Collision prediction and early warning method for multi-UAV logistics transportation

Abstract

The present invention discloses a multi-UAV logistics transportation collision prediction and early warning method, which belongs to the technical field of UAV transportation management. It solves the problem that the traditional UAV logistics transportation method in the prior art does not have a UAV early warning design; the present invention determines the coordinates of the logistics transportation point according to the UAV flight area, and determines all the logistics transportation points within the UAV logistics distribution range by calculating the distance relationship between the numbered logistics transportation point and the predicted UAV flight distance; obtains the minimum logistics transportation route distance by calculation, and selects the minimum element therein to obtain the shortest logistics transportation route for all UAVs; selects the area where the shortest logistics transportation routes of all UAVs intersect, analyzes the risk probability of direct collision between any two UAVs, and obtains the UAV early warning result. The present invention effectively improves the safety of UAV transportation, avoids collisions between UAVs during transportation, and can be applied to planning UAV logistics transportation routes.

Description

Technical Field

[0001] The invention relates to a multi-UAV logistics transportation collision prediction and early warning method, belonging to the technical field of UAV transportation management. Background Art

[0002] The application of drones in the logistics industry is changing the way goods are delivered, especially in terms of improving efficiency and reducing costs. In urban areas, drones can effectively avoid ground traffic congestion and directly deliver small goods to the destination quickly, greatly improving the speed and timeliness of delivery. In addition, drones have lower operating costs than traditional means of transportation such as trucks or motorcycles, especially in short-distance and small-batch delivery scenarios. This cost advantage is more obvious. Therefore, the use of drones not only improves the efficiency of logistics distribution, but also significantly reduces the overall logistics costs, especially in the express delivery industry and urgent delivery services.

[0003] From the perspective of environmental impact and sustainability, drone transportation also shows its advantages. Due to the reduction in the use of ground transportation, drone delivery plays an important role in alleviating traffic congestion and reducing carbon emissions. Especially when drones are powered by electricity and the energy comes from renewable energy, their negative impact on the environment will be further reduced, which not only helps the logistics industry to shift to a more environmentally friendly operating model, but also promotes the environmental sustainability of the entire society. As an efficient, environmentally friendly and flexible delivery method, drones can play a significant role in alleviating urban traffic congestion, reducing carbon emissions, increasing delivery speed and reducing operating costs. By analyzing the transportation capacity of drones in detail, companies and policymakers can better evaluate their applicability and benefits in specific scenarios. Analysis of drone transportation capacity can also reveal the challenges and opportunities faced by drone transportation in terms of technology, economies of scale and regulatory compliance, providing data support and decision-making basis for future technological improvements and policy formulation.

[0004] The current drone multi-point transportation system faces many challenges. The main problems are low transportation efficiency, insufficient ability to cope with dynamic environments, limited scale expansion and poor cost control. These are due to the shortcomings of existing transportation allocation strategies in path planning, task scheduling and cost-effectiveness. At the same time, when multiple drones are used to fly simultaneously in logistics transportation, the risk of collision naturally increases due to space sharing and possible overlapping flight routes. Drone collisions may cause damage to or complete failure of the drones, thereby affecting the safety of the goods carried, causing material losses or delayed delivery. If drones collide in densely populated areas or over high-value facilities, their falling fragments or the entire drone may cause harm to people on the ground, or damage to property such as buildings and vehicles. Therefore, the risk of drone collisions not only concerns the safety of drones and goods, but also involves broader issues of public safety and property protection.

[0005] In the prior art, a patent document with a publication (announcement) number of CN112520040A discloses a drone and a drone system for logistics transportation, which designs a drone for logistics transportation and has the ability to automatically load and unload goods, which is achieved through a cargo bin and an intelligent sliding box at the bottom of the frame. The box design includes a reverse-threaded slide plate, which fixes the goods by clamping them with springs, thereby preventing the goods from falling during transportation and achieving shock absorption protection, significantly improving the efficiency and quality of logistics transportation; however, the prior art does not consider the impact of various factors on the drone's transportation capacity during the drone's flight, resulting in a low actual logistics transportation capacity of drones that have not been assigned for transportation, making it difficult to meet actual transportation needs, and does not consider drone flight collision warnings.

