A uav landing site selection system based on landing point candidate priority

Abstract

The application relates to the technical field of unmanned planes, in particular to an unmanned plane landing site selection system based on landing site candidate priority, which comprises a processor and a memory storing preset prediction models, an initial unmanned plane landing site set, an initial unmanned plane landing site code set, a predicted flight route set, a landing site weight set, a route weight set, a node intermediate number set, a preset key landing site code set and a preset reference vector; the number of elements 1 in a third landing site code reference vector is used to inversely deduce the number of elements greater than 0 in a second landing site code reference vector, and then the number of corresponding key landing sites in a first landing site code reference vector is inversely deduced; a prediction loss is constructed according to a preset code threshold and the number of elements 1 in the third landing site code reference vector; the parameters of the prediction models are updated; the matching degree of a target priority order vector and a preset precondition is improved; and the site selection rationality of a target unmanned plane landing site is improved.

Description

A UAV take-off and landing point selection system based on candidate take-off and landing point priority Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a UAV take-off and landing point selection system based on candidate priority of take-off and landing points. Background Technology

[0002] Since its inception, the concept of UAM (Urban Air Mobility) has attracted attention from academia, industry, and governments. Its use of eVTOL (Electric Vertical Takeoff and Landing) and unmanned aircraft for transportation can not only meet the transportation needs of various distances within cities, but also improve transportation safety and efficiency.

[0003] Based on urban take-off and landing facilities, drones can alleviate ground traffic pressure to some extent, optimize the work structure of logistics delivery personnel, and meet the high-value, small-volume, and time-sensitive logistics needs of urban residents for express delivery, food delivery, and medicine. Therefore, with the widespread application of drones, the rational selection of drone take-off and landing sites has become crucial.

[0004] As a collection of various new technologies, drones have seen rapid technological development. However, research on automated airport facilities that can be widely adapted to drone-piloted aviation systems in urban areas remains lacking. When selecting drone take-off and landing sites, current technologies typically rely on professionals analyzing local geographical coordinates, altitude, topography, airspace conditions, and other geographical factors to determine the location. This requires significant time and manpower, making it difficult to meet the need for rapid site selection. Furthermore, the scientific validity and rationality of the selected sites are poor, failing to guarantee the reliability, stability, and safety of the drones.

[0005] Therefore, the planning and layout of drone take-off and landing facilities is crucial to meeting urban logistics needs, and how to improve the rationality of drone take-off and landing site selection has become an urgent problem to be solved. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention provides a UAV take-off and landing point selection system based on candidate priority, comprising a processor and a memory storing a computer program. The memory also stores a preset prediction model and an initial set of UAV take-off and landing points A = {A1, A2, ..., A...}. n , ..., A N Initial UAV take-off and landing point code set A 1 ={A 1 1, A1 2, ..., A 1 n , ..., A 1 N The set of predicted flight paths is B = {B1, B2, ..., B}. n , ..., B N}, The set of takeoff and landing point weights Q = {Q1, Q2, ..., Q} n Q N The set of route weights, P = {P1, P2, ..., P} n ..., P N}, Node betweenness set JS = {JS1, JS2, ..., JS} n , ..., JS N The preset key take-off and landing point code set GJ = {GJ1, GJ2, ..., GJ} μ ..., GJ M} and a preset reference vector YS = (1, 1, ..., 1, 0, 0, ..., 0), where A n This refers to the nth initial UAV take-off and landing point, A. 1 n Refers to A n The corresponding initial take-off and landing point code, B n ={B n1 B n2 , ..., B ni , ..., B nI(n)}, B ni Refers to A n The corresponding i-th predicted flight path, Q n Refers to A n The corresponding take-off and landing point weights, P n ={P n1 P n2 ..., P ni ..., P nI(n)}, P ni It refers to B ni The corresponding route weights, JS n Refers to A n The corresponding node betweenness, GJ μ This refers to the key take-off and landing point code corresponding to the preset μ-th key take-off and landing point, where n = 1, 2, ..., N, and N represents the total number of initial UAV take-off and landing points; i = 1, 2, ..., I(n), where I(n) represents the total number of predicted flight paths corresponding to the n-th initial UAV take-off and landing point; and μ = 1, 2, ..., M, where M represents the total number of preset key take-off and landing points. The first Щ elements in YS are 1, and the last N-Щ elements are 0. Щ represents the preset number of target UAV take-off and landing points. When the computer program is executed by the processor, the following steps are implemented:

[0007] S100, Input A, B, Q, P, and JS into the preset prediction model to obtain the prediction priority ranking vector PX = (PX1, PX2, ..., PX...). n ..., PX N ), of which PX n This refers to the initial drone take-off and landing point with a prediction priority of nth.

[0008] S200, perform vector multiplication on PX and YS to obtain the takeoff and landing point reference vector CK. 1 = (PX1, PX2, ..., PX) σ ..., PX Щ ,0,0,...,0), where PX σ This refers to the prediction of the initial UAV take-off and landing point with a priority of σth position, PX Щ This refers to the prediction of the initial UAV take-off and landing point with a priority of Щ, where σ = 1, 2, ..., Щ.

[0009] S300, according to CK 1 A 1 And GJ, obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ ,0,0,...,0), where BM σ It refers to PX σ The corresponding first takeoff and landing point code, BM Щ It refers to PX Щ The corresponding first take-off and landing point code.

[0010] S400, the transpose of GJ and BM 1 Multiply to obtain the second take-off and landing point coded reference vector BM 2 .

[0011] S500, compared to BM 2 Binarization is performed to obtain the third take-off and landing point coded reference vector BM. 3 .

[0012] S600, according to BM 3 Given a preset encoding threshold YZ, the prediction loss Loss = ρ × (YZ - SUM(BM)) is obtained. 3 ), where ρ refers to the preset loss coefficient, SUM(BM 3 ) refers to BM 3 The sum of all elements in .

[0013] S700 updates the parameters of the preset prediction model based on the loss until the loss = 0, and obtains the target prediction model.

[0014] S800 inputs A, B, Q, P and JS into the target prediction model to obtain the target priority ranking vector.

[0015] S900 determines the first 3 initial UAV take-off and landing points in the target priority sorting vector as the target UAV take-off and landing points.

[0016] The present invention has at least the following beneficial effects: A, B, Q, P, and JS are input into a preset prediction model to obtain a prediction priority ranking vector PX = (PX1, PX2, ..., PX...). n ..., PX N By multiplying PX and YS by vectors, the takeoff and landing point reference vector CK is obtained. 1 = (PX1, PX2, ..., PX) σ ..., PX Щ ,0,0,……,0), according to CK 1 A 1 And GJ, obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ (0, 0, ..., 0), the transpose of GJ and BM 1 Multiply to obtain the second take-off and landing point coded reference vector BM 2 , for BM 2 Binarization is performed to obtain the third take-off and landing point coded reference vector BM. 3 According to BM 3 Given a preset encoding threshold YZ, the prediction loss Loss = ρ × (YZ - SUM(BM)) is obtained. 3 The parameters of the preset prediction model are updated based on the loss until loss = 0, thus obtaining the target prediction model. A, B, Q, P, and JS are input into the target prediction model to obtain the target priority ranking vector. The first 3 initial UAV take-off and landing points in the target priority ranking vector are determined as the target UAV take-off and landing points. It can be seen that through BM... 3 The number of elements 1 in the middle, and the reverse calculation of BM. 2 The number of elements greater than 0 in the matrix, and then the BM can be deduced from this. 1 The number of corresponding key take-off and landing points, based on preset encoding thresholds YZ and BM. 3The number of elements 1 in the middle is used to construct the prediction loss, and the parameters of the prediction model are updated accordingly. This improves the matching degree between the target priority ranking vector and the preconditions, thereby improving the rationality of the selection of the target UAV take-off and landing points. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 is a flowchart of the execution computer program of a UAV take-off and landing point selection system based on candidate priority of take-off and landing points provided in Embodiment 1 of the present invention.

[0019] Figure 2 is a flowchart of a method for determining the priority of UAV take-off and landing points based on a complex network model, provided in Embodiment 2 of the present invention.

[0020] Figure 3 is a flowchart of the computer program for constructing a complex network model for UAV take-off and landing point selection, provided in Embodiment 3 of the present invention.

[0021] Figure 4 is a flowchart of a method for generating initial location of UAV take-off and landing points based on multiple factors, provided in Embodiment 4 of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It is understood that, where appropriate, the terms used to distinguish similar objects can be interchanged so that the invention can also be implemented in other embodiments besides the illustrated or described embodiments. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products, or devices.

