A method for layout of unmanned aerial vehicle transfer station based on topology optimization

Abstract

The application discloses a kind of based on topology optimization's unmanned plane transfer station layout method.Method includes that target area is gridded and available range is revised;Historical order start, end point is mapped as grid center pair weighted line segment, and main transport corridor is extracted by clustering;Candidate transfer station is laid along corridor under the constraint of available range and is connected to form skeleton network;With covering radius and available range as constraint, according to demand intensity, skeleton network is iteratively repaired, all demand grid centers are covered by translation or station insertion, and final layout is obtained.The method of the application considers skeleton to fit actual flow direction and number control of setting, and ensures the reachability of unmanned plane transfer station network.

Description

Technical Field

[0001] This invention belongs to the field of UAV logistics and transportation network optimization technology, and specifically relates to a UAV transfer station layout method based on topology optimization. Background Technology

[0002] With the rapid development of e-commerce, instant delivery, and the low-altitude economy, the exploration of drones' participation in urban and regional logistics is constantly advancing. Drones, with their strong navigation capabilities, short response times, and lack of ground congestion constraints, have demonstrated application potential in scenarios such as resupply in remote areas, delivery in mountainous and island regions, and the delivery of medical emergency supplies. To alleviate inherent bottlenecks such as limited range and payload, research and practice are gradually introducing drone transit facilities to extend service radius and improve mission completion rates.

[0003] Against this backdrop, extensive research has been conducted on the design and optimization of drone logistics networks. On the one hand, some studies, starting from traditional logistics site selection-allocation and hub-and-radius networks, treat drones as a new type of transport capacity, optimizing the location of central stations, radial structures, and capacity configuration given candidate nodes to balance construction costs and service levels. On the other hand, addressing range constraints and refueling needs, existing research has introduced take-off and landing points and charging stations, jointly modeling and solving site selection and flight paths, achieving high coverage and timeliness in specific scenarios. At the engineering practice and patent layout levels, various forms of drone take-off and landing platforms, logistics cabinets, and transfer boxes have emerged, integrating take-off and landing, charging (battery swapping), temporary storage, and unmanned handover functions into a single facility to support last-mile or localized drone logistics services.

[0004] However, existing technologies still have certain limitations. One type of approach focuses on the structural design and control process improvement of individual facilities or local systems, without considering the logistics network planning for multi-site drone relay transportation. Another type of network design method only addresses single-link transportation under certain special circumstances, and conducts site selection and route planning based on this, without considering the regional-level transit station network layout for drone relay transportation. In particular, for transit station systems that simultaneously undertake comprehensive functions such as energy replenishment and cargo relay, existing research has not yet provided a method for automatically extracting the main transportation corridor from historical orders and generating a globally reachable transit station layout scheme based on this, while taking into account the effective range of drones and the spatial distribution of demand. Summary of the Invention

[0005] To address the problems existing in the background technology, this invention provides a method for the layout of UAV transfer stations based on topology optimization, which solves the technical problems in the prior art where the layout of UAV transfer stations does not consider regional network connectivity and does not fully explore historical needs.

[0006] This invention comprehensively considers the distribution characteristics of historical order data and the coverage of candidate transit station networks for overall demand, and realizes the construction of a drone transit station network framework that meets the main needs, and achieves full coverage of the demand area by the drone transit station network.

[0007] The technical solution adopted by the UAV transfer station layout method of the present invention is as follows:

[0008] S1. Grid division and range correction: Divide the target area into multiple grids and determine the available range of the UAV under the grid division based on the UAV's nominal maximum range.

[0009] S2. Extraction of main transportation corridors: Map the start and end points of each historical order to the center of the grid, and obtain weighted line segments composed of grid center pairs. Cluster all weighted line segments and extract several main transportation corridors based on the clustering results.

[0010] S3. Backbone Network Construction: On each main transport corridor, multiple candidate transfer stations are deployed according to the available flight range, and all candidate transfer stations on the main transport corridors are connected to construct a candidate transfer station backbone network covering all main transport corridors.

[0011] S4. Demand Point Access: Based on historical orders, determine all demand grid centers and their demand intensity. With coverage radius and available range as constraints, iteratively repair the candidate transfer station skeleton network, prioritizing the uncovered demand grid centers with higher demand intensity, until all demand grid centers are covered, and finally obtain the UAV transfer station network.

[0012] Step S1 specifically involves:

[0013] S11. Based on the UAV's nominal maximum range and the preset grid ratio coefficient, determine the grid side length and divide the target area into multiple grids, with the geometric center of each grid defined as the grid center.

[0014] S12. Based on the nominal maximum range, the preset range safety factor, and the grid side length, the available range used to constrain the flight distance between transfer stations is calculated.

[0015] The available range is obtained by processing it using the following formula:

[0016] ;

[0017] in, Available range; This is the nominal maximum range; This is the grid scale factor; The grid side length; This is the safety factor for the flight range.