[0006] Therefore, a multi-UAV logistics transportation collision prediction and warning method is needed. Summary of the invention

[0007] A brief overview of the present invention is provided below in order to provide a basic understanding of certain aspects of the present invention. It should be understood that this overview is not an exhaustive overview of the present invention. It is not intended to identify key or important parts of the present invention, nor is it intended to limit the scope of the present invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to a more detailed description discussed later.

[0008] In view of this, in order to solve the problem that the traditional UAV logistics transportation method in the prior art does not have a UAV early warning design, the present invention provides a multi-UAV logistics transportation collision prediction and early warning method.

[0009] The technical solution is as follows: A multi-UAV logistics transportation collision prediction and early warning method includes the following steps:

[0010] S1. Collect data based on the evaluation indicators, build a quantitative analysis evaluation indicator system for UAV flight distance, and obtain the element matrix corresponding to the evaluation indicators;

[0011] S2. According to the element matrix corresponding to the evaluation index, identify the abnormal data of the UAV influencing factors and obtain the updated UAV evaluation index element matrix;

[0012] S3. Build a UAV flight distance prediction model, input the updated UAV evaluation index element matrix to train the model, and output the predicted UAV flight distance;

[0013] S4. Determine the coordinates of the logistics transportation points according to the drone flight area, and determine all logistics transportation points within the drone logistics distribution range by calculating the distance relationship between the numbered logistics transportation points and the predicted drone flight distance;

[0014] S5. According to all the logistics transportation points within the UAV logistics distribution range, the minimum logistics transportation route distance is obtained by calculation, and the minimum element is selected to obtain the shortest logistics transportation route for all UAVs;

[0015] S6. Select the area where the shortest logistics transportation routes of all drones intersect, analyze the risk probability of direct collision between any two drones, and obtain the drone warning results.

[0016] Further, in said S1, the evaluation indexes include battery capacity, motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity, flight altitude and flight distance;

[0017] The quantitative analysis and evaluation index system of UAV flight distance is specifically expressed as follows:

[0018] The battery capacity element matrix corresponding to the battery capacity is , , is the number of drones, For the Battery capacity corresponding to each drone;

[0019] The motor efficiency element matrix corresponding to the motor efficiency is: , , For the Motor efficiency corresponding to each drone;

[0020] The propeller diameter element matrix corresponding to the propeller diameter is , , For the Propeller diameter corresponding to each drone;

[0021] The fuselage mass element matrix corresponding to the fuselage mass is: , , For the The body mass corresponding to each drone;

[0022] The load element matrix corresponding to the load is: , , For the The corresponding load of each drone;

[0023] The flight speed element matrix corresponding to the flight speed is: , , For the The flight speed of each drone;

[0024] The flight mode element matrix corresponding to the flight mode is , , For the The flight modes corresponding to each drone;

[0025] The wind speed element matrix corresponding to the wind speed is , , For the Wind speed corresponding to each drone;

[0026] The wind direction element matrix corresponding to the wind direction is , , For the The wind direction corresponding to each drone;

[0027] The temperature element matrix corresponding to the temperature is , , For the The temperature corresponding to each drone;

[0028] The humidity element matrix corresponding to humidity is , , For the Humidity corresponding to each drone;

[0029] The flight height element matrix corresponding to the flight height is: , , For the The flight altitude corresponding to each drone;

[0030] The flight distance element matrix corresponding to the flight distance is , , For the The flight distance of each drone.