[0024] Example 1

[0025] This embodiment provides a UAV take-off and landing point selection system based on candidate take-off and landing point priority, as shown in Figure 1. This system includes a processor and a memory storing a computer program. The memory also stores a preset prediction model and an initial set of UAV take-off and landing points A = {A1, A2, ..., A...}. n , ..., A N Initial UAV take-off and landing point code set A 1 ={A 1 1, A 1 2, ..., A 1 n , ..., A 1 N The set of predicted flight paths is B = {B1, B2, ..., B}. n , ..., B N}, The set of takeoff and landing point weights Q = {Q1, Q2, ..., Q} n Q N The set of route weights, P = {P1, P2, ..., P} n ..., P N}, Node betweenness set JS = {JS1, JS2, ..., JS} n , ..., JS N The preset key take-off and landing point code set GJ = {GJ1, GJ2, ..., GJ} μ ..., GJ M} and a preset reference vector YS = (1, 1, ..., 1, 0, 0, ..., 0), where A n This refers to the nth initial UAV take-off and landing point, A. 1 n Refers to A n The corresponding initial take-off and landing point code, B n ={B n1 B n2 , ..., B ni , ..., B nI(n)}, B ni Refers to A n The corresponding i-th predicted flight path, Q n Refers to A n The corresponding take-off and landing point weights, P n ={P n1 P n2 ..., P ni ..., P nI(n)}, P ni It refers to B ni The corresponding route weights, JS n Refers to An The corresponding node betweenness, GJ μ YS refers to the key take-off and landing point code corresponding to the preset μ-th key take-off and landing point, n = 1, 2, ..., N, where N is the total number of initial UAV take-off and landing points, i = 1, 2, ..., I(n), where I(n) is the total number of predicted flight paths corresponding to the n-th initial UAV take-off and landing point, μ = 1, 2, ..., M, where M is the total number of preset key take-off and landing points. The first Щ elements in YS are 1, and the last N-Щ elements are 0. Щ is the preset number of target UAV take-off and landing points.

[0026] The initial drone take-off and landing point can refer to the take-off and landing point in the geographical area where drone take-off and landing facilities need to be selected, which can serve as the location for the layout of drone take-off and landing facilities. There are predicted flight routes between the initial drone take-off and landing points that are suitable for drones to fly to transport goods. The location of the initial drone take-off and landing point and the trajectory of the predicted flight route can be determined by the implementer according to the actual situation.

[0027] The initial UAV take-off and landing point code can be a code that is pre-set by the implementer according to the actual situation, used to identify the initial UAV take-off and landing point, such as 1, 2, 3...

[0028] The take-off and landing point weight can represent the probability of setting up drone take-off and landing facilities at the corresponding initial drone take-off and landing point, the route weight can represent the probability of the predicted flight route being the drone's flight path, and the node betweenness can represent the influence of the corresponding initial drone take-off and landing point relative to other initial drone take-off and landing points.

[0029] Critical take-off and landing points can be the initial UAV take-off and landing points provided by the implementer that require the installation of UAV take-off and landing facilities. These points can be determined by professionals based on the actual conditions of each initial UAV take-off and landing point and on their experience. This serves as a prerequisite for selecting UAV take-off and landing points, meaning that the prediction priority of the M critical take-off and landing points in the final prediction priority ranking vector belongs to the top 3 positions. Preset critical take-off and landing point codes can be pre-defined by the implementer based on the actual situation, used to identify critical take-off and landing points, such as -1, -2, -3, etc.

[0030] The pre-defined prediction model is used to extract and map features from the input unordered initial UAV take-off and landing points, predicted flight paths, node betweenness, take-off and landing point weights, and flight path weights, and outputs an ordered prediction priority ranking vector of the initial UAV take-off and landing points.

[0031] The preset reference vector YS = (1, 1, ..., 1, 0, 0, ..., 0) has the first 3 elements of YS as the value 1 and the last N-3 elements as the value 0. It is used to process the predicted priority ranking vector and extract the first 3 elements of the predicted priority ranking vector as the evaluation of the degree of matching between the ranking result of the predicted priority ranking vector and the preconditions.

[0032] When a computer program is executed by a processor, the following steps are performed:

[0033] S100, Input A, B, Q, P, and JS into the preset prediction model to obtain the prediction priority ranking vector PX = (PX1, PX2, ..., PX...). n ..., PX N ), of which PX n This refers to the initial drone take-off and landing point with a prediction priority of nth.

[0034] The higher the priority ranking, the higher the likelihood that the corresponding initial drone take-off and landing point will be the target drone take-off and landing point for setting up drone take-off and landing facilities.

[0035] In one specific embodiment, the memory also stores a set H = {H1, H2, ..., H...} of the importance of the take-off and landing point locations. n H N The first traffic flow set C = {C1, C2, ..., C} n , ..., C N The set of disturbance levels of the first building is G = {G1, G2, ..., G}. n , ..., G N}, where H n Refers to A n The corresponding importance of the take-off and landing point location, C n Refers to A n The corresponding first traffic flow, G n Refers to A n The corresponding first building interference level, the set of take-off and landing point weights Q, is obtained through the following steps:

[0036] S1, based on H, C, and G, obtain the set of takeoff and landing point weights Q = {Q1, Q2, ..., Q...} n Q N}, where A n The corresponding take-off and landing point weights Q n Meets the following conditions:

[0037] Q n =α1×H n +α2×C n +α3×e^(-G n), where α1 refers to the preset first position weight, α2 refers to the preset first traffic weight, α3 refers to the preset building interference weight, and e refers to the natural constant.

[0038] Among them, the importance of the take-off and landing point location can be used to characterize the importance of the location of the corresponding initial drone take-off and landing point in the drone transportation mission. The first traffic flow can refer to the traffic flow within a certain range corresponding to the initial drone take-off and landing point, which can be used to characterize the demand of the initial drone take-off and landing point for drone cargo transportation. The first building interference level can be used to characterize the degree of interference of buildings near the initial drone take-off and landing point on the drone take-off and landing when the drone takes off and lands at the corresponding initial drone take-off and landing point.

[0039] The greater the importance of the initial drone take-off and landing point, the greater the first traffic flow, and the less interference from the first building, the higher the likelihood of setting up drone take-off and landing facilities at that initial drone take-off and landing point.

[0040] Therefore, the importance of the take-off and landing point location and the first traffic flow are positively correlated with the take-off and landing point weights, while the degree of first building interference is negatively correlated with the take-off and landing point weights. Thus, the set of take-off and landing point weights Q can be obtained based on H, C, and G, which serves as the basis for constructing complex network models and selecting UAV take-off and landing point locations.

[0041] The specific values ​​of α1, α2, and α3 can be set by the implementer according to the actual situation.

[0042] The above measures the impact of the importance of the take-off and landing point location, the first traffic flow, and the first building interference on the possibility of setting up drone take-off and landing facilities at the initial drone take-off and landing point, obtains the take-off and landing point weight corresponding to each initial drone take-off and landing point, improves the accuracy of the take-off and landing point weight, and thus improves the accuracy of the construction of complex network models and the rationality of the selection of drone take-off and landing points.

[0043] In one specific embodiment, the memory also stores a set of flight route anomaly degrees D = {D1, D2, ..., D...} n , ..., D N The set of route lengths E = {E1, E2, ..., E} n , ..., E N}, D n ={D n1 D n2 , ..., D ni , ..., D nI(n)}, D ni It refers to B ni The corresponding degree of flight route abnormality, E n ={E n1 E n2, ..., E ni , ..., E nI(n)}, E ni It refers to B ni The corresponding route length and route weight set P are obtained through the following steps:

[0044] S2, based on C, obtain the second traffic flow set F = {F1, F2, ..., F} corresponding to B. n , ..., F N}, where B n The corresponding second traffic flow list F n ={F n1 F n2 , ..., F ni , ..., F nI(n)}, F ni equals B ni The average of the first traffic flow at the two corresponding initial drone take-off and landing points.

[0045] S3, based on B and H, obtain the set of route location importance corresponding to B, K = {K1, K2, ..., K...} n , ..., K N}, where B n List of corresponding route location importance K n ={K n1 K n2 , ..., K ni , ..., K nI(n)}, K ni equals B ni The average importance of the two initial drone take-off and landing points.

[0046] S4, based on B, D, E, F, and K, obtain the set of route weights P = {P1, P2, ..., P} corresponding to B. n ..., P N}, where P n ={P n1 P n2 ..., P ni ..., P nI(n)}, B ni The corresponding route weight P ni Meets the following conditions:

[0047] P ni =β1×e^(-D ni )+β2×E ni / ((∑ n＝1 N (Σ i＝1 I(n) E ni)) / (Σ n＝1 N I(n)))+β3×F ni +β4×K ni , where β

[0048] 1 refers to the preset abnormal weight, β2 refers to the preset route length weight, β3 refers to the preset second traffic weight, and β4 refers to the preset second position weight.

[0049] Among them, the degree of flight path anomaly can be used to characterize the extent of damage to buildings, people, and other lives and property when a drone experiences flight anomalies on the corresponding predicted flight path.

[0050] The average of the first traffic flow at the two initial drone take-off and landing points corresponding to each predicted flight path can be used as the second traffic flow for the corresponding predicted flight path to characterize the degree of guarantee provided by the predicted flight path for drone cargo transportation, and serve as the basis for obtaining the route weight corresponding to the predicted flight path.

[0051] The average of the importance of the two initial UAV take-off and landing points corresponding to each predicted flight path can characterize the importance of the corresponding predicted flight path location, serving as the basis for obtaining the flight path weights corresponding to the predicted flight path.