[0018] Step S2 specifically involves:

[0019] S21. Map the start and end points of each historical order to the center of the grid, construct line segments between grid center pairs, and assign weights to the line segments according to the historical order volume to obtain weighted line segments.

[0020] S22. For each weighted line segment, extract the direction features and position features, and concatenate them to form the feature vector of the corresponding weighted line segment.

[0021] S23. Based on the geometric distance between feature vectors, perform density clustering on all weighted line segments to obtain several clusters, and select multiple principal clusters that meet the preset conditions.

[0022] S24. Perform principal component analysis on each main cluster, and determine the start and end points of the main transport corridor corresponding to the current main cluster based on the direction of the first principal axis of the analysis results, thereby obtaining the main transport corridor.

[0023] In step S22:

[0024] The eigenvector of each weighted line segment is obtained by processing it using the following formula:

[0025]

[0026] , , ,

[0027] in, For indexing; For the first eigenvectors of weighted line segments; and All are the first The directional characteristics of weighted line segments; and All are the first Positional characteristics of weighted line segments; For the first The direction angle of the weighted line segments; and For the first The x and y coordinates of the midpoint of the weighted line segment; Position weights; This represents the side length of the grid region.

[0028] Step S23 specifically involves:

[0029] S231. Preset the neighborhood radius and minimum number of core segments, and use the geometric distance between feature vectors as the weighted distance between segments.

[0030] S232. For each weighted line segment, determine whether each weighted line segment is a core weighted line segment based on the distance between weighted line segments: if the number of weighted line segments contained in the neighborhood radius of the current weighted line segment is not less than the minimum number of core line segments, then the current weighted line segment is a core weighted line segment; otherwise, the current weighted line segment is a non-core weighted line segment.

[0031] S233. Based on all core weighted line segments and density-connected weighted line segments, the weighted line segments are divided into several clusters using a density clustering algorithm.

[0032] Step S3 specifically involves:

[0033] S31. For each main transport corridor, from the starting point to the end point, candidate transfer stations are determined and deployed based on the available flight distance, such that the distance between adjacent candidate transfer stations on the same main transport corridor does not exceed the available flight distance.

[0034] S32. Determine the connection relationship between different main transport corridors, and select the candidate transfer station that is closest to each of the two main transport corridors that need to be connected as the connection point.

[0035] S33. With available flight range as a constraint, new candidate transfer stations are set up between connection points to connect all main transport corridors, thereby forming a connected candidate transfer station backbone network.

[0036] Step S4 specifically involves:

[0037] S41. Determine all demand grid centers based on historical orders, and determine the demand intensity of each demand grid center.

[0038] S42. Verify whether the distance between each demand grid center and the nearest candidate transfer station in the candidate transfer station skeleton network meets the preset coverage radius. Mark demand grid centers that do not meet the coverage radius as uncovered demand grid centers, and mark demand grid centers that meet the coverage radius as covered demand grid centers.

[0039] S43. Based on the magnitude of the demand intensity corresponding to the uncovered demand grid center, each uncovered demand grid center is treated as a processing object for repair processing in turn. When all uncovered demand grid centers have been repaired, the candidate transfer station skeleton network obtained is used as the UAV transfer station network.

[0040] The specific process for repairing each uncovered demand grid center as the processing target is as follows:

[0041] F1. For each uncovered demand grid center, prioritize shifting the nearest candidate transfer station towards the current uncovered demand grid center along the line connecting the current uncovered demand grid center and the nearest candidate transfer station, using the neighbor set constraint and the original coverage constraint as common constraints, so that the current uncovered demand grid center is within the coverage radius of the shifted candidate transfer station.

[0042] F2. If the translation operation in step F1 cannot make the center of the currently uncovered demand grid within the coverage radius of the translated candidate transfer station, then, with the available range and coverage radius as constraints, insert several new candidate transfer stations at equal intervals on the line connecting the center of the uncovered demand grid and the nearest candidate transfer station, so that the distance between any two adjacent candidate transfer stations on the line does not exceed the available range and the distance from the last new candidate transfer station on the line to the center of the currently uncovered demand grid is within the coverage radius.

[0043] The neighbor set constraint is that the distance between the candidate transfer station after translation and all adjacent candidate transfer stations in the candidate transfer station skeleton network does not exceed the available flight range; the original coverage constraint is that the centers of all covered demand grids within the coverage radius of the candidate transfer station before translation must still be within the coverage radius of the candidate transfer station after translation.

[0044] The number of new candidate transfer stations inserted in step F2 and the position of each new candidate transfer station are set according to the following formula:

[0045]

[0046]

[0047] in, The number of new candidate transit stations to insert; Indicates rounding up; For currently uncovered demand grid centers Its most recent candidate transit station The distance between them; The coverage radius; Available range; Using indexes, retrieve values ​​from 1 to... ; Indicates the first The location of a new candidate transit point; and These represent the currently uncovered demand grid centers. and recent candidate transit stations The location.

