Method for locating transfer point for urban public transport and unmanned aerial vehicle cooperative transportation

Abstract

This invention discloses a method for selecting connection points in collaborative transportation between urban public transport and unmanned aerial vehicles (UAVs), belonging to the technical field of logistics and distribution systems. The method includes: acquiring relevant data; preprocessing urban road network traffic flow data, performing K-means clustering on the preprocessed data to divide the time periods of different daily traffic flow patterns; extracting urban public transport trajectory data features within the target road section, obtaining the road segment operation trajectory features for the time period of that road section through clustering; modeling and analyzing each type of road segment, constructing the time series of each type of road segment using difference equations; calculating the expected value of comprehensive road segments based on the road segment operation trajectory features and characteristic travel times using fuzzy comprehensive evaluation; using a fuzzy compromise method to determine the demand weight of POI data; and using an annealing algorithm for connection point selection. This invention achieves scientific and intelligent connection point selection, effectively improving the efficiency of last-mile collaborative logistics distribution.

Description

Technical Field

[0001] This invention relates to the field of logistics and distribution system technology, and in particular to a method for selecting the location of connection points for collaborative transportation between urban public transport and drones. Background Technology

[0002] Currently, express delivery still relies mainly on manual delivery. Meanwhile, despite smart upgrades, urban public transportation still faces the dual pressures of declining attractiveness, underutilized resources, and high operating costs.

[0003] The low-altitude economy offers a new path to overcome the above-mentioned difficulties. Drones can overcome obstacles in the ground road network and achieve efficient point-to-point transportation. Urban public transport, with its extensive station network, can serve as a resupply, take-off, landing and transfer node for drones, forming a three-dimensional collaborative model of ground public transport combined with low-altitude drones.

[0004] Currently, the collaborative transportation model of public transport and drones has been explored in theory and preliminary practice, and related studies generally acknowledge its potential in improving delivery efficiency and alleviating traffic congestion. However, existing research largely focuses on the overall architecture and feasibility analysis of the collaborative model, lacking systematic and operational technical solutions for the specific site selection of key connection points in the collaborative system. Existing methods often rely on empirical rules or simple geographical proximity principles, failing to comprehensively consider complex factors such as dynamic urban traffic conditions, public transport operation characteristics, and multi-dimensional demand distribution. This results in insufficient scientific rigor in the layout of connection points, low matching with actual delivery needs, and difficulty in fully realizing the effectiveness of the collaborative system.

[0005] Therefore, promoting the deep integration of public transportation and drones to build intelligent joint delivery solutions urgently requires a scientific, precise, and systematic method for selecting connection points to address the methodological deficiencies in existing research and achieve the efficient implementation of collaborative transportation systems. Summary of the Invention

[0006] Purpose of the invention: To address the above problems, the purpose of this invention is to provide a method for selecting connection points for collaborative transportation between urban public transport and drones, thereby solving the problems of express delivery and drone battery life through complementary advantages.

[0007] Technical solution: The method for selecting the location of connection points for urban public transport and drone-assisted transportation of the present invention includes the following steps:

[0008] Step 1: Obtain relevant data, including urban road network traffic flow data, urban public transport operation trajectory data, POI data of the target area, and population heat map data;

[0009] Step 2: Preprocess the traffic flow data of the urban road network, perform K-means clustering on the preprocessed data, and divide the time periods of different daily traffic flow patterns.

[0010] Step 3: Under different daily traffic flow patterns, extract the urban bus trajectory data features within the target road section, and obtain the road segment operation trajectory features for that road section during different time periods through clustering.

[0011] Step 4: Model and analyze each type of road segment, construct the time series of each type of road segment using difference equations, fit it into a first-order autoregressive model, and obtain the characteristic travel time by calculating the equilibrium point.

[0012] Step 5: Based on the characteristics of the road segment's operating trajectory and characteristic travel time, calculate the expected value of the comprehensive road segment using the fuzzy comprehensive evaluation method;

[0013] Step 6: Based on the POI data and population heat map data of the target area, rank the influencing indicators by importance, construct a preference matrix, and use the fuzzy compromise method to find the demand weight of POI data.

[0014] Step 7: Construct a spatially unified demand density field based on the expected value of road segments, the demand weight of POI data, the geometric information and attributes of urban roads, and use the annealing algorithm to select the connection point.

[0015] Preferably, step 4 includes:

[0016] Treating the dwell time within a time-segment traffic flow pattern as a curve of a dynamic system, and assuming a linear correlation between the dwell time in the next time segment and the dwell time in the current time segment, a first-order linear difference equation is constructed, as follows:

[0017] ,

[0018] In the formula, Indicates the first Average dwell time during the period The inertia coefficient represents the strength of the influence of the current state on the next state. External driving coefficient;

[0019] The system reaches equilibrium when it stabilizes and no longer changes over time, i.e., when the formula is satisfied. The characteristic travel time is obtained by solving the problem and is expressed as:

[0020] ,

[0021] In the formula, This represents the stable value to which the dwell time of the current type of road segment will converge under long-term operation.