[0052] The less abnormal the predicted flight path is, the longer the flight path is, the more secondary traffic flow there is, and the higher the importance of the flight path location, the higher the probability that the predicted flight path is the flight path of the UAV.

[0053] Therefore, the predicted flight path weights are negatively correlated with the degree of flight path anomaly and positively correlated with flight path length, secondary traffic flow, and the importance of flight path location. Thus, based on B, D, E, F, and K, the set of flight path weights P corresponding to B can be obtained, which serves as the basis for constructing complex network models and selecting UAV take-off and landing sites.

[0054] The specific values ​​of β1, β2, β3 and β4 can be set by the implementer according to the actual situation.

[0055] The above measures the impact of flight path anomaly, flight path length, secondary traffic flow, and flight path location importance on the likelihood of a predicted flight path serving as the flight path of a UAV. By obtaining the flight path weight corresponding to each predicted flight path, the accuracy of the flight path weight is improved, thereby improving the accuracy of the construction of complex network models and the rationality of the selection of UAV take-off and landing points.

[0056] In one specific implementation, the node betweenness set JS is obtained through the following steps:

[0057] S5. Based on A, B, Q and P, the initial UAV take-off and landing point is taken as a node, the predicted flight path is taken as an edge, the take-off and landing point weight corresponding to the initial UAV take-off and landing point is taken as the node value of the corresponding node, and the flight path weight corresponding to the predicted flight path is taken as the edge value of the corresponding edge, thus obtaining a complex network model T. The complex network model T is used to filter the target UAV take-off and landing point from A.

[0058] S61, based on T and E, obtain the shortest path between any two nodes in T.

[0059] S62 determines the number of shortest paths passing through the nodes corresponding to the nth initial drone take-off and landing point as the node betweenness JS of the nth initial drone take-off and landing point. n .

[0060] S63, iterate through n = 1, 2, ..., N, and obtain the node betweenness set JS = {JS1, JS2, ..., JS...} n , ..., JS N}

[0061] In one specific embodiment, the memory also stores a clustering coefficient set JL = {JL1, JL2, ..., JL...} n ..., JL N}, JL n Refers to A n The corresponding clustering coefficients and the M preset key take-off and landing points are obtained through the following steps:

[0062] S2000, based on A, B, Q, P, JS, JL and the preset initial weight set QZ = {QZ1, QZ2, QZ3, QZ4, QZ5}, obtain the candidate priority set HX = {HX1, HX2, ..., HX...}. n ..., HX N}, where QZ1 refers to the preset weight of take-off and landing points, QZ2 refers to the preset weight of flight routes, QZ3 refers to the preset weight of the number of flight routes, QZ4 refers to the preset weight of node betweenness, QZ5 refers to the preset weight of clustering coefficient, and A n Corresponding candidate priority HX n Meets the following conditions:

[0063] HX n =QZ1×Q n +QZ2×∑ i＝1 I(n) (P ni / I(n))+QZ3×I(n)+QZ4×JS n +QZ5×JL n .

[0064] S3000, based on HX, selects M candidate drone take-off and landing points from N initial drone take-off and landing points.

[0065] S4000, based on D, E, and F, obtain the candidate take-off and landing point prediction score set FZ = {FZ1, FZ2, ..., FZ}. n ..., FZ N}, where A n The corresponding candidate takeoff and landing point prediction score FZ n Meets the following conditions:

[0066] FZ n =β1×(Σ i＝1 I(n) e^(-D ni ))+β2×(∑ n＝1 N (Σ i＝1 I(n) E ni ))+β3×(∑ n＝1 N (Σ i＝1 I(n) F ni ), where β1

[0067] β1 refers to the preset abnormal weight, β2 refers to the preset route length weight, and β3 refers to the preset second traffic weight.

[0068] S5000, based on FZ, obtain the screening scores FS for M candidate UAV take-off and landing points, where FS meets the following conditions:

[0069] FS=(Σ n＝1 N (FZ n -FZ0) 2 / N, where the predicted mean of candidate take-off and landing points FZ0=Σ n＝1 N (FZ n ) / N.

[0070] S7000, if FS > FS 0 If the initial weight set is not updated, then QZ is updated and replaced with the updated QZ. Step S2000 is repeated until FS ≤ FS. 0 FS≤FS 0 The M candidate UAV take-off and landing points were determined as the M critical take-off and landing points, FS 0 This refers to the preset screening score threshold.

[0071] The more shortest paths there are through the nodes corresponding to the nth initial drone take-off and landing point, the greater the influence of the nth initial drone take-off and landing point, and the higher the probability of setting up drone take-off and landing facilities at that initial drone take-off and landing point.

[0072] S200, perform vector multiplication on PX and YS to obtain the takeoff and landing point reference vector CK. 1 = (PX1, PX2, ..., PX) σ ..., PX Щ ,0,0,...,0), where PX σ This refers to the prediction of the initial UAV take-off and landing point with a priority of σth position, PX Щ This refers to the prediction of the initial UAV take-off and landing point with a priority of Щ, where σ = 1, 2, ..., Щ.

[0073] In this task, based on the selection of Щ target UAV take-off and landing points, PX and YS are multiplied by vector to obtain the take-off and landing point reference vector CK. 1 = (PX1, PX2, ..., PX) σ ..., PX Щ ,0,0,...,0), to extract the initial UAV take-off and landing points ranked in the top Щ positions according to the prediction priority, which serves as the basis for evaluating the ranking effect of the prediction priority ranking vector PX.

[0074] S300, according to CK 1 A 1 And GJ, obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ ,0,0,...,0), where BM σ It refers to PX σ The corresponding first takeoff and landing point code, BM Щ It refers to PX Щ The corresponding first take-off and landing point code.

[0075] In one specific embodiment, S300 further includes the following steps:

[0076] S310, if PX σ If it is a critical takeoff and landing point, then PX will be used. σ The corresponding critical take-off and landing point code is determined to be PX. σ The corresponding first take-off and landing point code.

[0077] S320, if PX σ If it is not a critical takeoff and landing point, then PX will be used. σ The corresponding initial takeoff and landing point code is determined to be PX.σ The corresponding first take-off and landing point code.

[0078] S330, iterate through PX1, PX2, ..., PX σ ..., PX Щ , obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ ,0,0,...,0).

[0079] Specifically, based on the type of the initial UAV take-off and landing points extracted from the first 3 positions, i.e., critical take-off and landing points or non-critical take-off and landing points, the first take-off and landing point code corresponding to each initial UAV take-off and landing point in the first 3 positions is obtained. The code is used to determine the degree of matching between the initial UAV take-off and landing points of the first 3 positions and the preconditions, providing a data basis for measuring the ranking effect of the prediction priority ranking vector PX.

[0080] In one specific implementation, the critical take-off and landing point code is less than 0, and the initial take-off and landing point code is greater than 0.

[0081] The above-mentioned conversion of the initial take-off and landing points of the previous 3 UAVs into corresponding take-off and landing point codes provides a data foundation for judging the degree of matching between the initial take-off and landing points of the previous 3 UAVs and the preconditions through coding, and thus measuring the ranking effect of the prediction priority ranking vector PX.

[0082] S400, the transpose of GJ and BM 1 Multiply to obtain the second take-off and landing point coded reference vector BM 2 .

[0083] Wherein, the matrix dimension of GJ is 1×M, and the transpose of GJ is (GJ). T The dimension is M×1, BM 1 If the matrix dimension is 1×N, then (GJ) T ×BM 1 The obtained second take-off and landing point coded reference vector BM 2 The matrix dimension is M×N.

[0084] Among them, (GJ) T The codes for each key take-off and landing point in the middle are less than 0, BM 1 The first takeoff and landing point code corresponding to the critical takeoff and landing point in the middle is less than 0, BM 1 If the first takeoff and landing point code of the corresponding non-critical takeoff and landing point is greater than 0, then BM 1 The first takeoff and landing point code corresponding to the key takeoff and landing point in (GJ) T The product of the codes for each key take-off and landing point in the BM is greater than 0.1 The corresponding non-critical takeoff and landing points and (GJ) T The product of the codes for each key takeoff and landing point is less than 0; therefore, it can be determined through BM. 2 The number of elements greater than 0 in the matrix, and the reverse calculation of BM. 1 The number of key take-off and landing points corresponds to the prediction priority ranking vector PX, which in turn represents the degree of matching between the prediction priority ranking vector PX and the preconditions.

[0085] For example, BM 2 If the number of elements greater than 0 is M×Amount, then BM can be determined. 1 The number of critical take-off and landing points corresponding to this is Amount.

[0086] S500, compared to BM 2 Binarization is performed to obtain the third take-off and landing point coded reference vector BM. 3 .

[0087] In one specific embodiment, S500 further includes the following steps:

[0088] S510, if BM 2 If the element in the μ-th row and σ-th column is greater than 0, then BM will be... 3 The element in the μ-th row and σ-th column is set to 1.

[0089] S520, if BM 2 If the element in the μ-th row and σ-th column is less than 0, then BM will be... 3 The element in the μ-th row and σ-th column is set to 0.

[0090] S530, BM 3 Set the elements in columns Щ+1 to Д to 0.