[0048] The beneficial effects of this invention are:

[0049] 1. This invention extracts the corresponding feature vectors from the weighted line segments between the grid center pairs formed by historical orders, obtains the main transportation corridor through density clustering, and deploys a candidate transfer station skeleton network on the main transportation corridor, so that the layout of UAV transfer stations is more in line with the actual cargo flow and reduces low-utilization transfer stations and invalid flight segments.

[0050] 2. This invention targets massive historical orders and extracts the main transportation corridors through grid division and classification, feature extraction, and line segment clustering, which greatly reduces the computational scale and time complexity of network construction and is suitable for large-scale order scenarios at the city level.

[0051] 3. In the modeling process, this invention distinguishes between the nominal maximum range of UAVs and the available range of operations. It uses the available range to uniformly constrain the side lengths and site coverage radii between candidate transfer stations, and limits the station movement distance during translation repair. It introduces flight redundancy from the planning stage, thereby improving the feasibility and flight safety of the planned network in actual operation.

[0052] 4. This invention only performs reachability checks on demand grid centers that appear in historical orders. For uncovered demand grid centers, it prioritizes repair by shifting the nearest candidate transfer station. Only when the range and coverage constraints cannot be met will a new station be inserted. Under the premise of ensuring that all demand grids are connected to the network, the number of newly added candidate transfer stations is controlled, thereby reducing the subsequent facility construction and operation and maintenance costs. Attached Figure Description

[0053] Figure 1 This is a flowchart of the method of the present invention.

[0054] Figure 2 This is a distribution map of historical orders generated in Example 1.

[0055] Figure 3 This is a schematic diagram showing the square grid division of the study area and the location of the grid center.

[0056] Figure 4 To generate an aggregated result graph centered on the order grid.

[0057] Figure 5 This is a schematic diagram of mapping historical orders to weighted line segments at the grid center.

[0058] Figure 6 This is the final network diagram of the drone transfer station obtained in Example 1.

[0059] Figure 7 The diagram shows the layout of transfer stations obtained using the random location method in Comparative Example 1.

[0060] Figure 8 The diagram shows the layout of the transfer station obtained using the midpoint layout method as a comparison example 2. Detailed Implementation

[0061] The present invention will now be described in more detail with reference to the accompanying drawings and embodiments. However, the present invention is not limited thereto. For those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention. Contents not described in detail in this specification are prior art known to those skilled in the art.

[0062] Example 1:

[0063] The study area in this embodiment is a square area of ​​40km × 40km, and the nominal maximum range of the UAV is 8km; there are 600 historical orders, which tend to be distributed along the diagonal direction, with the probability of generation increasing as they are closer together. The orders are generated randomly, and the generation results are as follows: Figure 2 As shown.

[0064] like Figure 1 As shown, the drone transfer station layout method of this embodiment is implemented according to the following steps:

[0065] S11. Assume the nominal maximum range of the UAV is... To discretize the urban space (target area), a grid scaling factor is selected. Define the grid side length as:

[0066]

[0067] In this embodiment, take Therefore, the grid side length It is 1km.

[0068] like Figure 3 As shown, in this embodiment, the research area (target area) is first embedded in an outer rectangle with a side length of... A regular square grid is used to cover the entire study area, and the geometric center of each grid is taken as the "grid center". The start and end positions of subsequent historical orders are mapped to the corresponding grid centers, thus transforming the demand in continuous space into discrete nodes at the grid level. The processing result is as follows: Figure 4 As shown, the darker the color of the center point of the grid, the more start or end points of the aggregated orders there are.

[0069] S12, the nominal maximum range of the drone is Introducing a range safety factor (If you take) The nominal maximum range is reduced to the safe range. :

[0070]

[0071] On the other hand, the diagonal length of a single square grid is In the worst-case scenario, the round trip from the transit station to the grid center, then to the specific service location, and back again, for a single grid, can be considered to be at most [missing information - likely a number]. Therefore, the available range was adjusted. for:

[0072]

[0073] This indicates the longest flight length that can be safely used on the network after considering safety redundancy and intra-grid round trip.

[0074] S21. Map the start and end points of each historical order to the center of the corresponding grid, construct line segments between grid center pairs, and assign weights to the line segments based on the historical order volume to obtain weighted line segments. Further, S21 specifically involves:

[0075] Weighted line segment mapping. Let the historical order set be... (in, (for indexing), each historical order The start and end points are projected onto the center of their respective grids. and , forming a line segment :

[0076]

[0077] Each line segment weight All are assigned a value of 1. When the origin and destination of multiple historical orders are mapped to the same pair of grid centers... At that time, the weights of these historical orders are linearly summed to obtain a weighted line segment. (in, (for index), weighted line segment The weight is equal to the number of orders.

[0078] Subsequently, all the obtained weighted line segments constitute a weighted line segment set. This lays the foundation for subsequent geometric similarity calculations and line segment clustering. Weighted line segment set. Each weighted line segment in It has a start and end point (grid center), geometric direction, length and weight, and subsequent clustering is carried out based on this batch of line segments.