[0022] Preferably, step 5 includes:

[0023] Step 51: Construct a fuzzy evaluation system for the characteristics of the road segment's operating trajectory, determine the factor set and evaluation level, and use the expert evaluation method to determine the weights of each clustering index.

[0024] Step 52: A weight calculation method based on deviation metric is used to model and calculate the bus trajectory features and road segment features, and then integrate them to form a comprehensive evaluation road segment weight, expressed as:

[0025] ,

[0026] In the formula, For road section The actual travel time. For road section Featured passage time, For road section In clustering The relative importance score within, For clustering Importance score in the entire system clustering, , , As the weighting coefficients, the expert evaluation method is used to satisfy... ;

[0027] Step 53, based on road segment weights Based on the time percentage of that period, the expected value of the comprehensive road segment in a day is calculated using the following formula:

[0028] ,

[0029] In the formula, Representative road section Expected value Representative road section Weights for clustering daily traffic patterns in category J; Representative road section Duration under the J-type daily traffic pattern cluster; This represents the total duration of all types of daily traffic pattern clusters.

[0030] Preferably, step 6 includes:

[0031] The influencing indicators are ranked according to their importance based on the characteristics of the target region, and a first preference matrix is ​​constructed. The construction rules are as follows:

[0032] When for the first The degree of preference for certain indicators Much smaller than the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 0, the first Line number Column elements Take 2;

[0033] When for the first The degree of preference for certain indicators Less than the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 0, the first Line number Column elements Take 1;

[0034] When for the first The degree of preference for certain indicators Approximate to the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 1, the first Line number Column elements Take 1;

[0035] According to the first preference matrix Constructing the second preference matrix , represented as:

[0036] ,

[0037] In the formula, Let be a real number between (0,1), and , ;

[0038] From the second preference matrix Define a directed weighted graph A is a node in the graph, representing the indicator to be evaluated, and the second preference matrix. The elements represent the weights of directed edges. The outgoing edge values ​​in this graph are set as follows:

[0039] ,

[0040] In the formula, For variables c is a variable.

[0041] Therefore, the weight values ​​of the factors influencing POI are calculated and expressed as follows:

[0042] ,

[0043] In the formula, The index number represents the influencing indicator. For variables.

[0044] Preferably, step 7 includes:

[0045] Spatialization of expected values ​​for road segments:

[0046] Each road segment is abstracted as a series of equally spaced virtual points, each virtual point inheriting the expected value of that road segment. Kriging interpolation is then used to interpolate these point values ​​into a continuous grid surface covering the study area, resulting in a continuous demand density field for the road segment. ;

[0047] POI weight spatialization:

[0048] For each POI, a kernel density estimate is created centered on its location. The kernel function radius is adjusted according to the POI type, and the POI weights are used as the height coefficient of the kernel density to obtain the continuous demand density field of the POI. .

[0049] Preferably, step 7 further includes:

[0050] Set filtering conditions to find candidate connection points within the target area. Then, fuse the two types of data within these candidate connection points into a continuous, comprehensive demand field, represented as:

[0051] ,

[0052] In the formula, and This is a weighting coefficient, representing the relative importance of the two demands.

[0053] Continuous integrated demand field Discretize the data into a set of points and assign values ​​to obtain the current position. Value, get a containing A set of candidate connection points, each candidate connection point having location coordinates. and corresponding demand value .

[0054] Preferably, step 7 further includes:

[0055] Select the number of connection points based on the characteristics of the city and the demand. Set decision variables and ,when Time indicates at candidate pick-up point Establish a connection point, otherwise set to 0; when Time indicates demand point From the connection point Service, otherwise 0;

[0056] The objective function is constructed with the goal of minimizing the total cost, and is expressed as:

[0057] ,

[0058] In the formula, Indicates the connection point Fixed construction costs; Indicate demand points to the transfer point The distance; Indicates unit transportation cost; Indicates distribution center Unit storage fee rate; Indicate demand points Annual demand;

[0059] Establish constraints that each demand point must be served by one and only one connection point;

[0060] By solving the objective function, the optimal allocation method is found, which serves as the optimal connection point location.

[0061] Preferably, step 3 includes:

[0062] In various time periods, road sections are set up, and urban bus trajectory data features within these sections are extracted, including dwell time, average speed, speed standard deviation, idling ratio, number of braking events, number of rapid acceleration events, and speed ratio of the preceding road section.

[0063] The characteristics of each trajectory data are standardized, the number of clusters is set, and the characteristics of each road segment in the road network are obtained by K-means clustering, including high-speed, aggressive and stable types.