[0031] Furthermore, in S2, the process of identifying abnormal battery capacity data is as follows: The average battery capacity of each drone , The standard deviation of the battery capacity of each drone and the relative distance between any data and the mean , comparing the relative distance between individual data points and the mean The relationship with 3 is , to determine the battery capacity of the drone Normal data is retained. , to determine the battery capacity of the drone The above operation is performed on all data, and all abnormal data are deleted to obtain the updated battery capacity element matrix. , , The updated battery capacity data corresponding to the WNth drone;

[0032] The average battery capacity of each drone It is expressed as:

[0033] ;

[0034] The standard deviation of the battery capacity of each drone It is expressed as:

[0035] ;

[0036] The relative distance between any data and the mean It is expressed as:

[0037] ;

[0038] According to the abnormal battery capacity data identification process, the updated motor efficiency element matrix is ​​established , Updated propeller diameter element matrix , Updated fuselage mass element matrix , updated load element matrix , Updated flight speed element matrix , Updated flight mode element matrix , Updated wind speed element matrix , Updated wind direction element matrix , the updated temperature element matrix , updated humidity element matrix , updated flight height element matrix and the updated flight distance element matrix .

[0039] Furthermore, in S3, the UAV flight distance prediction model includes an input layer, 7 hidden layers, a self-attention layer, 2 residual layers and an output layer;

[0040] The neural network structure of the established UAV flight distance prediction model is connected in the following order:

[0041] Input layer: Updated battery capacity element matrix , Updated motor efficiency element matrix , Updated propeller diameter element matrix , Updated fuselage mass element matrix , updated load element matrix , Updated flight speed element matrix , Updated flight mode element matrix , Updated wind speed element matrix , Updated wind direction element matrix , the updated temperature element matrix , updated humidity element matrix , updated flight height element matrix and the updated flight distance element matrix As input;

[0042] Hidden layer 1: Number of neurons: 512, activation function: ReLU, batch normalization;

[0043] Hidden layer 2: number of neurons: 256, activation function: ReLU, batch normalization;

[0044] Self-attention layer: input the output matrix X of hidden layer 2; initialize the weight matrix and establish the initial weight query matrix in turn , initial weight key matrix and the initial weight matrix , transform the output matrix X into a query matrix , key matrix K and value matrix V, , , ; Calculate the query matrix The similarity with the key matrix K gives the attention score matrix A, ,in, is the scaling factor; the softmax function is used to normalize the attention score matrix A so that the sum of the elements in each row is 1, and the normalized matrix B is obtained and output. ;

[0045] Hidden layer 3: Number of neurons: 128, activation function: ReLU, batch normalization;

[0046] Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection.

[0047] Hidden layer 4: number of neurons: 64, activation function: ReLU, batch normalization;

[0048] Hidden layer 5: number of neurons: 32, activation function: ReLU, batch normalization;

[0049] Hidden layer 6: number of neurons: 16, activation function: ReLU, batch normalization;

[0050] Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection;

[0051] Hidden layer 7: Number of neurons: 8, activation function: ReLU, batch normalization;

[0052] Output layer: Number of neurons: 1, outputs the predicted flight distance of the drone.

[0053] Furthermore, in said S4, a three-dimensional coordinate system is established for the logistics transportation area, the coordinates of the logistics transportation points are determined, and the first The coordinates of the logistics transport points are marked as , ,in, =1, 2, 3, ..., F, where F is the total number of logistics transportation points, For the The horizontal coordinate of the logistics transportation point, For the The vertical coordinate of the logistics transportation point, For the The vertical coordinates of the logistics transportation points are marked as the coordinates of the first drone landing pad. , ,in, is the horizontal coordinate of the drone landing pad, is the y value of the coordinate of the drone landing pad, is the z value of the coordinates of the drone landing pad;

[0054] For the first logistics transport point , , calculate the distance between the first logistics transportation point and the drone landing pad ;

[0055] The first distance between the first logistics transport point and the drone landing pad It is expressed as:

[0056] ;

[0057] The predicted flight distance of the first drone based on the output , compare the first distance The predicted flight distance of the first drone The relationship between When , it indicates that the distance between the first logistics transportation point and the drone landing pad exceeds the flight range of the first drone, and it is determined that the first logistics transportation point is not within the logistics distribution range of the first drone. When , it indicates that the first logistics transportation point is within the logistics distribution range of the first drone. Based on this, the distance between other logistics transportation points and the first logistics transportation point is calculated. ;

[0058] The distance between other logistics transportation points and the first logistics transportation point It is expressed as:

[0059] ;