[0079] For example, Figure 5 As shown, when only the first line segment and the line segment The start and end points are mapped to the same pair of grid centers. hour:

[0080] No. line segment weight For 1, the first line segment weight The value is 1, thus the two line segments and The obtained weighted line segment The weight is The value is 2.

[0081] S22. For each weighted line segment, extract its direction and position features, and concatenate them to form the feature vector of the corresponding weighted line segment. Further, S22 specifically involves:

[0082] For each weighted line segment Calculate the direction angle and with direction angle Cosine and sine are used as directional features:

[0083]

[0084] Weighted line segments midpoint , Normalize and multiply by position weight (like ), to obtain location features:

[0085] ,

[0086] in, The side length of the grid region under study.

[0087] S23. Based on the geometric distance between feature vectors, perform density clustering on all weighted line segments to obtain several clusters, and select multiple principal clusters that meet preset conditions. Further, S23 specifically involves:

[0088] S231, Preset neighborhood radius and minimum number of core segments The geometric distance between eigenvectors is used as the weighted distance between line segments.

[0089] The geometric distance between any two feature vectors is set according to the following formula:

[0090]

[0091] in, and All are indexes; Let Euclidean distance be the eigenvectors corresponding to the i-th and j-th weighted line segments; For indexing, It is the m-th vector in the eigenvectors corresponding to the i-th weighted line segment; It is the m-th vector in the eigenvectors corresponding to the j-th weighted line segment.

[0092] S232. For each weighted line segment, determine whether each weighted line segment is a core weighted line segment based on the distance between weighted line segments:

[0093] If the neighborhood radius of the current weighted line segment The number of weighted line segments contained within is not less than the number of minimum core line segments. If the value is true, then the current weighted segment is a core weighted segment; otherwise, the current weighted segment is a non-core weighted segment.

[0094] S233. Based on all core weighted line segments and density-connected weighted line segments, the weighted line segments are divided into several clusters using a density clustering algorithm. (in, For indexing, For the first (Number of clusters); In specific implementation, the DBSCAN density clustering algorithm is used.

[0095] In practice, the neighborhood radius of each core weighted line segment All weighted line segments within are considered to be connected to the corresponding core weighted line segment density.

[0096] S234. For each cluster Count the number of line segments included. As cluster weight Then, calculate the total weight of all clusters and select the clusters that reach the preset proportion of the total weight as the main clusters.

[0097] In this embodiment, clusters that account for 15% of the total weight are selected as the master clusters, and clusters that are not selected are considered noise and do not participate in the subsequent skeleton construction.

[0098] S24. Perform principal component analysis on each main cluster. Based on the direction of the first principal axis of the analysis results, determine the start and end points of the main transport corridor corresponding to the current main cluster, thereby obtaining the main transport corridor. Further, S24 specifically involves:

[0099] For each main cluster, collect the endpoint coordinates of all weighted line segments within the cluster, perform principal component analysis (PCA) using the endpoint set as a sample, and obtain the first principal axis direction; project all endpoints along the first principal axis direction onto a one-dimensional coordinate axis, and take the endpoints corresponding to the minimum and maximum values ​​of the projection as the start and end points of the main transport corridor corresponding to the current main cluster, respectively, and use the total number of weighted line segments contained in the current main cluster as the weight of the main transport corridor.

[0100] S31. For each main transport corridor, from the starting point to the end point, candidate transfer stations are determined and deployed based on available flight distance, ensuring that the distance between adjacent candidate transfer stations on the same main transport corridor does not exceed the available flight distance. Further, S31 specifically includes:

[0101] For each main transport corridor, calculate the distance from the starting point to the ending point.

[0102] If the distance does not exceed the available flight range, both the origin and destination will be considered as candidate transit stations.

[0103] If the distance is greater than the available flight range, then N intermediate points are inserted at equal intervals on the current main transport corridor, so that the main transport corridor is divided into N+1 equally spaced sub-transport corridors, and the length of the sub-transport corridors does not exceed the available flight range; the starting point, the ending point and all intermediate points are all considered as candidate transit stations.

[0104] In practice, to address resource constraints, the minimum number of N intermediate points should be inserted at equal intervals to meet the requirement (ensuring that the distance between adjacent candidate transfer stations on the same main transport corridor does not exceed the available flight range).

[0105] Further, after processing all main transport corridors according to the method in step S31, a sequence of candidate transfer stations arranged along each main transport corridor is obtained, and all candidate transfer stations constitute a set of candidate transfer stations.

[0106] S32. Determine the connection relationships between different main transport corridors, and select the nearest candidate transfer station as the connection point on the two main transport corridors that need to be connected. Further, S32 specifically involves:

[0107] Obtain the geometric center of each main transport corridor, construct the distance matrix between corridors based on the geometric centers of all main corridors, and use the minimum spanning tree algorithm based on the distance matrix to determine the main transport corridor pairs that need to be connected.

[0108] In this embodiment, the average of the coordinates of the starting point and the ending point of the main transportation corridor is taken as the coordinate value of the geometric center.