[0064] Beneficial effects: Compared with the prior art, the significant advantages of this invention are:

[0065] 1. High scientific rigor: This invention comprehensively considers dynamic traffic patterns, public transport operation characteristics, and multi-dimensional urban demand data. It divides daily traffic flow patterns through K-means clustering, obtains characteristic travel times by combining public transport trajectory feature clustering and a first-order autoregressive model, and calculates the expected value of road segments using the fuzzy comprehensive evaluation method, making the site selection decision closer to the real situation of complex urban environments.

[0066] 2. High accuracy: This invention constructs a road segment expected density field and a POI demand density field to form a comprehensive demand field, and uses an annealing algorithm for global optimization. While meeting the goal of minimizing costs, it also takes into account spatial distribution and demand intensity, ensuring that the selected connection points have both scientific validity and practical service value.

[0067] 3. Strong systematicity: This invention constructs a complete method chain from data preprocessing, multi-source feature extraction, multi-stage clustering, fuzzy weight calculation to annealing algorithm site selection, realizing the deep integration and systematic analysis of traffic dynamic data and urban static POI data, and has good operability and promotion value. Attached Figure Description

[0068] Figure 1 This is a flowchart of the present invention;

[0069] Figure 2 A clustering result diagram of daily traffic flow;

[0070] Figure 3 A heatmap showing the temporal distribution of daily traffic flow clusters;

[0071] Figure 4 Heatmap of correlation matrix for bus trajectory features;

[0072] Figure 5 To determine the optimal clustering number map for the elbow rule in bus trajectory features;

[0073] Figure 6 This is a clustering result diagram of bus trajectory features;

[0074] Figure 7 Radar chart for clustering bus trajectory features. Detailed Implementation

[0075] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and not intended to limit the scope of the invention. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the embodiments of the present invention, and not all structures.

[0076] In the following description, specific details such as target system architecture and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0077] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0078] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0079] Furthermore, in the description of this application and the appended claims, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0080] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include the target features, structures, or characteristics described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0081] Combination Figure 1 As shown in this embodiment, the method for selecting the location of the connection point for urban public transport and drone collaborative transportation includes the following steps:

[0082] Step 1: Obtain relevant data, including urban road network traffic flow data, urban public transport operation trajectory data, POI data of the target area, and population heat map data.

[0083] In one example, relevant traffic flow data, urban public transport operation trajectory data, and data on points of interest (POIs) such as companies, schools, business centers, and communities, as well as population heat map data, are obtained within a target area.

[0084] Step 2: Preprocess the traffic flow data of the urban road network, perform K-means clustering on the preprocessed data, and divide the time periods of different daily traffic flow patterns.

[0085] In one example, traffic flow data related to the urban road network is recorded in time series format, with missing traffic flow data filled using linear interpolation. Based on the characteristics of each city, traffic flow data for different time intervals is selected, and K-means clustering is performed on the preprocessed data. The K-means clustering formula is expressed as:

[0086] ,

[0087] ,

[0088] In the formula, Indicates the first The daily traffic distribution pattern for the first The membership degree of each cluster; That is, when the first The daily traffic distribution pattern belongs to the first When clustering ,otherwise ; Indicates the first Daily traffic distribution pattern With the Cluster centroid The square of the Euclidean distance between them.

[0089] In this example, traffic flow data from 6:00 AM to 10:00 PM daily over 7 days was selected for the target area. K-means clustering analysis was performed, grouping traffic flow data into 15-minute intervals. The clustering results are as follows: Figure 2 , Figure 3 As shown in Table 1, the time periods of the daily traffic flow pattern are divided into four groups, including the off-peak period, the morning peak period, the daytime off-peak period, and the evening peak period.

[0090] like Figure 2 As shown, the traffic flow clustering results are presented as a scatter plot and an average traffic flow curve. The horizontal axis is plotted at time points (6:00–22:00, one point every 15 minutes), and the vertical axis is plotted at average traffic flow (number of vehicles / 15 minutes). Different scatter plot colors represent the clusters (four in total, Cluster1–Cluster4) to which each time point belongs. The black curve (Average Traffic) represents the 7-day average traffic flow trend. The curve exhibits a clear bimodal characteristic, with morning peak (approximately 7:30–9:00) and evening peak (approximately 17:00–19:00), consistent with urban commuting patterns. Points of the same color cluster within consecutive time periods, indicating that the clustering algorithm identified time periods with similar traffic flow characteristics.