[0060] The distance between all other logistics transportation points and the first logistics transportation point Sort from small to large and number the logistics transport points in the above order. The coordinates of the logistics transport points are , ,in, After the number The horizontal coordinate of the logistics transportation point, After the number The vertical coordinate of the logistics transportation point, After the number The vertical coordinates of the logistics transportation points, the coordinates of the first logistics transportation point after numbering are , ,in, is the horizontal coordinate of the first logistics transportation point after numbering, is the vertical coordinate of the first logistics transportation point after numbering, is the vertical coordinate of the first logistics transportation point after numbering, and the first logistics transportation point is calculated The first logistics transport point after the number The first center point coordinates between , ,in, is the horizontal coordinate of the first center point, , is the ordinate of the first center point, , is the vertical coordinate of the first center point, , calculate the distance between the first center point and the first drone landing pad ;

[0061] The second distance between the first center point and the first drone landing pad It is expressed as:

[0062] ;

[0063] Compare the second distance The predicted flight distance of the first drone The distance between When it is, it indicates the first logistics transportation point after the number Not within the range of the first drone logistics delivery, when When it is, it indicates the first logistics transportation point after the number In the first drone logistics delivery range, Based on this, the coordinates of the second logistics transportation point are , ,in, is the horizontal coordinate of the second logistics transportation point after numbering, is the ordinate of the second logistics transport point after numbering. The vertical coordinate of the second logistics transportation point after numbering is calculated to obtain the first logistics transportation point The first logistics transport point after the number And the second logistics transport point after the number The coordinates of the second center point between , ,in, is the horizontal coordinate of the second midpoint, , is the ordinate of the second midpoint, , is the vertical coordinate of the second midpoint, , calculate the distance between the second center point and the first drone landing pad ;

[0064] The third distance between the second center point and the first drone landing pad It is expressed as:

[0065] ;

[0066] Compare the third distance The predicted flight distance of the first drone The distance between When it is, it indicates the second logistics transportation point after the number Not within the range of the first drone logistics delivery, when When it is, it indicates the second logistics transportation point after the number Within the first drone logistics delivery range;

[0067] Repeat the above steps to determine all logistics transportation points within the entire drone logistics distribution range.

[0068] Furthermore, in S5, all logistics transportation points within the logistics distribution range of the first drone are numbered, and all logistics transportation points within the logistics distribution range of the first drone are recorded as , , …, ,in, =The number of all logistics transport points within the logistics distribution range of the first drone, and the number of the first drone landing point Defined as the starting point of the transportation route, for the remaining logistics transportation points except the first logistics transportation point within the logistics delivery range of the first drone Random sorting is performed, and the remaining logistics transportation points except the first logistics transportation point within the logistics distribution range of the first drone after random sorting are recorded as , the random sorting order is that the drone starts from the first logistics transportation point within the logistics delivery range of the first drone The order of logistics delivery after takeoff, and the total flight distance of the drone in the randomly sorted order ;

[0069] Total flight distance of the drone in the randomly sorted order It is expressed as:

[0070] ;

[0071] in, It is expressed as the flight distance between the first drone landing point and the first logistics transportation point within the first drone logistics distribution range. It represents the flight distance between the first logistics transportation point within the logistics delivery range of the first drone and the second logistics transportation point within the logistics delivery range of the first drone after random sorting. Indicates the second to last logistics transport point within the logistics delivery range of the first drone after random sorting To the last logistics transport point within the logistics delivery range of the first drone after random sorting The flight distance between

[0072] All logistics transport points within the first drone's logistics delivery range As the first logistics delivery point of the drone, it will deliver the remaining logistics transportation points within the logistics delivery range of the first drone except the first logistics transportation point. Perform GM random sorts, and the order after all random sorts is different. Calculate the total flight distance of all randomly sorted drones , , …, ,in, The first logistics transportation point for the drone from the first drone logistics delivery range After takeoff, the total flight distance of the drone required for flight is calculated according to the random sorting order obtained in the GMth time, and the total flight distance matrix of the drone is established. , ;