[0091] S530, traverse μ = 1, 2, ..., M and σ = 1, 2, ..., Щ to obtain the third take-off and landing point coded reference vector BM. 3 .

[0092] Among them, for BM 2 Binarization is performed, setting elements greater than 0 to 1 and elements less than or equal to 0 to 0. This allows us to modify the BM... 3 The data scale is normalized to avoid the influence of the magnitude of each element value on the prediction loss, thereby improving the accuracy of the prediction loss and thus improving the accuracy of judging the matching degree between the prediction priority ranking vector PX and the preconditions.

[0093] S600, according to BM 3 Given a preset encoding threshold YZ, the prediction loss Loss = ρ × (YZ - SUM(BM)) is obtained. 3), where ρ refers to the preset loss coefficient, SUM(BM 3 ) refers to BM 3 The sum of all elements in .

[0094] In this embodiment, the prerequisite is that the prediction priorities of the M key take-off and landing points in the final prediction priority ranking vector belong to the top 3 positions, and SUM(BM) 3 ) can characterize BM 3 The number of element 1s in BM, which represents the number of elements in BM. 1 The number of key take-off and landing points in the prediction priority ranking vector PX represents the number of key take-off and landing points in the prediction priority ranking vector PX. The more key take-off and landing points there are in the prediction priority ranking vector PX, the higher the degree of matching between the prediction priority ranking vector PX and the preconditions.

[0095] SUM(BM 3 The range of values ​​for ) is [0, M]. 2 Therefore, based on the prediction loss Loss = ρ × (YZ - SUM(BM) 3 The prediction priority ranking vector PX is used to characterize the degree of matching between the prediction priority ranking vector PX and the preconditions, serving as the basis for updating the prediction model parameters to improve the prediction accuracy of the prediction model.

[0096] In one specific implementation, YZ = M 2 .

[0097] In one specific implementation, ρ = 1000.

[0098] The specific value of ρ can be set by the implementer according to the actual situation. For example, ρ = 100, 1000, 10000, etc., to amplify the prediction loss, improve the training effect of the prediction model, and improve the prediction accuracy of the prediction model.

[0099] The above-mentioned method of improving the prediction loss to characterize the degree of matching between the prediction priority ranking vector PX and the preconditions provides a data basis for updating the parameters of the prediction model to improve the prediction accuracy of the prediction model.

[0100] S700 updates the parameters of the preset prediction model based on the loss until the loss = 0, and obtains the target prediction model.

[0101] S800 inputs A, B, Q, P and JS into the target prediction model to obtain the target priority ranking vector.

[0102] S900 determines the first 3 initial UAV take-off and landing points in the target priority sorting vector as the target UAV take-off and landing points.

[0103] The target priority ranking vector obtained by Loss=0 indicates that the number of key take-off and landing points in the predicted priority ranking vector PX is M. This means that the target priority ranking vector matches the preconditions. Therefore, the first Щ initial UAV take-off and landing points in the target priority ranking vector are determined as the target UAV take-off and landing points, which serve as the location for setting up UAV take-off and landing facilities to support tasks such as UAV delivery logistics.

[0104] As described above, A, B, Q, P, and JS are input into a preset prediction model to obtain the prediction priority ranking vector PX = (PX1, PX2, ..., PX...). n ..., PX N By multiplying PX and YS by vectors, the takeoff and landing point reference vector CK is obtained. 1 = (PX1, PX2, ..., PX) σ ..., PX Щ ,0,0,……,0), according to CK 1 A 1 And GJ, obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ (0, 0, ..., 0), the transpose of GJ and BM 1 Multiply to obtain the second take-off and landing point coded reference vector BM 2 , for BM 2 Binarization is performed to obtain the third take-off and landing point coded reference vector BM. 3 According to BM 3 Given a preset encoding threshold YZ, the prediction loss Loss = ρ × (YZ - SUM(BM)) is obtained. 3 The parameters of the preset prediction model are updated based on the loss until loss = 0, thus obtaining the target prediction model. A, B, Q, P, and JS are input into the target prediction model to obtain the target priority ranking vector. The first 3 initial UAV take-off and landing points in the target priority ranking vector are determined as the target UAV take-off and landing points. It can be seen that through BM... 3 The number of elements 1 in the middle, and the reverse calculation of BM. 2 The number of elements greater than 0 in the matrix, and then the BM can be deduced from this. 1 The number of corresponding key take-off and landing points, based on preset encoding thresholds YZ and BM. 3 The number of elements 1 in the middle is used to construct the prediction loss, and the parameters of the prediction model are updated accordingly. This improves the matching degree between the target priority ranking vector and the preconditions, thereby improving the accuracy of the target UAV take-off and landing points.

[0105] Example 2

[0106] This second embodiment provides a method for determining the priority of candidate take-off and landing points for unmanned aerial vehicles (UAVs) based on a complex network model, as shown in Figure 2. This method includes:

[0107] S1000, Obtain the initial set of UAV take-off and landing points A = {A1, A2, ..., A...} n , ..., A N The set of predicted flight paths is B = {B1, B2, ..., B}. n , ..., B N}, The set of takeoff and landing point weights Q = {Q1, Q2, ..., Q} n Q N The set of route weights, P = {P1, P2, ..., P} n ..., P N}, Node betweenness set JS = {JS1, JS2, ..., JS} n , ..., JS N Clustering coefficient set JL = {JL1, JL2, ..., JL} n ..., JL N The set of abnormality levels of flight routes, D = {D1, D2, ..., D...} n , ..., D N The set of route lengths E = {E1, E2, ..., E} n , ..., E N The second traffic flow set F = {F1, F2, ..., F} n , ..., F N}, where A n This refers to the nth initial UAV take-off and landing point, B. n ={B n1 B n2 , ..., B ni , ..., B nI(n)}, B ni Refers to A n The corresponding i-th predicted flight path, Q n Refers to A n The corresponding take-off and landing point weights, P n ={P n1 P n2 ..., P ni ..., P nI(n)}, P ni It refers to B ni The corresponding route weights, JS n Refers to A n The corresponding node betweenness, JL n Refers to An The corresponding clustering coefficient, D n ={D n1 D n2 , ..., D ni , ..., D nI(n)}, D ni It refers to B ni The corresponding degree of flight route abnormality, E n ={E n1 E n2 , ..., E ni , ..., E nI(n)}, E ni It refers to B ni The corresponding route length, F n ={F n1 F n2 , ..., F ni , ..., F nI(n)}, F ni It refers to B ni The corresponding second traffic flow, n = 1, 2, ..., N, where N refers to the total number of initial UAV take-off and landing points, i = 1, 2, ..., I(n), where I(n) refers to the total number of predicted flight paths corresponding to the nth initial UAV take-off and landing point.

[0108] Among them, the take-off and landing point weights represent the probability of setting up drone take-off and landing facilities at the corresponding initial drone take-off and landing point; the flight path weights represent the probability of the predicted flight path being the drone's flight path; the node betweenness coefficients represent the influence of the corresponding initial drone take-off and landing point relative to other initial drone take-off and landing points; and the clustering coefficients represent the degree to which the corresponding initial drone take-off and landing point clusters with other initial drone take-off and landing points.

[0109] S2000, based on A, B, Q, P, JS, JL and the preset initial weight set QZ = {QZ1, QZ2, QZ3, QZ4, QZ5}, obtain the candidate priority set HX = {HX1, HX2, ..., HX...}. n ..., HX N}, where QZ1 refers to the preset weight of take-off and landing points, QZ2 refers to the preset weight of flight routes, QZ3 refers to the preset weight of the number of flight routes, QZ4 refers to the preset weight of node betweenness, QZ5 refers to the preset weight of clustering coefficient, and A n Corresponding candidate priority HX n Meets the following conditions:

[0110] HX n =QZ1×Q n +QZ2×∑ i＝1 I(n) (P ni / I(n))+QZ3×I(n)+QZ4×JS n +QZ5×JL n .

[0111] Among them, the higher the weight of the initial UAV take-off and landing point, the higher the average route weight of the predicted flight route connected to the initial UAV take-off and landing point, the more routes the predicted flight route connected to the initial UAV take-off and landing point, and the higher the node betweenness number and clustering coefficient of the initial UAV take-off and landing point, the higher the probability of setting up UAV take-off and landing facilities at the initial UAV take-off and landing point.

[0112] Therefore, by combining Q, P, JS, JL and the corresponding preset initial weights, the candidate priority corresponding to each initial UAV take-off and landing point is obtained, which indicates that the corresponding initial UAV take-off and landing point is determined as the target UAV take-off and landing point to set the priority of the UAV take-off and landing facility.

[0113] The specific values ​​of QZ1, QZ2, QZ3, QZ4, and QZ5 can be set by the implementer according to the actual situation.

[0114] The above-mentioned combination of the initial UAV take-off and landing point weight, average flight path weight, number of predicted flight paths, node betweenness and clustering coefficient, and preset initial weights, measures the candidate priority of each initial UAV take-off and landing point as the target UAV take-off and landing point, thereby improving the accuracy of candidate priority and thus improving the accuracy of target UAV selection.