[0091] like Figure 3As shown, the cluster time distribution heatmap displays the cluster to which each time point belongs. The horizontal axis is the time point, and the vertical axis is the cluster number (Cluster1-Cluster4). The location of the colored block indicates which cluster that time point belongs to. Clusters are represented by red blocks; the wider the block, the longer the cluster has lasted. The colored blocks form continuous bands on the time axis, indicating that each traffic mode is continuous in time, consistent with actual traffic operation characteristics. The boundaries between the colored blocks represent the moments of transition between different traffic modes. For example, the transition point from morning rush hour to daytime off-peak, and from daytime off-peak to evening rush hour. Figure 3 This allows for a direct verification of whether the clustering results are reasonable. If the color blocks are disordered and frequently change, it indicates that the clustering effect is poor; conversely, if the color blocks are continuous and have clear boundaries, it indicates that the clustering is successful.

[0092] Table 1. Results of K-means clustering analysis

[0093]

[0094] Step 3: Under different daily traffic flow patterns, extract the urban bus trajectory data features within the target road section, and obtain the road segment operation trajectory features for that road section during different time periods through clustering.

[0095] Furthermore, step 3 includes:

[0096] In various time periods, road sections are set up, and urban bus trajectory data features within these sections are extracted, including dwell time, average speed, speed standard deviation, idling ratio, number of braking events, number of rapid acceleration events, and speed ratio of the preceding road section.

[0097] The characteristics of each trajectory data are standardized, the number of clusters is set, and the characteristics of each road segment in the road network are obtained by K-means clustering, including high-speed, aggressive and stable types.

[0098] In one example, based on step 2, different traffic flow patterns are divided into time periods. Within each time period, appropriate road intervals are set. In this embodiment, intervals of 50m are used for judgment, and urban public transport trajectory data features are extracted.

[0099] (1) Basic speed characteristics: average speed, maximum speed, speed standard deviation (reflecting the degree of speed fluctuation).

[0100] (2) Speed ​​change characteristics: number of accelerations, number of decelerations, frequency of rapid acceleration / deceleration (the percentage of times the speed change rate exceeds the threshold). The initial number of clusters was set to 3. The number of clusters was optimized using the elbow rule to determine the final number of clusters, thus obtaining the characteristics of each road segment in the road network. The feature correlation heatmap is shown below. Figure 4 As shown in the diagram, the elbow rule is as follows: Figure 5 As shown in the figure, the clustering results are as follows: Figure 6 As shown in the radar charts of each cluster feature, Figure 7 As shown in Table 2, the final classification results in three driving types.

[0101] like Figure 4 The heatmap of the correlation matrix of driving behavior features shown uses color intensity (red→yellow→green→blue) to correspond to correlation coefficients (1→0→-1), which can intuitively show the degree of association between different driving features. This figure provides data basis for subsequent PCA dimensionality reduction and K-means clustering, ensuring that driving behavior classification covers multi-dimensional features while avoiding redundant information interference.

[0102] from Figure 4 The matrix reveals a strong intrinsic correlation among speed-related features (such as maximum speed, speed range, and 90th percentile), indicating information overlap in their representation of driving behavior. In contrast, acceleration-related features show a weaker correlation with speed-related features, exhibiting relatively independent distribution characteristics. This further explains the necessity of introducing PCA dimensionality reduction in this example: by integrating highly correlated speed features into a few principal components, multicollinearity among features can be effectively reduced, information redundancy can be decreased, thereby improving the efficiency and interpretability of subsequent clustering analysis.

[0103] This example uses the elbow rule to determine the optimal number of clusters for K-means clustering.

[0104] Specifically, the elbow rule is a visualization method for determining the optimal number of clusters in K-means clustering. The implementation process includes:

[0105] (1) Define the evaluation index: SSE (sum of squared errors) is used as the evaluation index. SSE represents the sum of squared distances from each data point to the centroid of its cluster. The smaller the SSE, the higher the similarity of data points within the cluster.

[0106] (2) Plot the SSE-K curve: Try different K values ​​in turn (e.g., K=1,2,3,...,n), run the K-means algorithm and calculate the corresponding SSE value, and plot the curve with K as the x-axis and SSE as the y-axis.

[0107] (3) Determine the optimal K value: The curve will decrease as K increases. When K is less than the optimal number of clusters, the SSE decreases faster. When K exceeds the optimal number of clusters, the SSE decreases significantly and the curve shows a clear elbow. The K value corresponding to this elbow is the optimal number of clusters.

[0108] like Figure 5As shown, the sum of squares within clusters (WCSS) decreases with increasing number of clusters, but the rate of decrease has a clear inflection point: when the number of clusters increases from 1 to 3, WCSS decreases rapidly (from approximately 1080 to 570), significantly improving the compactness of the data within clusters; when the number of clusters exceeds 3, the rate of decrease in WCSS slows down significantly, and further increasing the number of clusters has limited effect on optimizing the compactness of clusters. Therefore, the optimal number of clusters for this driving behavior classification is 3, which ensures a balance between the compactness of clusters and the discriminative power of the categories.

[0109] In this example, the elbow rule is used to optimize the number of clusters, avoiding the subjectivity of manually setting the K value, providing a reasonable number of categories for driving behavior classification, and ensuring the scientific nature of the clustering results.