[0073] Filter and obtain the total flight distance matrix of the drone The smallest element in That is, the first logistics transportation point within the logistics distribution range of the first drone As the minimum logistics transportation route distance corresponding to the first logistics delivery point of the drone, ;

[0074] Repeat the above steps to select the remaining logistics transport points except the first logistics transport point within the logistics delivery range of the first drone. As the first logistics delivery point of the drone, calculate the corresponding minimum logistics transportation route distance , , …, , establish the minimum logistics transportation route distance matrix , ;

[0075] Screening to obtain the minimum logistics transportation route distance matrix The smallest element in That is, the minimum flight distance of the first drone in the multi-point logistics transportation process, and its corresponding flight route is the shortest logistics transportation route;

[0076] According to the total number of drones WN, repeat the above steps WN times to obtain the shortest logistics transportation route for all drones.

[0077] Furthermore, the total number of the same logistics transportation points of the shortest logistics transportation route of the first drone and the shortest logistics transportation route of the second drone is recorded as , the same logistics transportation points of the shortest logistics transportation route of the first drone and the shortest logistics transportation route of the second drone are numbered 1, 2, ..., ;

[0078] According to the collected flight speed of the first drone and the flight speed of the second drone , traverse the logistics transportation points where the logistics transportation route of the first drone is the same as that of the second drone, and calculate the time it takes for the first drone to fly to the The time for the same logistics transport point and the second drone flew to The time for the same logistics transport point ,in, =1, 2, …, ;

[0079] The first drone flew to The time for the same logistics transport point It is expressed as:

[0080] ;

[0081] The second drone flew to The time for the same logistics transport point It is expressed as:

[0082] ;

[0083] in, Fly the first drone to The total flight distance of drones between the same logistics transportation points, For the second drone to fly to The total flight distance of drones between the same logistics transportation points;

[0084] when When , it indicates that there is a risk of collision between the two drones. For the safe time range, When , it indicates that there is no risk of collision between the two drones;

[0085] When there is a risk of collision between two drones, the data of all logistics transportation points of the first drone are collected, and the transfer probability of the first drone flying from the current position to the collision direction is calculated at each logistics transportation point. , that is, the transfer probability of the first UAV flying from the current position to the position with collision risk, and the probability of the first UAV flying from the current position to the collision position is calculated ;

[0086] The probability of the first drone flying from its current position to the collision position It is expressed as:

[0087] ;

[0088] Among them, KM is the number of logistics transportation points between the initial position of the first UAV and the position with collision risk;

[0089] Collect data from all logistics transport points of the second drone, and calculate the transfer probability of the second drone flying from its current position to the collision direction at each logistics transport point , that is, the transfer probability of the second UAV flying from the current position to the position with collision risk, and the probability of the second UAV flying from the current position to the collision position is calculated ;

[0090] The probability of the second drone flying from its current position to the collision position It is expressed as:

[0091] ;

[0092] in, is the number of logistics transport points between the initial position of the second UAV and the position with collision risk;

[0093] Calculate the risk probability of collision between the first UAV and the second UAV , set the risk probability threshold , compare the risk probability of collision between the first drone and the second drone Risk probability threshold The relationship between When , it indicates that the probability of collision between the two drones is high and an early warning is needed. , it indicates that the probability of a collision between the two drones is low and no warning is needed.

[0094] The beneficial effects of the present invention are as follows: the present invention systematically evaluates the performance of drones by establishing a quantitative analysis and evaluation index system for drone flight distance, and clarifies the flight capability of drones under different conditions; the present invention identifies abnormal data related to the data of factors affecting the flight distance of drones, ensures the accuracy and reliability of the data, and improves the accuracy of the estimation of the flight distance of drones; the present invention establishes a drone flight distance prediction model, which can predict the flight capability of drones under specific conditions, optimizes the use of drones, and ensures that drones can plan tasks within the range allowed by endurance, avoiding power exhaustion or other risks in the middle of the journey; the present invention clarifies the transportation capacity of drones by determining all logistics transportation points within the logistics distribution range of all drones, and effectively supports distribution efficiency optimization, transportation risk management and cost control; the present invention can formulate the optimal drone logistics transportation route method, effectively improve the drone logistics transportation capacity and logistics transportation efficiency, reduce transportation risks, and improve customer satisfaction; the present invention predicts and warns drone collisions, which can ensure the safety of multiple drones when performing logistics transportation at the same time, reduce the accident rate, and help improve logistics efficiency and reduce operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0096] Figure 1 Schematic diagram of the process of multi-UAV logistics transportation collision prediction and warning method. DETAILED DESCRIPTION