[0115] S3000, based on HX, selects M candidate drone take-off and landing points from N initial drone take-off and landing points.

[0116] In one specific embodiment, S3000 further includes the following steps:

[0117] S3100 determines the initial UAV take-off and landing points corresponding to the M highest candidate priorities in HX as candidate UAV take-off and landing points.

[0118] S4000, based on D, E, and F, obtain the candidate take-off and landing point prediction score set FZ = {FZ1, FZ2, ..., FZ}. n ..., FZ N}, where A n The corresponding candidate takeoff and landing point prediction score FZ n Meets the following conditions:

[0119] FZ n =β1×(Σ i＝1 I(n) e^(-D ni ))+β2×(∑ n＝1 N(Σ i＝1 I(n) E ni ))+β3×(∑ n＝1 N (Σ i＝1 I(n) F ni ), where β1

[0120] β1 refers to the preset abnormal weight, β2 refers to the preset route length weight, and β3 refers to the preset second traffic weight.

[0121] Among them, the smaller the degree of flight path anomaly, the longer the flight path length, and the greater the second traffic flow of all predicted flight paths corresponding to the candidate drone take-off and landing point, the lower the demand for drone take-off and landing facilities at the candidate drone take-off and landing point, and the higher the probability of setting up drone take-off and landing facilities at the candidate drone take-off and landing point. Correspondingly, the higher the prediction score of the candidate take-off and landing point.

[0122] The above-mentioned method measures the probability of setting up drone take-off and landing facilities at candidate drone take-off and landing points based on D, E, and F, thereby obtaining the candidate take-off and landing point prediction score set FZ, which improves the accuracy of the candidate take-off and landing point prediction scores.

[0123] S5000, based on FZ, obtain the screening scores FS for M candidate UAV take-off and landing points, where FS meets the following conditions:

[0124] FS=(Σ n＝1 N (FZ n -FZ0) 2 / N, where the predicted mean of candidate take-off and landing points FZ0=Σ n＝1 N (FZ n ) / N.

[0125] In order to ensure that the delivery efficiency and delivery intensity of drones at each take-off and landing point are as balanced as possible, this embodiment aims to make the predicted scores of candidate take-off and landing points among the selected target drone take-off and landing points as consistent as possible.

[0126] The above describes how the difference between the predicted scores of the M candidate drone take-off and landing points is measured to obtain the screening score FS of the M candidate drone take-off and landing points. This score represents the screening accuracy of the M candidate drone take-off and landing points and serves as the basis for updating the initial weight set. This allows for the re-screening of the M take-off and landing points as target drone take-off and landing points, providing locations for drone take-off and landing facilities. This achieves the effect of balancing the delivery efficiency and delivery intensity of drones corresponding to each target drone take-off and landing point, thereby improving the overall working efficiency of drones.

[0127] S6000, if FS > FS0 If the initial weight set is not updated, then QZ is updated and replaced with the updated QZ. Step S2000 is repeated until FS ≤ FS. 0 To obtain FS≤FS 0 The candidate priority set HX is given by FS ≤ FS. 0 The candidate priority set HX is used to filter the target UAV take-off and landing points from N initial UAV take-off and landing points, FS 0 This refers to the preset screening score threshold.

[0128] Among them, the preset screening score threshold FS 0 The specific values ​​can be set by the implementer based on the actual situation.

[0129] The candidate priority corresponding to each initial UAV take-off and landing point is obtained by comprehensively considering the take-off and landing point weight, average flight path weight, number of predicted flight paths, node betweenness coefficient and clustering coefficient, as well as the preset initial weight. Therefore, the accuracy of the preset initial weight affects the accuracy of the candidate priority, and thus affects the accuracy of the selection of candidate UAV take-off and landing points.

[0130] The larger the screening score FS for the M candidate drone take-off and landing points, the greater the difference in delivery efficiency and delivery intensity among the M candidate drone take-off and landing points. Therefore, when FS > FS 0 Update QZ periodically and replace the preset initial weight set with the updated QZ. Repeat step S2000 to obtain M new candidate UAV take-off and landing points until FS≤FS. 0 To obtain FS≤FS 0 The candidate priority set HX is used to select target drone take-off and landing points from N initial drone take-off and landing points, thereby improving the consistency of delivery efficiency and delivery intensity among target drone take-off and landing points and improving the accuracy of drone take-off and landing point selection.

[0131] Based on A, B, Q, P, JS, JL and the preset initial weight set QZ = {QZ1, QZ2, QZ3, QZ4, QZ5}, the candidate priority set HX = {HX1, HX2, ..., HX} is obtained. n ..., HX N According to HX, M candidate drone take-off and landing points are selected from N initial drone take-off and landing points. Based on D, E, and F, the predicted score set of the candidate take-off and landing points is obtained as FZ = {FZ1, FZ2, ..., FZ...}. n ..., FZ N According to FZ, obtain the screening score FS for M candidate drone take-off and landing points. If FS > FS 0If the initial weight set is not found, then update QZ and replace it with the updated QZ. Repeat the above steps until FS ≤ FS. 0 FS≤FS 0 M candidate drone take-off and landing points are determined as the target drone take-off and landing points. It can be seen that by measuring the difference between the predicted scores of the candidate take-off and landing points, the screening score FS of the M candidate drone take-off and landing points is obtained, which represents the screening accuracy of the M candidate drone take-off and landing points. This serves as the basis for updating the initial weight set and the candidate priority set, so as to screen out the target drone take-off and landing points and provide the setting location for drone take-off and landing facilities. This achieves the effect of balancing the delivery efficiency and delivery intensity of the drones corresponding to each target drone take-off and landing point, thereby improving the overall working efficiency of the drones.

[0132] Example 3

[0133] This embodiment provides a complex network model construction system for UAV take-off and landing point selection, as shown in Figure 3. This system includes a processor and a memory storing a computer program. The memory also stores an initial set of UAV take-off and landing points A = {A1, A2, ..., A...}. n , ..., A N The set of predicted flight paths is B = {B1, B2, ..., B}. n , ..., B N The set of importance of take-off and landing point locations is H = {H1, H2, ..., H...} n H N The first traffic flow set C = {C1, C2, ..., C} n , ..., C N The first set of building interference levels is G = {G1, G2, ..., G...} n , ..., G N The set of abnormality levels of flight routes, D = {D1, D2, ..., D...} n , ..., D N The set of route lengths E = {E1, E2, ..., E} n , ..., E N}, where A n This refers to the nth initial UAV take-off and landing point, B. n ={B n1 B n2 , ..., B ni , ..., B nI(n)}, B ni Refers to A n The corresponding i-th predicted flight path, H n Refers to A nThe corresponding importance of the take-off and landing point location, C n Refers to A n The corresponding first traffic flow, G n Refers to A n The corresponding first level of building interference, D n ={D n1 D n2 , ..., D ni , ..., D nI(n)}, D ni It refers to B ni The corresponding flight path anomaly level, E, is used to characterize the degree of damage to the external environment when a UAV experiences flight anomalies along its predicted flight path. n ={E n1 E n2 , ..., E ni , ..., E nI(n)}, E ni It refers to B ni The corresponding flight path length, n = 1, 2, ..., N, where N refers to the total number of initial UAV take-off and landing points, i = 1, 2, ..., I(n), where I(n) refers to the total number of predicted flight paths corresponding to the nth initial UAV take-off and landing point.

[0134] When a computer program is executed by a processor, the following steps are performed:

[0135] S1, based on H, C, and G, obtain the set of takeoff and landing point weights Q = {Q1, Q2, ..., Q...} n Q N}, where A n The corresponding take-off and landing point weights Q n Meets the following conditions:

[0136] Q n =α1×H n +α2×C n +α3×e^(-G n ), where α1 refers to the preset first position weight, α2 refers to the preset first traffic weight, α3 refers to the preset building interference weight, and e refers to the natural constant.

[0137] The specific values ​​of α1, α2, and α3 can be set by the implementer according to the actual situation.

[0138] The above measures the impact of the importance of the take-off and landing point location, the first traffic flow, and the first building interference on the possibility of setting up drone take-off and landing facilities at the initial drone take-off and landing point, obtains the take-off and landing point weight corresponding to each initial drone take-off and landing point, improves the accuracy of the take-off and landing point weight, and thus improves the accuracy of the construction of complex network models and the rationality of the selection of drone take-off and landing points.

[0139] S2, based on C, obtain the second traffic flow set F = {F1, F2, ..., F} corresponding to B. n , ..., F N}, where B n The corresponding second traffic flow list F n ={F n1 F n2 , ..., F ni , ..., F nI(n)}, F ni equals B ni The average of the first traffic flow at the two corresponding initial drone take-off and landing points.

[0140] S3, based on B and H, obtain the set of route location importance corresponding to B, K = {K1, K2, ..., K...} n , ..., K N}, where B n List of corresponding route location importance K n ={K n1 K n2 , ..., K ni , ..., K nI(n)}, K ni equals B ni The average importance of the two initial drone take-off and landing points.