[0110] like Figure 6 , Figure 7 The images shown are visualizations of the driving behavior clustering results, including a principal component scatter plot and a radar chart. Figure 6 It can be seen that the three clusters show a clear separation trend in the principal component space. Cluster 2 is concentrated in the higher region of the first principal component, cluster 3 is concentrated in the lower region of the first principal component, and cluster 1 is distributed in the middle region, indicating that the characteristics of the three types of driving behaviors are significantly different.

[0111] from Figure 7 As can be seen from the data, Cluster 1 (yellow) shows outstanding performance in speed-related features such as avg_speed (average speed), median_speed (median speed), and speed_90th (90th percentile speed), indicating that the overall driving speed is relatively high. Cluster 2 (orange) shows more prominence in acceleration features such as sharp_accel_count (number of rapid accelerations) and sharp_decel_count (number of rapid decelerations), reflecting a more aggressive driving style. Cluster 3 (blue) shows relatively mild performance in all features, indicating a more stable driving type.

[0112] Combination Figure 6 and Figure 7 As can be seen, through K-means clustering, driving behavior is divided into three categories: high-speed (cluster 1), aggressive (cluster 2), and stable (cluster 3). The characteristics of the three types of driving behavior are clearly distinguishable, providing an intuitive basis for subsequent driving behavior assessment. The results are shown in Table 2.

[0113] Table 2 Driving Type Results Table

[0114]

[0115] Step 4: Model and analyze each type of road segment, construct the time series of each type of road segment using difference equations, fit it into a first-order autoregressive model, and obtain the characteristic travel time by calculating the equilibrium point.

[0116] Furthermore, step 4 includes:

[0117] Treating the dwell time within a time-segment traffic flow pattern as a curve of a dynamic system, and assuming a linear correlation between the dwell time in the next time segment and the dwell time in the current time segment, a first-order linear difference equation is constructed, as follows:

[0118] ,

[0119] In the formula, Indicates the first Average dwell time during the period The inertia coefficient represents the strength of the influence of the current state on the next state. External driving coefficient;

[0120] The system reaches equilibrium when it stabilizes and no longer changes over time, i.e., when the formula is satisfied. The characteristic travel time is obtained by solving the problem and is expressed as:

[0121] ,

[0122] In the formula, This represents the stable value to which the dwell time of the current type of road segment will converge under long-term operation.

[0123] This example selects a section of road during a certain time period, as shown in Table 3.

[0124] Table 3 Example of road segment data

[0125]

[0126] The inertia coefficient is found by using the least squares method to minimize the prediction error. The external driving coefficient is The characteristic passage time is then calculated as follows:

[0127] ,

[0128] Step 5: Based on the characteristics of the road segment's operating trajectory and characteristic travel time, calculate the expected value of the comprehensive road segment using the fuzzy comprehensive evaluation method.

[0129] Furthermore, step 5 includes:

[0130] Step 51: Construct a fuzzy evaluation system for the characteristics of the road segment's operating trajectory, determine the factor set and evaluation level, and use the expert evaluation method to determine the weights of each clustering index.

[0131] In one example, a fuzzy evaluation system is constructed for the three clusters established based on the characteristics of the road segment operation trajectory in step 3. The factor set and evaluation level are determined, and the weights of the respective cluster indicators are determined using the expert evaluation method. For example, in the aggressive type (cluster 2), the number of rapid accelerations, the number of rapid decelerations, and the acceleration fluctuation index are fuzzily evaluated based on the relevant data of the bus operation trajectory to determine the general weights of this cluster.

[0132] Step 52: A weight calculation method based on deviation metric is used to model and calculate the bus trajectory features and road segment features, and then integrate them to form a comprehensive evaluation road segment weight, expressed as:

[0133] ,

[0134] In the formula, For road section The actual travel time. For road section Featured passage time, For road section In clustering The relative importance score within, For clustering Importance score in the entire system clustering, , , As the weighting coefficients, the expert evaluation method is used to satisfy... ;

[0135] Step 53, based on road segment weights Based on the time percentage of that period, the expected value of the comprehensive road segment in a day is calculated using the following formula:

[0136] ,

[0137] In the formula, Representative road section Expected value Representative road section Weights for clustering daily traffic patterns in category J; Representative road section Duration under the J-type daily traffic pattern cluster; This represents the total duration of all types of daily traffic pattern clusters.

[0138] Step 6: Based on the POI data and population heat map data of the target area, rank the influencing indicators by importance, construct a preference matrix, and use the fuzzy compromise method to find the demand weight of POI data.