[0109] For each pair of main transport corridors that need to be connected, each candidate transfer station in one main transport corridor is connected to all candidate transfer stations in the other main transport corridor, resulting in several corridor connection segments. Only the shortest corridor connection segments are selected and retained, thereby realizing the connection of the main transport corridor pairs.

[0110] In practice, when connecting main transport corridor pairs, the connections are made in descending order of the weight of the main transport corridor.

[0111] S33. For the shortest corridor connection segment in each main transport corridor pair, process it in the same way as in step S31 to ensure that the distance between any two adjacent candidate transfer stations in the shortest corridor connection segment does not exceed the available flight range. All newly constructed candidate transfer stations are aggregated into the candidate transfer station set to obtain a new candidate transfer station set, and thus also obtain an overall connected candidate transfer station skeleton network.

[0112] S33. Using available flight range as a constraint, new candidate transfer stations are established between connection points to connect all main transport corridors, thereby forming a connected candidate transfer station backbone network. Further, step S33 specifically involves:

[0113] S331. For the shortest corridor connection segment in each main transport corridor pair, calculate the distance between the two endpoints.

[0114] If the distance does not exceed the available flight range, both endpoints will be directly designated as candidate transit stations.

[0115] If the distance is greater than the available flight range, M intermediate points are inserted at equal intervals on the current corridor connecting line segment, so that the corridor connecting line segment is divided into M+1 equally spaced sub-transport corridors, and the length of the sub-transport corridor does not exceed the available flight range; the two endpoints and all intermediate points of the corridor connecting line segment are all regarded as candidate transfer stations.

[0116] In practice, to address resource constraints, step S331 involves inserting a minimum number of M intermediate points at equal intervals.

[0117] S332. After processing the shortest connecting line segment between all main transport corridor pairs according to the method in step S331, all candidate transfer stations form a new set of candidate transfer stations, resulting in an overall connected candidate transfer station skeleton network.

[0118] S41. Based on historical orders, determine all demand grid centers and the demand intensity for each demand grid center. Further, S41 specifically involves:

[0119] Grid centers that appear at least once as a start or end point in historical orders will be selected as demand grid centers and filtered out to form a demand grid center set. and determine the set of demand grid centers. The demand intensity at the center of each demand grid.

[0120] Demand intensity is the weight of the weighted line segment corresponding to the center of the demand grid as the starting or ending point.

[0121] For grid centers that have never appeared in historical orders, they are not considered as demand objects, thereby avoiding unnecessary coverage and repair calculations for areas without historical demand.

[0122] S42. Verify whether the distance between each demand grid center and the nearest candidate transfer station in the candidate transfer station skeleton network meets the preset coverage radius. Mark demand grid centers that do not meet the coverage radius as uncovered demand grid centers, and mark demand grid centers that meet the coverage radius as covered demand grid centers. Further, S42 specifically includes:

[0123] In the latest list of candidate transit stations Given its spatial location, the coverage radius is introduced. This is used to represent the service range of a single candidate transfer station to surrounding demand grid centers. In specific implementation, the coverage radius... =0.5 × available range .

[0124] For each demand grid center Calculate the minimum distance from it to the current set of candidate transit stations:

[0125]

[0126] in, Indicates the demand grid center The minimum distance to the set of candidate transit stations; For the current demand grid center; The index represents a candidate transit station in the candidate transit station set; Represents the set of candidate transit stations; This indicates the current demand grid center to the candidate transfer station. Euclidean distance.

[0127] when At that time, the current demand grid center will be... Mark the center of the covered demand grid and add it to the covered set. .

[0128] In practice, this makes Each candidate transit station at the time of establishment All can serve as the current grid center for demand. Provide services.

[0129] when At that time, the current demand grid center will be... Mark the center of the uncovered demand grid and add it to the uncovered set. This leads to the set partitioning:

[0130] ;

[0131] in, For the demand grid center set; For the already covered set; This represents taking the union of sets; This is an uncovered set; This indicates finding the intersection; This represents the empty set.

[0132] S43, For uncovered sets For each uncovered demand grid center, repair processing is carried out sequentially according to the demand intensity corresponding to the uncovered demand grid center. When all uncovered demand grid centers have been repaired, the resulting candidate transfer station skeleton network is used as the UAV transfer station network.

[0133] The specific process for repairing each uncovered demand grid center as the processing target is as follows:

[0134] F1. For each uncovered demand grid center, prioritize shifting the nearest candidate relay station towards the current uncovered demand grid center along the line connecting the current uncovered demand grid center and the nearest candidate relay station, using neighbor set constraints and existing coverage constraints as common constraints. This ensures that the current uncovered demand grid center is within the coverage radius of the shifted candidate relay station. Further, F1 specifically involves:

[0135] Assume the current processing object is the uncovered requirement grid center. Gathering at candidate transfer stations Find the demand center The nearest candidate transit station :

[0136]

[0137] in, This indicates that the demand grid center is not covered. The nearest candidate transit point; This indicates that the objective function The independent variable when the minimum value is reached The value of ; For grid centers that are currently not covered; The index represents a candidate transit station in the set of candidate transit stations.

