new node, calculate the attraction and break conditions between this node and all nodes in the network; Step F3.3: Determine how many nodes in the network this new node should be connected to based on the attraction obtained in step F3.2; Step F3.4: Determine the connection edges between the existing node and several nodes that should be broken according to the breaking conditions obtained in step F3.2: Step F3.4: Based on the attraction obtained in step F3.2, use the wheel method to determine the connection and break conditions; Step F4: Repeat the process of step F3 until the truck mobility network is constructed.

Citation Information

Patent Citations

  • Generative adversarial network model and vehicle trajectory prediction method using same

    CN111931902A

  • Spatial clustering feature extraction method for incoming logistics cluster

    CN112365217A

  • Two-stage trajectory generation method based on generative adversarial network

    CN115858896A

  • Urban traffic flow network analysis method based on complex network theory

    CN110111575A

  • Urban traffic track data set generation method based on taxi data and urban road network

    CN110298500A