[0139] Furthermore, step 6 includes:

[0140] The influencing indicators are ranked according to their importance based on the characteristics of the target region, and a first preference matrix is ​​constructed. The construction rules are as follows:

[0141] When for the first The degree of preference for certain indicators Much smaller than the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 0, the first Line number Column elements Take 2;

[0142] When for the first The degree of preference for certain indicators Less than the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 0, the first Line number Column elements Take 1;

[0143] When for the first The degree of preference for certain indicators Approximate to the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 1, the first Line number Column elements Take 1;

[0144] The preference matrix Ra is then expressed as:

[0145] ,

[0146] According to the first preference matrix Constructing the second preference matrix , represented as:

[0147] ,

[0148] In the formula, Let be a real number between (0,1), and , ;

[0149] From the second preference matrix Define a directed weighted graph A is a node in the graph, representing the indicator to be evaluated, and the second preference matrix. The elements represent the weights of directed edges. The outgoing edge values ​​in this graph are set as follows:

[0150] ,

[0151] In the formula, For variables ; As variables, ;

[0152] Therefore, the weight values ​​of the factors influencing POI are calculated and expressed as follows:

[0153] ,

[0154] In the formula, The index number represents the influencing indicator. For variables.

[0155] In one example, the level of preference is set to: community. The school is The company is Business Center The characteristics of a certain area are: communities and schools are of similar importance; schools are less important than companies; schools are less important than companies; and companies are far less important than commercial centers, i.e., communities ( ) School( ),School( ) company( ),company( ) Business Center ), thereby constructing a preference matrix.

[0156] Preset parameters: .

[0157] according to The conversion rules yield the values ​​of the preference matrix R shown in Table 4.

[0158] Table 4. Values ​​of the Preference Matrix R

[0159]

[0160] Calculate the boundary values ​​of each region They are respectively:

[0161] ,

[0162] ,

[0163] ,

[0164] ,

[0165] The sum of all outgoing boundary values ​​is:

[0166] ,

[0167] Calculate the weighting coefficients for each region They are respectively:

[0168] ,

[0169] ,

[0170] ,

[0171] .

[0172] Step 7: Construct a spatially unified demand density field based on the expected value of road segments, the demand weight of POI data, the geometric information and attributes of urban roads, and use the annealing algorithm to select the connection point.

[0173] Further, step 7 includes:

[0174] (1) Spatialization of expected values ​​for road segments:

[0175] Each road segment is abstracted as a series of equally spaced virtual points, each virtual point inheriting the expected value of that road segment. Kriging interpolation is then used to interpolate these point values ​​into a continuous grid surface covering the study area, resulting in a continuous demand density field for the road segment. .

[0176] Specifically, the road segment is first discretized into virtual points. For each road segment, virtual points are generated at equal intervals along the road centerline, and each point inherits the expected value of its respective road segment. The sampling interval is selected based on the road curvature and the gradient of the expected value change; the sampling interval should be less than half of the minimum curvature radius of the road. In special cases, such as curved road segments, sampling should be performed at equal intervals along the road. When there are multiple points from different road segments near an intersection, all points should be retained. For road segment endpoints, sampling points should be ensured at both the start and end points. In this example, a point is taken every 10 meters. The empirical variability function is calculated: the distance and difference between all pairs of points are calculated. For each pair of known points... and Calculate their spatial distance The empirical variability function formula is as follows:

[0177] ,

[0178] In the formula: This represents discrete points calculated from real data. For distance equal to Number of point pairs For in position Observations at that location Indicates distance for The value at another location.

[0179] A variogram is plotted with distance h on the horizontal axis and semivariance γ(h) on the vertical axis. The theoretical variogram model is fitted using a spherical model, where the expression for the spherical model is:

[0180] ,

[0181] ,

[0182] In the formula, Indicates distance as Correlation coefficient at time; (Nuclear value) is the variation value as h→0, reflecting measurement error or microscale variation; (Partial sill values) represent the maximum spatial variation; (Range) represents the maximum distance of spatial autocorrelation; beyond this distance, correlation ceases. (Staple value) This represents the total degree of variation within the region.

[0183] A grid is created, dividing the entire study area into small grids, with a grid size of [size missing]. (Same sampling interval as virtual points). Each cell has a center point, and a value is estimated for that center point. For the current cell center point to be estimated... Use known points whose distance is less than the search radius, and set the search radius to 2-3 times the range obtained from the variogram analysis. Set the weights for the kriging. Two conditions must be met:

[0184] Optimality: Predicted value The error is minimized, that is, the variance is minimized;

[0185] Unbiasedness: The expected value of the prediction equals the expected value of the actual value, i.e. The weights must sum to 1, i.e. .

[0186] Let the prediction point be... There are around Known points , The weight to be determined; It is a Lagrange multiplier, which is used to satisfy unbiasedness. Introduced parameters. Constructing the Kriging equations (minimum variance and unbiasedness): To minimize the variance of the prediction error, the final derived linear equations are shown below:

[0187] ,

[0188] In the formula, For known points and The semivariogram values ​​between; For point The difference between itself and the original is 0; For known points and prediction points The semivariogram values ​​between; The weight to be determined; These are auxiliary parameters (ensuring the sum of weights is 1).