[0138] Without changing the number of transfer stations, prioritize attempting to move the nearest transfer station within a finite step size. Repair the overlay in this way.

[0139] Set the maximum translation step size (For example ), and take the translation direction as from the nearest transfer station. Pointing to the demand grid center Directional unit vector :

[0140]

[0141] exist Within the range, construct the location of the nearest candidate transfer station after translation:

[0142]

[0143] in, For the nearest candidate transit point Candidate transfer stations after relocation; This is the translation step size; It is a unit vector in direction.

[0144] Given a translation step size The following two types of constraints are checked for the translation scheme:

[0145] Neighbor set constraint: The distance between the translated candidate transfer station and all its adjacent candidate transfer stations in the candidate transfer station skeleton network does not exceed the available flight range. More specifically:

[0146] Based on the candidate relay station skeleton network and the nearest candidate relay station All connected nodes The set formed is called the neighbor set. Check the candidate transfer stations after the translation position Gather with neighbors Each node in Are the distances between them still within the available flight range? ,Right now:

[0147]

[0148] This ensures that the translation operation does not disrupt the local connectivity of the candidate transfer station skeleton network.

[0149] Existing coverage constraints: All existing coverage requirement grid centers within the original candidate transfer station coverage radius must remain within the original candidate transfer station coverage radius. Specifically:

[0150] For those originally in the covered set And from the nearest candidate transit station For each demand grid center m providing services, check the candidate transfer stations after the translation position. To the nearest candidate transit station Does the distance m from the center m of each demand grid providing services still not exceed the coverage radius? This avoids "pulling off" the original covered grid center during the translation process.

[0151] If all the above constraints are satisfied, then the candidate transfer station after translation... Enables currently uncovered demand grid centers satisfy:

[0152]

[0153] Then accept the relocation plan and use the candidate transfer station at the relocated position. Replace the original candidate transfer station The location is determined, and the candidate transfer station set and coverage status are updated.

[0154] If in There is no grid center within the range that simultaneously satisfies the above constraints and can cover the currently uncovered requirements. If the translation position is determined, it is assumed that translating the nearest transit station cannot repair the currently uncovered grid center. .

[0155] F2. If the translation operation in step F1 cannot ensure that the center of the currently uncovered demand grid is within the coverage radius of the translated candidate transfer station, then, constrained by available range and coverage radius, several new candidate transfer stations are inserted at equal intervals along the line connecting the center of the uncovered demand grid and the nearest candidate transfer station. This ensures that the distance between any two adjacent candidate transfer stations on the line does not exceed the available range, and the distance from the last new candidate transfer station on the line to the center of the currently uncovered demand grid is within the coverage radius. In practice, to address resource constraints, a minimum number of new candidate transfer stations are inserted at equal intervals to meet the requirements (ensuring that the distance between any two adjacent candidate transfer stations on the line does not exceed the available range, and the distance from the last new candidate transfer station on the line to the center of the currently uncovered demand grid is within the coverage radius).

[0156] Furthermore, F2 specifically refers to:

[0157] When targeting a certain uncovered set Uncovered demand grid center When the operation of moving the nearest candidate transfer station is not feasible under neighbor set constraints or original coverage constraints, the current uncovered demand grid center is selected. A fallback mechanism is used to repair the connection between the nearest candidate transfer station and the nearest candidate transfer station by inserting a new candidate transfer station.

[0158] set up For currently uncovered demand grid centers The nearest candidate transit station, the distance between the two is denoted as:

[0159]

[0160] Within available range and coverage radius Given the known circumstances, to ensure from After passing through several new candidate transit points, it arrives at the currently uncovered demand grid center. In the path, the method for inserting a new candidate transit station must satisfy:

[0161] The distance between any two adjacent new candidate transit stations shall not exceed the available flight range. And it is far from the currently uncovered demand center. Recent new candidate transit points to grid centers for currently uncovered needs The distance does not exceed the coverage radius .

[0162] The number of newly inserted candidate transfer stations is denoted as . It must meet the following requirements:

[0163]

[0164] in, The number of new candidate transit stations to insert; Indicates rounding up; For currently uncovered demand grid centers Its most recent candidate transit station The distance between them; The coverage radius; Available range.

[0165] Therefore, we can take:

[0166]

[0167] when When this occurs, it indicates that no additional candidate transit stations need to be inserted, and the nearest candidate transit station can be used. In coverage radius Internal direct service to grid centers that do not currently cover demand ;when At that time, it will be at the nearest candidate transit station. Grid center for currently uncovered needs Insert evenly along the connecting line The location of a new candidate transit station can be represented as:

[0168]

[0169] in, Using indexes, retrieve values ​​from 1 to... ; Indicates the first The location of a new candidate transit point; Indicates the position of the point; and These represent the currently uncovered demand grid centers. Location and nearest candidate transit points The location.

[0170] At this point, the distance between adjacent candidate transfer stations does not exceed the available flight range. Furthermore, the last new candidate transfer station is located in a grid center that is currently not covered by the grid. The distance between them does not exceed the coverage radius .