[0141] S4, based on B, D, E, F, and K, obtain the set of route weights P = {P1, P2, ..., P} corresponding to B. n ..., P N}, where P n ={P n1 P n2 ..., P ni ..., P nI(n)}, B ni The corresponding route weight P ni Meets the following conditions:

[0142] P ni =β1×e^(-D ni )+β2×E ni / ((∑ n＝1 N (Σi＝1 I(n) E ni )) / (Σ n＝1 N I(n)))+β3×F ni +β4×K ni , where β

[0143] 1 refers to the preset abnormal weight, β2 refers to the preset route length weight, β3 refers to the preset second traffic weight, and β4 refers to the preset second position weight.

[0144] The above measures the impact of flight path anomaly, flight path length, secondary traffic flow, and flight path location importance on the likelihood of a predicted flight path serving as the flight path of a UAV. By obtaining the flight path weight corresponding to each predicted flight path, the accuracy of the flight path weight is improved, thereby improving the accuracy of the construction of complex network models and the rationality of the selection of UAV take-off and landing points.

[0145] S5. Based on A, B, Q and P, the initial UAV take-off and landing point is taken as a node, the predicted flight path is taken as an edge, the take-off and landing point weight corresponding to the initial UAV take-off and landing point is taken as the node value of the corresponding node, and the flight path weight corresponding to the predicted flight path is taken as the edge value of the corresponding edge, thus obtaining a complex network model T. The complex network model T is used to filter the target UAV take-off and landing point from A.

[0146] Among them, the complex network model T is used to filter out the target drone take-off and landing points from A. The target drone take-off and landing points can serve as the location for setting up drone take-off and landing facilities to support tasks such as drone delivery logistics.

[0147] Based on H, C, and G, the set of takeoff and landing point weights Q = {Q1, Q2, ..., Q} is obtained. n Q N The study measures the impact of the importance of the take-off and landing point location, the first traffic flow, and the first building interference on the probability of setting up drone take-off and landing facilities at the initial drone take-off and landing point. It obtains the take-off and landing point weights corresponding to each initial drone take-off and landing point, improving the accuracy of the take-off and landing point weights. Based on C, it obtains the second traffic flow set F = {F1, F2, ..., F} corresponding to B. n , ..., F N} is used to characterize the degree of assurance that the corresponding predicted flight path provides for the transport of goods by the UAV. Based on B and H, the set of importance of the flight path position corresponding to B is obtained as K = {K1, K2, ..., K}. n , ..., K N Based on B, D, E, F, and K, obtain the set of route weights P = {P1, P2, ..., P} corresponding to B. n ..., P NThis method measures the impact of flight path anomaly degree, flight path length, secondary traffic flow, and flight path location importance on the likelihood of a predicted flight path serving as the flight path of a UAV. It obtains the flight path weights corresponding to each predicted flight path, improving the accuracy of these weights. Based on A, B, Q, and P, it uses the initial UAV take-off and landing points as nodes, the predicted flight paths as edges, the take-off and landing point weights corresponding to the initial UAV take-off and landing points as the node values ​​of the corresponding nodes, and the flight path weights corresponding to the predicted flight paths as the edge values ​​of the corresponding edges, thus obtaining a complex network model T. This model is used to filter target UAV take-off and landing points from A, improving the accuracy of complex network model T construction and consequently improving the accuracy of target UAV take-off and landing point selection, thereby enhancing the rationality of UAV take-off and landing point location selection.

[0148] Example 4

[0149] This embodiment four provides a method for generating initial locations for UAV take-off and landing points based on multiple factors, as shown in Figure 4. This method includes the following steps:

[0150] S10: Obtain Θ original UAV take-off and landing points, several predicted flight paths corresponding to each original UAV take-off and landing point, and the flight path length, second traffic flow, and flight path anomaly degree corresponding to each predicted flight path. The second traffic flow refers to the average of the first traffic flow corresponding to the two original UAV take-off and landing points corresponding to the predicted flight path. The flight path anomaly degree is used to characterize the degree of damage to the outside world when the UAV experiences flight anomalies on the corresponding predicted flight path. Θ is an integer greater than 0.

[0151] In one specific embodiment, S10 further includes the following steps:

[0152] S11, obtain the map data corresponding to Θ original drone take-off and landing points and the first traffic flow corresponding to each original drone take-off and landing point.

[0153] S12, input the map data corresponding to Θ original UAV take-off and landing points and the first traffic flow corresponding to Θ original UAV take-off and landing points into the preset UAV route setting model to obtain several predicted flight routes corresponding to each original UAV take-off and landing point.

[0154] The implementer can use path planning algorithms, such as A* search algorithm, Dijkstra algorithm, genetic algorithm, etc., to quickly generate drone flight paths between the original drone take-off and landing points based on the map data and the first traffic flow corresponding to the original drone take-off and landing points, and use the predicted flight paths for the drone take-off and landing point selection task.

[0155] Map data can include factors such as terrain data, obstacle data, and weather conditions.

[0156] S20: Based on the several predicted flight routes corresponding to each original UAV take-off and landing point, the length of each predicted flight route, and the second traffic flow, obtain the number of predicted flight routes, the total length of the routes, and the total traffic volume corresponding to each original UAV take-off and landing point.

[0157] Specifically, for any original UAV take-off and landing point, the sum of the lengths of all predicted flight routes corresponding to the current original UAV take-off and landing point is determined as the total length of the flight route corresponding to the current original UAV take-off and landing point, and the sum of the second traffic flow of all predicted flight routes corresponding to the current original UAV take-off and landing point is determined as the total traffic volume corresponding to the current original UAV take-off and landing point, which serves as the basis for selecting UAV take-off and landing points.

[0158] S30: Based on the predicted number of flight routes, total route length, total traffic volume, preset route number threshold, preset route length threshold, and preset traffic flow threshold corresponding to each original UAV take-off and landing point, select ζ take-off and landing points to be merged from Θ original UAV take-off and landing points, where 0 < ζ < Θ.

[0159] The specific values ​​of the preset threshold for the number of routes, the preset threshold for the length of routes, and the preset threshold for traffic flow can be set by the implementer according to the actual situation.

[0160] In one specific embodiment, S30 further includes the following steps:

[0161] S31. For any original UAV take-off and landing point, if the number of predicted flight routes corresponding to the current original UAV take-off and landing point is less than the preset number of routes threshold, the total length of routes is less than the preset route length threshold, and the total traffic volume is less than the preset flow threshold, then the current original UAV take-off and landing point is determined as a take-off and landing point to be merged.

[0162] S32, iterate through Θ original drone take-off and landing points to obtain ζ take-off and landing points to be merged.

[0163] The more predicted flight routes, the longer the total length of the routes, and the larger the total traffic volume corresponding to the original drone take-off and landing point, the lower the demand for drone take-off and landing facilities at the original drone take-off and landing point, and the higher the probability of setting up drone take-off and landing facilities at the original drone take-off and landing point. Conversely, the fewer predicted flight routes, the higher the demand for drone take-off and landing facilities, and the lower the probability of setting up drone take-off and landing facilities at the original drone take-off and landing point.

[0164] Therefore, original drone take-off and landing points with low demand and probability of setting up drone take-off and landing facilities can be designated as take-off and landing points to be merged into connected take-off and landing points with high demand and probability of setting up drone take-off and landing facilities. This improves the correlation between take-off and landing points, ensures that the merged take-off and landing points have a higher probability of setting up drone take-off and landing facilities, and improves the coverage and rationality of drone take-off and landing facilities through take-off and landing point merging.

[0165] S40: Based on all predicted flight paths corresponding to each take-off and landing point to be merged, obtain η original UAV take-off and landing points connected to each take-off and landing point to be merged, where η is an integer greater than 0.

[0166] S50: For any take-off and landing point to be merged, take any original UAV take-off and landing point connected to the current take-off and landing point as an intermediate take-off and landing point. Based on the route length between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points and the intermediate take-off and landing point, as well as the route length between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points, obtain the route length priority corresponding to the intermediate take-off and landing point.

[0167] In one specific embodiment, S50 further includes the following steps:

[0168] S51, the average of the flight path lengths between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points is determined as the first flight path length.

[0169] S52, the average of the flight path lengths between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points is determined as the second flight path length.

[0170] S53, the sum of the route length between the current take-off and landing point to be merged and the intermediate take-off and landing point and the second route length is determined as the third route length.

[0171] S54 determines the ratio of the length of the first route to the length of the third route as the route length priority corresponding to the intermediate take-off and landing point.

[0172] The first route length represents the average distance from the current take-off and landing point to the other η-1 original drone take-off and landing points. The second route length represents the average distance from the current intermediate take-off and landing point to the other η-1 original drone take-off and landing points. The third route length represents the average distance from the current take-off and landing point to the intermediate take-off and landing point and then to the other η-1 original drone take-off and landing points, with the intermediate take-off and landing point as a transit point. If the third route length is less than the corresponding first route length, it means that compared with the current take-off and landing point to the other η-1 original drone take-off and landing points, the drone's flight distance is shorter when the intermediate take-off and landing point is used as a transit point. This indicates that the intermediate take-off and landing point has a higher priority than the current take-off and landing point, and merging the take-off and landing point to the intermediate take-off and landing point has a smaller impact on the drone flight demand from the current take-off and landing point to the other η-1 original drone take-off and landing points.