[0189] Solve the system of linear equations to obtain the weights. and The estimated value is calculated using the following formula:

[0190] ,

[0191] The formula for calculating the estimated variance is:

[0192] ,

[0193] Traverse all grid points and repeat the above process for each grid point to obtain the complete raster surface. Interpolation at the edges of the study area may be unreliable and requires boundary processing. In the buffer, interpolation is performed at a certain distance outside the study area, and then clipped; the estimated values ​​at the edges are given a lower confidence level; if the results are noisy, a 3×3 window can be used for mean filtering, and finally the values ​​are mapped to the interval [0, 1].

[0194] (2) Spatialization of POI weights:

[0195] For each POI, a kernel density estimate is created centered on its location. The kernel function radius is adjusted according to the POI type, and the POI weights are used as the height coefficient of the kernel density to obtain the continuous demand density field of the POI. .

[0196] Specifically, each point has different levels of importance, and weights are introduced. Then the weighted kernel density formula is:

[0197] ,

[0198] In the formula, In spatial location POI demand density value at the location; This represents the total number of POIs within the study area. For the first The weight of each POI represents its importance or attractiveness. For the first The bandwidth (radius of influence) of a POI is usually determined based on the POI type. For point To the The Euclidean distance between the POIs.

[0199] Furthermore, step 7 also includes:

[0200] Set filtering conditions to find candidate connection points within the target area. Then, fuse the two types of data within these candidate connection points into a continuous, comprehensive demand field, represented as:

[0201] ,

[0202] In the formula, and This is a weighting coefficient, representing the relative importance of the two demands.

[0203] Continuous integrated demand field Discretize the data into a set of points and assign values ​​to obtain the current position. Value, get a containing A set of candidate connection points, each candidate connection point having location coordinates. and corresponding demand value .

[0204] Furthermore, step 7 also includes:

[0205] Select the number of connection points based on the characteristics of the city and the demand. Set decision variables and ,when Time indicates at candidate pick-up point Establish a connection point, otherwise set to 0; when Time indicates demand point From the connection point Service, otherwise 0;

[0206] The objective function is constructed with the goal of minimizing the total cost, and is expressed as:

[0207] ,

[0208] In the formula, Indicates the connection point Fixed construction costs; Indicate demand points to the transfer point The distance; Indicates unit transportation cost; Indicates distribution center Unit storage fee rate; Indicate demand points Annual demand;

[0209] Establish constraints that each demand point must be served by one and only one connection point;

[0210] By solving the objective function, the optimal allocation method is found, which serves as the optimal connection point location.

Claims

1. A method for selecting the location of a transfer point for collaborative urban public transport and unmanned aerial vehicle (UAV) transportation, characterized in that, Includes the following steps: Step 1: Obtain relevant data, including urban road network traffic flow data, urban public transport operation trajectory data, POI data of the target area, and population heat map data; Step 2: Preprocess the traffic flow data of the urban road network, perform K-means clustering on the preprocessed data, and divide the time periods of different daily traffic flow patterns. Step 3: Under different daily traffic flow patterns, extract the urban bus trajectory data features within the target road section, and obtain the road segment operation trajectory features for that road section during different time periods through clustering. Step 4: Model and analyze each type of road segment, construct the time series of each type of road segment using difference equations, fit it into a first-order autoregressive model, and obtain the characteristic travel time by calculating the equilibrium point. Step 5: Based on the characteristics of the road segment's operating trajectory and characteristic travel time, calculate the expected value of the comprehensive road segment using the fuzzy comprehensive evaluation method; Step 6: Based on the POI data and population heat map data of the target area, rank the influencing indicators by importance, construct a preference matrix, and use the fuzzy compromise method to find the demand weight of POI data. Step 7: Construct a spatially unified demand density field based on the expected value of road segments, the demand weight of POI data, the geometric information and attributes of urban roads, and use the annealing algorithm to select the connection point.

2. The method for selecting the location of the connection point for urban public transport and drone-assisted transportation according to claim 1, characterized in that, Step 4 includes: Treating the dwell time within a time-segment traffic flow pattern as a curve of a dynamic system, and assuming a linear correlation between the dwell time in the next time segment and the dwell time in the current time segment, a first-order linear difference equation is constructed, as follows: ， In the formula, Indicates the first Average dwell time during the period The inertia coefficient represents the strength of the influence of the current state on the next state. External driving coefficient; The system reaches equilibrium when it stabilizes and no longer changes over time, i.e., when the formula is satisfied. The characteristic travel time is obtained by solving the problem and is expressed as: ， In the formula, This represents the stable value to which the dwell time of the current type of road segment will converge under long-term operation.