[0171] If a new candidate transfer station needs to be inserted, after inserting the new candidate transfer station, the currently uncovered demand grid centers will be... Its most recent candidate transit station connection segment Treating it as a newly added connecting corridor (a repair path that satisfies the available range constraint), new candidate transit stations are inserted into the newly added connecting corridor in the same way as in step S31, thereby obtaining a new set of candidate transit stations, and thus also a new candidate transit station skeleton network.

[0172] During each repair process, after each translation or insertion of a new transfer station, the candidate transfer station set, candidate transfer station skeleton network, and coverage status are updated, and the demand grid centers that have been converted to the covered status are removed from the uncovered set.

[0173] When the uncovered set When empty, it indicates that all demand grid centers have been connected to the candidate transfer station skeleton network under the constraints of coverage radius and available range, and the repair process is terminated. The candidate transfer station skeleton network obtained at this time is used as the UAV transfer station network, that is, the UAV transfer station layout is obtained.

[0174] This embodiment generates 51 candidate transfer stations, covering 499 demand grid centers, achieving 100% coverage of the demand grid centers and realizing complete coverage. The results are as follows: Figure 6 As shown, data-driven corridor extraction better reflects the direction of demand concentration and constructs a connected network of transit stations.

[0175] Furthermore, by constructing drone transfer stations based on the obtained drone transfer station layout, it is possible to realize the transfer of drones during transport flights in the target area.

[0176] Comparative Example 1:

[0177] This comparative example uses the same study area, the same nominal maximum range of the UAV, the same number of historical orders, and the same steps S11 and S12 as Example 1.

[0178] Then, this comparative example directly selected the same number (51) of candidate transfer stations as the output of Example 1, and randomly selected stations within the study area, such as... Figure 7 As shown, historical order demand is not considered. Under random location selection, 194 demand centers are outside the coverage radius (approximately 2.89 km) and therefore cannot be served, resulting in an actual demand grid center coverage rate of only 61.62%. Furthermore, the random location selection method in this comparative example cannot guarantee connectivity between candidate transit stations. In contrast, the drone transit station network obtained in Example 1 covers all demand grid centers with the same number of stations (51).

[0179] Comparative Example 2:

[0180] This comparative example uses the same study area, the same nominal maximum range of the UAV, the same number of historical orders, and the same steps S11 and S12 as Example 1.

[0181] Then, based on the origin and destination of each historical order, the midpoint of the grid center is calculated, and these unique midpoints are directly used as the corresponding candidate transfer stations. The results are as follows: Figure 8 As shown, there are 489 unique midpoints as intermediate stations (candidate transit stations), but 23 demand centers are still not covered, resulting in an actual demand center coverage rate of only 95.35%. In comparison, the midpoint deployment method in this comparative example generates a huge number of stations, leading to extremely low efficiency. Furthermore, the midpoint deployment method in this comparative example cannot guarantee connectivity between transit stations.

[0182] The method of this invention can be used for planning drone logistics networks within cities or regions, and is suitable for deploying various types of drone transfer station nodes. The transfer stations have both battery swapping and cargo relay functions.

[0183] This invention comprehensively considers the distribution characteristics of historical order data and the coverage of candidate transit station networks for overall demand, and realizes the construction of a drone transit station network framework that meets the main needs, and achieves full coverage of the demand area by the drone transit station network.

[0184] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for layouting unmanned aerial vehicle (UAV) transfer stations based on topology optimization, characterized in that, Includes the following steps: S1. Divide the target area into multiple grids and determine the available range of the UAV under the grid division based on the UAV's nominal maximum range; S2. Map the start and end points of each historical order to the center of the grid, and obtain weighted line segments composed of grid center pairs. Cluster all weighted line segments and extract several main transportation corridors based on the clustering results. S3. On each main transport corridor, multiple candidate transfer stations are deployed according to the available flight range, and all candidate transfer stations on the main transport corridors are connected to form a candidate transfer station backbone network covering all main transport corridors. S4. Based on historical orders, determine all demand grid centers and their demand intensity. With coverage radius and available range as constraints, iteratively repair the candidate transfer station skeleton network. Process the uncovered demand grid centers in order of demand intensity until all demand grid centers are covered, and finally obtain the UAV transfer station network.

2. The method for layout of UAV transfer stations based on topology optimization according to claim 1, characterized in that, Step S1 specifically involves: S11. Based on the UAV's nominal maximum range and the preset grid ratio coefficient, determine the grid side length and divide the target area into multiple grids, with the geometric center of each grid defined as the grid center. S12. Based on the nominal maximum range, the preset range safety factor, and the grid side length, the available range used to constrain the flight distance between transfer stations is calculated.