[0173] Correspondingly, the larger the ratio of the length of the first route to the length of the third route, the higher the priority of the route length corresponding to the intermediate take-off and landing points.

[0174] The above-mentioned method obtains the priority of the route length between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points and the intermediate take-off and landing point, as well as the route length between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points, to characterize the priority of the intermediate take-off and landing point compared with the current take-off and landing point to be merged, and the impact of merging the take-off and landing point to be merged into the intermediate take-off and landing point on the UAV flight demand from the current take-off and landing point to be merged to the other η-1 original UAV take-off and landing points. This serves as the basis for merging the take-off and landing points to be merged, thus improving the rationality of take-off and landing point merging.

[0175] S60: Based on the degree of route anomaly of the predicted flight path between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points and the intermediate take-off and landing point, as well as the degree of route anomaly of the predicted flight path between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points, obtain the priority of the route anomaly degree corresponding to the intermediate take-off and landing point.

[0176] In one specific embodiment, S60 further includes the following steps:

[0177] S61, the average of the flight path anomaly levels between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points is determined as the first flight path anomaly level.

[0178] S62, the average of the flight path anomaly levels between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points is determined as the second flight path anomaly level.

[0179] S63, the sum of the abnormality level of the current take-off and landing point to be merged and the intermediate take-off and landing point and the abnormality level of the second route is determined as the abnormality level of the third route.

[0180] S64 determines the ratio of the abnormality level of the first route to that of the third route as the priority of the abnormality level of the route corresponding to the intermediate take-off and landing point.

[0181] The first route anomaly level can characterize the average anomaly level of a drone traveling directly from the current take-off and landing point to the other η-1 original drone take-off and landing points. The second route anomaly level can characterize the average anomaly level of a drone traveling directly from the current intermediate take-off and landing point to the other η-1 original drone take-off and landing points. The third route anomaly level can characterize the average anomaly level of a drone traveling from the current take-off and landing point to the intermediate take-off and landing point and then to the other η-1 original drone take-off and landing points, using the intermediate take-off and landing point as a transit point. If the third route anomaly level is less than the corresponding first route anomaly level, it means that compared with the drone traveling directly from the current take-off and landing point to the other η-1 original drone take-off and landing points, the drone's flight anomaly level is lower when the intermediate take-off and landing point is used as a transit point. This indicates that the intermediate take-off and landing point has a higher priority than the current take-off and landing point to be merged, and merging the take-off and landing point to be merged into the intermediate take-off and landing point has a smaller impact on the safety of the drone.

[0182] Correspondingly, the higher the ratio of the abnormality level of the first route to that of the third route, the higher the priority of the abnormality level of the route corresponding to the intermediate take-off and landing point.

[0183] The above-mentioned method obtains the priority of the flight path anomaly between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points and the intermediate take-off and landing point, as well as the priority of the flight path anomaly between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points. This is used to characterize the priority of the intermediate take-off and landing point compared to the current take-off and landing point to be merged, and the impact of merging the take-off and landing point to be merged into the intermediate take-off and landing point on the safety of the UAV. This serves as the basis for merging the take-off and landing points to be merged, thus improving the rationality of the take-off and landing point merging.

[0184] S70 obtains the merging priority corresponding to the current intermediate take-off and landing point based on the priority of route length and the priority of route anomaly.

[0185] In one specific embodiment, S70 further includes the following steps:

[0186] S71, obtain the preset route length weight and the preset anomaly degree weight.

[0187] S72, based on the route length priority, the preset route length weight, the route anomaly priority, and the preset anomaly weight, obtains the merging priority corresponding to the current intermediate take-off and landing point.

[0188] Specifically, the first product of the route length priority and the preset route length weight, and the second product of the route anomaly priority and the preset anomaly weight are calculated. The sum of the first product and the second product is determined as the merging priority corresponding to the current intermediate take-off and landing point, which represents the priority of merging the current take-off and landing point to be merged into the current intermediate take-off and landing point.

[0189] The specific values ​​of the preset route length weight and the preset anomaly degree weight can be set by the implementer according to the actual situation.

[0190] The above-mentioned method combines route length priority, preset route length weight, route anomaly priority, and preset anomaly weight to obtain the merging priority corresponding to the current intermediate take-off and landing point. This priority is used to represent the priority of merging the current take-off and landing point to be merged into the current intermediate take-off and landing point, which serves as the basis for merging the current take-off and landing point to be merged and improves the rationality of take-off and landing point merging.

[0191] S80: Traverse the η original UAV take-off and landing points connected to the current take-off and landing points to be merged, and obtain η merging priorities.

[0192] S90 merges the current take-off and landing point to be merged with the original UAV take-off and landing point corresponding to the highest merging priority, and determines the merged take-off and landing point as the initial UAV take-off and landing point.

[0193] In this process, take-off and landing points with low demand for drone take-off and landing facilities and low probability of setting up such facilities are merged into other take-off and landing points with high demand for such facilities and high probability of setting up such facilities. This process obtains initial drone take-off and landing points as locations for drone take-off and landing facilities, thereby generating a batch of initial locations that can be used for drone take-off and landing site selection. This ensures that the locations of drone take-off and landing facilities match the demand for take-off and landing points, thereby improving the relative coverage and rationality of drone take-off and landing facilities.

[0194] As described above, based on the number of predicted flight routes, total route length, and total traffic volume corresponding to each original UAV take-off and landing point, and combined with preset route number thresholds, preset route length thresholds, and preset traffic volume thresholds, ζ take-off and landing points to be merged are selected from Θ original UAV take-off and landing points. Based on all predicted flight routes corresponding to each take-off and landing point to be merged, η original UAV take-off and landing points connected to each take-off and landing point to be merged are obtained. For any take-off and landing point to be merged, any original UAV take-off and landing point connected to the current take-off and landing point is taken as an intermediate take-off and landing point. Based on the route length and route anomaly degree between the current take-off and landing point to be merged and the other η-1 original UAV take-off and landing points and the intermediate take-off and landing point, as well as the route length and route anomaly degree between the intermediate take-off and landing point and the other η-1 original UAV take-off and landing points, the route length priority and route anomaly degree corresponding to the intermediate take-off and landing point are obtained. Based on the priority of flight path length and the priority of flight path anomaly, the merging priority corresponding to the current intermediate take-off and landing point is obtained. The current take-off and landing point to be merged is merged with the original drone take-off and landing point corresponding to the highest merging priority, and the merged take-off and landing point is determined as the initial drone take-off and landing point. It can be seen that take-off and landing points with low probability of setting up drone take-off and landing facilities, such as those with a predicted number of flight paths less than the preset number of flight paths threshold, a total flight path length less than the preset flight path length threshold, and a total traffic volume less than the preset traffic flow threshold, are merged into other take-off and landing points with a higher probability of setting up drone take-off and landing facilities. This generates a batch of initial points that can be used for drone take-off and landing site selection, so that the setting up of drone take-off and landing facilities at the initial drone take-off and landing points can match the take-off and landing point requirements, thereby improving the relative coverage and rationality of the drone take-off and landing facilities.

[0195] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. Those skilled in the art should also understand that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of this invention is defined by the appended claims.

Claims

1. A UAV take-off and landing point selection system based on candidate priority of take-off and landing points, characterized in that, The UAV take-off and landing point selection system based on candidate priority includes a processor and a memory storing a computer program. The memory also stores a preset prediction model and an initial set of UAV take-off and landing points A = {A1, A2, ..., A...}. n , ..., A N Initial UAV take-off and landing point code set A 1 ={A 1 1, A 1 2, ..., A 1 n , ..., A 1 N The set of predicted flight paths is B = {B1, B2, ..., B}. n , ..., B N }, The set of takeoff and landing point weights Q = {Q1, Q2, ..., Q} n Q N The set of route weights, P = {P1, P2, ..., P} n ..., P N }, Node betweenness set JS = {JS1, JS2, ..., JS} n , ..., JS N The preset key take-off and landing point code set GJ = {GJ1, GJ2, ..., GJ} μ ..., GJ M } and a preset reference vector YS = (1, 1, ..., 1, 0, 0, ..., 0), where A n This refers to the nth initial UAV take-off and landing point, A. 1 n Refers to A n The corresponding initial take-off and landing point code, B n ={B n1 B n2 , ..., B ni , ..., B nI(n) }, B ni Refers to A n The corresponding i-th predicted flight path, Q n Refers to A n The corresponding takeoff and landing point weights, P n ={P n1 P n2 ..., P ni ..., P nI(n) }, P ni It refers to B ni The corresponding route weights, JS n Refers to A n The corresponding node betweenness, GJ μ This refers to the key take-off and landing point code corresponding to the preset μ-th key take-off and landing point, n = 1, 2, ..., N, where N is the total number of initial UAV take-off and landing points, i = 1, 2, ..., I(n), where I(n) is the total number of predicted flight paths corresponding to the n-th initial UAV take-off and landing point, μ = 1, 2, ..., M, where M is the total number of preset key take-off and landing points. The first Щ elements in YS are 1, and the last N-Щ elements are 0, where Щ is the preset number of target UAV take-off and landing points. When the computer program is executed by the processor, the following steps are implemented: S100, A, B, Q, P, and JS are input into the preset prediction model to obtain the prediction priority sorting vector PX = (PX1, PX2, ..., PX...). n ..., PX N ), of which PX n This refers to predicting the initial UAV take-off and landing point with the nth priority; S200, performs vector multiplication on PX and YS to obtain the take-off and landing point reference vector CK. 1 = (PX1, PX2, ..., PX) σ ..., PX Щ ,0,0,...,0), where PX σ This refers to the prediction of the initial UAV take-off and landing point with a priority of σth position, PX Щ This refers to the prediction of the initial UAV take-off and landing point with a priority of Щ, where σ = 1, 2, ..., Щ; S300, based on CK 1 A 1 And GJ, obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ ,0,0,...,0), where BM σ It refers to PX σ The corresponding first takeoff and landing point code, BM Щ It refers to PX Щ The corresponding first takeoff and landing point code; S400, the transpose of GJ and BM 1 Multiply to obtain the second take-off and landing point coded reference vector BM 2 S500, compared to BM 2 Binarization is performed to obtain the third take-off and landing point coded reference vector BM. 3 S600, according to BM 3 Given a preset encoding threshold YZ, the prediction loss Loss = ρ × (YZ - SUM(BM)) is obtained. 3 ), where ρ refers to the preset loss coefficient, SUM(BM 3 ) refers to BM 3 S700: Sum of all elements; S800: Update the parameters of the preset prediction model according to Loss until Loss = 0, and obtain the target prediction model; S900: Input A, B, Q, P and JS into the target prediction model to obtain the target priority sorting vector; S900: Determine the first 3 initial UAV take-off and landing points in the target priority sorting vector as the target UAV take-off and landing points.

2. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 1, characterized in that, The memory also stores a set H = {H1, H2, ..., H...} of the importance of the take-off and landing points. n H N The first traffic flow set C = {C1, C2, ..., C} n , ..., C N The set of disturbance levels of the first building is G = {G1, G2, ..., G}. n , ..., G N }, where H n Refers to A n The corresponding importance of the take-off and landing point location, C n Refers to A n The corresponding first traffic flow, G n Refers to A n The corresponding first building interference level, the take-off and landing point weight set Q is obtained through the following steps: S1, based on H, C, and G, obtain the take-off and landing point weight set Q = {Q1, Q2, ..., Q...} n Q N }, where A n The corresponding take-off and landing point weights Q n The following conditions must be met: Q n =α1×H n +α2×C n +α3×e^(-G n ), where α1 refers to the preset first position weight, α2 refers to the preset first traffic weight, α3 refers to the preset building interference weight, and e refers to the natural constant.

3. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 2, characterized in that, The memory also stores a set of flight route anomaly levels D = {D1, D2, ..., D...} n , ..., D N The set of route lengths E = {E1, E2, ..., E} n , ..., E N }, D n ={D n1 D n2 , ..., D ni , ..., D nI(n) }, D ni It refers to B ni The corresponding degree of flight route abnormality, E n ={E n1 E n2 , ..., E ni , ..., E nI(n) }, E ni It refers to B ni The corresponding route length and route weight set P are obtained through the following steps: S2, based on C, obtain the second traffic flow set F = {F1, F2, ..., F} corresponding to B. n , ..., F N }, where B n The corresponding second traffic flow list F n ={F n1 F n2 , ..., F ni , ..., F nI(n) }, F ni equals B ni The average of the first traffic flow at the two corresponding initial drone take-off and landing points; S3, based on B and H, obtain the set of importance of the flight path location corresponding to B, K = {K1, K2, ..., K...} n , ..., K N }, where B n List of corresponding route location importance K n ={K n1 K n2 ..., K ni ..., K nI(n) }, K ni equals B ni The average importance of the two initial UAV take-off and landing points; S4, based on B, D, E, F, and K, obtain the set of route weights P = {P1, P2, ..., P...} corresponding to B. n ..., P N }, where P n ={P n1 P n2 ..., P ni ..., P nI(n) }, B ni The corresponding route weight P ni P meets the following conditions: ni =β1×e^(-D ni )+β2×E ni / ((∑ n＝1 N (Σ i＝1 I(n) E ni )) / (Σ n＝1 N I(n)))+β3×F ni +β4×K ni Wherein, β1 refers to the preset abnormal weight, β2 refers to the preset route length weight, β3 refers to the preset second traffic weight, and β4 refers to the preset second location weight.

4. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 3, characterized in that, The node betweenness set JS is obtained through the following steps: S5, based on A, B, Q, and P, the initial UAV take-off and landing points are taken as nodes, the predicted flight paths are taken as edges, the take-off and landing point weights corresponding to the initial UAV take-off and landing points are taken as the node values ​​of the corresponding nodes, and the flight path weights corresponding to the predicted flight paths are taken as the edge values ​​of the corresponding edges, thus obtaining a complex network model T, wherein the complex network model T is used to filter the target UAV take-off and landing points from A; S61, based on T and E, the shortest path between any two nodes in T is obtained; S62, the number of shortest paths passing through the nodes corresponding to the nth initial UAV take-off and landing point is determined as the node betweenness set JS corresponding to the nth initial UAV take-off and landing point. n S63, iterate through n = 1, 2, ..., N to obtain the node betweenness set JS = {JS1, JS2, ..., JS...} n , ..., JS N } 5. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 3, characterized in that, The memory also stores a clustering coefficient set JL = {JL1, JL2, ..., JL...} n ..., JL N }, JL n Refers to A n The corresponding clustering coefficients and M preset key take-off and landing points are obtained through the following steps: S2000, based on A, B, Q, P, JS, JL and the preset initial weight set QZ = {QZ1, QZ2, QZ3, QZ4, QZ5}, the candidate priority set HX = {HX1, HX2, ..., HX} is obtained. n ..., HX N }, where QZ1 refers to the preset weight of take-off and landing points, QZ2 refers to the preset weight of flight routes, QZ3 refers to the preset weight of the number of flight routes, QZ4 refers to the preset weight of node betweenness, QZ5 refers to the preset weight of clustering coefficient, and A n Corresponding candidate priority HX n Meets the following conditions: HX n =QZ1×Q n +QZ2×∑ i＝1 I(n) (P ni / I(n))+QZ3×I(n)+QZ4×JS n +QZ5×JL n S3000, based on HX, select M candidate UAV take-off and landing points from N initial UAV take-off and landing points; S4000, based on D, E, and F, obtain the candidate take-off and landing point prediction score set FZ = {FZ1, FZ2, ..., FZ...} n ..., FZ N }, where A n The corresponding candidate takeoff and landing point prediction score FZ n FZ meets the following conditions: n =β1×(Σ i＝1 I(n) e^(-D ni ))+β2×(∑ n＝1 N (Σ i＝1 I(n) E ni ))+β3×(∑ n＝1 N (Σ i＝1 I(n) F ni ), where β1 refers to the preset anomaly weight, β2 refers to the preset flight path length weight, and β3 refers to the preset second traffic weight; S5000, according to FZ, obtain the screening score FS of M candidate UAV take-off and landing points, where FS meets the following condition: FS=(Σ n＝1 N (FZ n -FZ0) 2 / N, where the predicted mean of candidate take-off and landing points FZ0=Σ n＝1 N (FZ n ) / N; S7000, if FS > FS 0 If the initial weight set is not updated, then QZ is updated and replaced with the updated QZ. Step S2000 is repeated until FS ≤ FS. 0 FS≤FS 0 The M candidate UAV take-off and landing points were determined as the M critical take-off and landing points, FS 0 This refers to the preset screening score threshold.

6. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 1, characterized in that, S300 also includes the following step: S310, if PX σ If it is a critical takeoff and landing point, then PX will be used. σ The corresponding critical take-off and landing point code is determined to be PX. σ The corresponding first takeoff and landing point code; S320, if PX σ If it is not a critical takeoff and landing point, then PX will be used. σ The corresponding initial takeoff and landing point code is determined to be PX. σ The corresponding first takeoff and landing point code; S330, traversing PX1, PX2, ..., PX σ ..., PX Щ , obtain CK 1 The corresponding first take-off and landing point coded reference vector BM 1 = (BM1, BM2, ..., BM σ , ..., BM Щ ,0,0,...,0).

7. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 6, characterized in that, S500 also includes the following step: S510, if BM 2 If the element in the μ-th row and σ-th column is greater than 0, then BM will be... 3 Set the element in the μ-th row and σ-th column of S520 to 1; 2 If the element in the μ-th row and σ-th column is less than 0, then BM will be... 3 Set the element in the μ-th row and σ-th column of S530 to 0; 3 Set the elements in columns Щ+1 to Д to 0; S530, traverse μ = 1, 2, ..., M and σ = 1, 2, ..., Щ to obtain the third take-off and landing point encoding reference vector BM. 3 .

8. The UAV take-off and landing point selection system based on candidate take-off and landing point priority according to claim 7, characterized in that, YZ=M 2 。

Images