3. The method for selecting the location of the connection point for urban public transport and drone-assisted transportation according to claim 2, characterized in that, Step 5 includes: Step 51: Construct a fuzzy evaluation system for the characteristics of the road segment's operating trajectory, determine the factor set and evaluation level, and use the expert evaluation method to determine the weights of each clustering index. Step 52: A weight calculation method based on deviation metric is used to model and calculate the bus trajectory features and road segment features, and then integrate them to form a comprehensive evaluation road segment weight, expressed as: ， In the formula, For road section The actual travel time. For road section Featured passage time, For road section In clustering The relative importance score within, For clustering Importance score in the entire system clustering, , , As the weighting coefficients, the expert evaluation method is used to satisfy... ; Step 53, based on road segment weights Based on the time percentage of that period, the expected value of the comprehensive road segment in a day is calculated using the following formula: ， In the formula, Representative road section Expected value Representative road section Weights for clustering daily traffic patterns in category J; Representative road section Duration under the J-type daily traffic pattern cluster; This represents the total duration of all types of daily traffic pattern clusters.

4. The method for selecting the location of the connection point for urban public transport and drone-assisted transportation according to claim 3, characterized in that, Step 6 includes: The influencing indicators are ranked according to their importance based on the characteristics of the target region, and a first preference matrix is ​​constructed. The construction rules are as follows: When for the first The degree of preference for certain indicators Much smaller than the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 0, the first Line number Column elements Take 2; When for the first The degree of preference for certain indicators Less than the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 0, the first Line number Column elements Take 1; When for the first The degree of preference for certain indicators Approximate to the first The degree of preference for certain indicators When, the first in the preference matrix Line number Column elements Take 1, the first Line number Column elements Take 1; According to the first preference matrix Constructing the second preference matrix , represented as: ， In the formula, Let be a real number between (0,1), and , ; From the second preference matrix Define a directed weighted graph A is a node in the graph, representing the indicator to be evaluated, and the second preference matrix. The elements represent the weights of directed edges. The outgoing edge values ​​in this graph are set as follows: ， In the formula, For variables c is a variable. Therefore, the weight values ​​of the factors influencing POI are calculated and expressed as follows: ， In the formula, The index number represents the influencing indicator. For variables.

5. The method for selecting the location of the connection point for coordinated urban public transport and unmanned aerial vehicle (UAV) transportation according to claim 1, characterized in that, Step 7 includes: Spatialization of expected values ​​for road segments: Each road segment is abstracted as a series of equally spaced virtual points, each virtual point inheriting the expected value of that road segment. Kriging interpolation is then used to interpolate these point values ​​into a continuous grid surface covering the study area, resulting in a continuous demand density field for the road segment. ; POI weight spatialization: For each POI, a kernel density estimate is created centered on its location. The kernel function radius is adjusted according to the POI type, and the POI weights are used as the height coefficient of the kernel density to obtain the continuous demand density field of the POI. .

6. The method for selecting the location of the connection point for coordinated urban public transport and unmanned aerial vehicle (UAV) transportation according to claim 5, characterized in that, Step 7 also includes: Set filtering conditions to find candidate connection points within the target area. Then, fuse the two types of data within these candidate connection points into a continuous, comprehensive demand field, represented as: ， In the formula, and This is a weighting coefficient, representing the relative importance of the two demands. Continuous integrated demand field Discretize the data into a set of points and assign values ​​to obtain the current position. Value, get a containing A set of candidate connection points, each candidate connection point having location coordinates. and corresponding demand value .

7. The method for selecting the location of the connection point for coordinated urban public transport and unmanned aerial vehicle (UAV) transportation according to claim 5, characterized in that, Step 7 also includes: Select the number of connection points based on the characteristics of the city and the demand. Set decision variables and ,when Time indicates at candidate pick-up point Establish a connection point, otherwise set to 0; when Time indicates demand point From the connection point Service, otherwise 0; The objective function is constructed with the goal of minimizing the total cost, and is expressed as: ， In the formula, Indicates the connection point Fixed construction costs; Indicate demand points to the transfer point The distance; Indicates unit transportation cost; Indicates distribution center Unit storage fee rate; Indicate demand points Annual demand; Establish constraints that each demand point must be served by one and only one connection point; By solving the objective function, the optimal allocation method is found, which serves as the optimal connection point location.

8. The method for selecting the location of the connection point for urban public transport and drone-assisted transportation according to claim 1, characterized in that, Step 3 includes: In various time periods, road sections are set up, and urban bus trajectory data features within these sections are extracted, including dwell time, average speed, speed standard deviation, idling ratio, number of braking events, number of rapid acceleration events, and speed ratio of the preceding road section. The characteristics of each trajectory data are standardized, the number of clusters is set, and the characteristics of each road segment in the road network are obtained by K-means clustering, including high-speed, aggressive and stable types.

Images