[0097] In order to make the technical solutions and advantages of the embodiments of the present invention more clearly understood, the exemplary embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than an exhaustive list of all the embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0098] refer to Figure 1 The present embodiment is described in detail, and the method for predicting and warning collisions of multiple UAVs in logistics transportation specifically includes the following steps:

[0099] S1. Collect data based on the evaluation indicators, build a quantitative analysis evaluation indicator system for UAV flight distance, and obtain the element matrix corresponding to the evaluation indicators;

[0100] S2. According to the element matrix corresponding to the evaluation index, identify the abnormal data of the UAV influencing factors and obtain the updated UAV evaluation index element matrix;

[0101] S3. Build a UAV flight distance prediction model, input the updated UAV evaluation index element matrix to train the model, and output the predicted UAV flight distance;

[0102] S4. Determine the coordinates of the logistics transportation points according to the drone flight area, and determine all the logistics transportation points within the drone logistics distribution range by calculating the distance relationship between the numbered logistics transportation points and the predicted drone flight distance.

[0103] S5. According to all the logistics transportation points within the UAV logistics distribution range, the minimum logistics transportation route distance is obtained by calculation, and the minimum element is selected to obtain the shortest logistics transportation route for all UAVs;

[0104] S6. Select the area where the shortest logistics transportation routes of all drones intersect, analyze the risk probability of direct collision between any two drones, and obtain the drone warning results.

[0105] Furthermore, in S1, the quantitative analysis and evaluation index system of the UAV flight distance includes battery capacity, motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity, flight altitude and flight distance;

[0106] The battery capacity determines the total amount of electrical energy that the drone can use, thus affecting its continuous flight time. The battery capacity element matrix corresponding to the battery capacity is: , , is the number of drones, For the Battery capacity corresponding to each drone;

[0107] Motor efficiency refers to the efficiency of the motor in converting electrical energy into mechanical energy. A high-efficiency motor can increase the flight time. The motor efficiency element matrix corresponding to the motor efficiency is: , , For the Motor efficiency corresponding to each drone;

[0108] The propeller size affects the thrust and flight efficiency of the drone. The propeller diameter directly affects the thrust and aerodynamic performance it provides. The propeller diameter element matrix corresponding to the propeller diameter is: , , For the Propeller diameter corresponding to each drone;

[0109] The fuselage mass affects the take-off and flight stability of the UAV. The fuselage mass element matrix corresponding to the fuselage mass is: , , For the The body mass corresponding to each drone;

[0110] The load refers to the extra weight that the drone can carry in addition to its own weight. The load element matrix corresponding to the load is , , For the The corresponding load of each drone;

[0111] The flight speed directly affects the flight distance and time. The flight speed element matrix corresponding to the flight speed is: , , For the The flight speed of each drone;

[0112] The flight mode determines different energy consumption and operating conditions. The flight modes include manual mode, stable mode, hovering mode, heading mode, return mode, follow mode, waypoint flight, and track flight. For quantitative comparison, the above modes are coded as 1-8 in sequence. The flight mode element matrix corresponding to the flight mode is , , For the The flight modes corresponding to each drone;

[0113] Wind speed will affect the flight efficiency and control of the drone. The wind speed element matrix corresponding to the wind speed is: , , For the Wind speed corresponding to each drone;

[0114] Wind direction affects the flight path and energy consumption of the drone. The wind direction angle is used to quantitatively analyze and compare the wind direction. The wind direction element matrix corresponding to the wind direction is: , , For the The wind direction corresponding to each drone;