3. The method for layout of UAV transfer stations based on topology optimization according to claim 1, characterized in that, Step S2 specifically involves: S21. Map the start and end points of each historical order to the center of the grid, construct line segments between grid center pairs, and assign weights to the line segments according to the historical order volume to obtain weighted line segments; S22. For each weighted line segment, extract the direction features and position features, and concatenate them to form the feature vector of the corresponding weighted line segment; S23. Based on the geometric distance between feature vectors, perform density clustering on all weighted line segments to obtain several clusters, and select multiple principal clusters that meet the preset conditions. S24. Perform principal component analysis on each main cluster, and determine the start and end points of the main transport corridor corresponding to the current main cluster based on the direction of the first principal axis of the analysis results, thereby obtaining the main transport corridor.

4. The method for layout of UAV transfer stations based on topology optimization according to claim 3, characterized in that, In step S22: The eigenvector of each weighted line segment is obtained by processing it using the following formula: ， ， ， in, For indexing; For the first The eigenvectors of a weighted line segment; and All are the first The directional characteristics of weighted line segments; and All are the first Positional characteristics of weighted line segments; For the first The direction angle of the weighted line segments; and For the first The x and y coordinates of the midpoint of the weighted line segment; Position weights; denoted as the side length of the grid region.

5. The method for layout of UAV transfer stations based on topology optimization according to claim 3, characterized in that, Step S23 specifically involves: S231. Preset the neighborhood radius and minimum number of core line segments, and use the geometric distance between feature vectors as the weighted distance between line segments; S232. For each weighted line segment, determine whether each weighted line segment is a core weighted line segment based on the distance between weighted line segments: If the number of weighted segments contained within the neighborhood radius of the current weighted segment is not less than the minimum number of core segments, then the current weighted segment is a core weighted segment. Conversely, the current weighted segment is a non-core weighted segment; S233. Based on all core weighted line segments and density-connected weighted line segments, the weighted line segments are divided into several clusters using a density clustering algorithm.

6. The method for layout of UAV transfer stations based on topology optimization according to claim 1, characterized in that, Step S3 specifically involves: S31. For each main transport corridor, from the starting point to the end point, candidate transfer stations are determined and deployed based on the available flight distance, such that the distance between adjacent candidate transfer stations on the same main transport corridor does not exceed the available flight distance. S32. Determine the connection relationship between different main transport corridors, and select the candidate transfer station with the closest distance between them as the connection point on the two main transport corridors that need to be connected; S33. With available flight range as a constraint, new candidate transfer stations are set up between connection points to connect all main transport corridors, thereby forming a connected candidate transfer station backbone network.

7. The method for layout of UAV transfer stations based on topology optimization according to claim 1, characterized in that, Step S4 specifically involves: S41. Determine all demand grid centers based on historical orders, and determine the demand intensity of each demand grid center; S42. Verify whether the distance between each demand grid center and the nearest candidate transfer station in the candidate transfer station skeleton network meets the preset coverage radius. Mark demand grid centers that do not meet the coverage radius as uncovered demand grid centers, and mark demand grid centers that meet the coverage radius as covered demand grid centers. S43. Based on the magnitude of the demand intensity corresponding to the uncovered demand grid center, each uncovered demand grid center is treated as a processing object for repair processing in turn. When all uncovered demand grid centers have been repaired, the candidate transfer station skeleton network obtained is used as the UAV transfer station network.

8. The method for layout of UAV transfer stations based on topology optimization according to claim 7, characterized in that, The specific process for repairing each uncovered demand grid center as the processing target is as follows: F1. For each uncovered demand grid center, prioritize shifting the nearest candidate transfer station towards the current uncovered demand grid center along the line connecting the current uncovered demand grid center and the nearest candidate transfer station, using the neighbor set constraint and the original coverage constraint as common constraints, so that the current uncovered demand grid center is within the coverage radius of the shifted candidate transfer station. F2. If the translation operation in step F1 cannot make the center of the currently uncovered demand grid within the coverage radius of the translated candidate transfer station, then, with the available range and coverage radius as constraints, insert several new candidate transfer stations at equal intervals on the line connecting the center of the uncovered demand grid and the nearest candidate transfer station, so that the distance between any two adjacent candidate transfer stations on the line does not exceed the available range and the distance from the last new candidate transfer station on the line to the center of the currently uncovered demand grid is within the coverage radius.

9. The method for layout of UAV transfer stations based on topology optimization according to claim 8, characterized in that: The neighbor set constraint is that the distance between the translated candidate transfer station and all adjacent candidate transfer stations in the candidate transfer station skeleton network does not exceed the available flight range; The original coverage constraint is that all covered grid centers within the coverage radius of the candidate transfer station before the translation must still be within the coverage radius of the candidate transfer station after the translation.

10. A method for layout of UAV transfer stations based on topology optimization according to claim 8, characterized in that: The number of new candidate transfer stations inserted in step F2 and the position of each new candidate transfer station are set according to the following formula: in, The number of new candidate transit stations to insert; Indicates rounding up; For currently uncovered demand grid centers Its most recent candidate transit station The distance between them; The coverage radius; Available range; Using indexes, retrieve values ​​from 1 to... ; Indicates the first The location of a new candidate transit point; and These represent the currently uncovered demand grid centers. and recent candidate transit stations The location.

Images