[0115] Temperature affects battery performance and air density, which in turn affects flight. The temperature element matrix corresponding to temperature is: , , For the The temperature corresponding to each drone;

[0116] High humidity will affect the efficiency and reliability of electronic equipment. The humidity element matrix corresponding to humidity is: , , For the Humidity corresponding to each drone;

[0117] The flight altitude affects the air density and flight efficiency. The flight altitude element matrix corresponding to the flight altitude is: , , For the The flight altitude corresponding to each drone;

[0118] The flight distance element matrix corresponding to the flight distance is , , For the The flight distance corresponding to each drone;

[0119] Specifically, the flight distance of the drone directly affects the logistics transportation capacity. Therefore, the flight distance of the drone is used as a quantitative indicator of the logistics transportation capacity of the drone. At the same time, there are many factors that affect the flight distance of the drone, including battery capacity, motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity, flight altitude, etc. In order to quantify the impact of each factor on the flight distance of the drone, it is necessary to first clarify the quantitative analysis indicators of each factor, and establish the corresponding element matrix based on the data collected from public channels.

[0120] Furthermore, in S2, the process of identifying abnormal battery capacity data is as follows: The average battery capacity of each drone , The standard deviation of the battery capacity of each drone and the relative distance between any data and the mean , comparing the relative distance between individual data points and the mean The relationship with 3 is , to determine the battery capacity of the drone Normal data is retained. , to determine the battery capacity of the drone The above operation is performed on all data, and all abnormal data are deleted to obtain the updated battery capacity element matrix. , , For the updated Data on the battery capacity of each drone;

[0121] The average battery capacity of each drone It is expressed as:

[0122] ;

[0123] The standard deviation of the battery capacity of each drone It is expressed as:

[0124] ;

[0125] The relative distance between any data and the mean It is expressed as:

[0126] ;

[0127] According to the abnormal battery capacity data identification process, the updated motor efficiency element matrix is ​​established , Updated propeller diameter element matrix , Updated fuselage mass element matrix , updated load element matrix , Updated flight speed element matrix , Updated flight mode element matrix , Updated wind speed element matrix , Updated wind direction element matrix , the updated temperature element matrix , updated humidity element matrix , updated flight height element matrix and the updated flight distance element matrix , , , , , , , , , , , , ,in, For the updated The motor efficiency data corresponding to each drone, For the updated The motor efficiency data corresponding to each drone, For the updated The data of the fuselage mass corresponding to each drone, For the updated The corresponding load data of each drone, For the updated The flight speed data of each drone, For the updated The data of the flight modes corresponding to each drone, For the updated The wind speed data corresponding to each drone, For the updated The wind direction data corresponding to each drone, For the updated The temperature data corresponding to each drone, For the updated The humidity data corresponding to each drone, For the updated The data of the flight altitude corresponding to each drone, For the updated The data of flight distance corresponding to each drone;

[0128] Specifically, the number of elements in all element matrices needs to remain the same, so the number of elements in the updated element matrix, that is, the amount of data after deleting abnormal data, remains .

[0129] Furthermore, in S3, the UAV flight distance prediction model includes an input layer, 7 hidden layers, a self-attention layer, 2 residual layers and an output layer;

[0130] The neural network structure of the established UAV flight distance prediction model is connected in the following order:

[0131] Input layer: Updated battery capacity element matrix , Updated motor efficiency element matrix , Updated propeller diameter element matrix , Updated fuselage mass element matrix , updated load element matrix , Updated flight speed element matrix , Updated flight mode element matrix , Updated wind speed element matrix , Updated wind direction element matrix , the updated temperature element matrix , updated humidity element matrix , updated flight height element matrix and the updated flight distance element matrix As input;

[0132] Hidden layer 1: Number of neurons: 512, activation function: ReLU, batch normalization;

[0133] Hidden layer 2: number of neurons: 256, activation function: ReLU, batch normalization;

[0134] Self-attention layer: input the output matrix X of hidden layer 2; initialize the weight matrix and establish the initial weight query matrix in turn , initial weight key matrix and the initial weight matrix , transform the output matrix X into a query matrix , key matrix K and value matrix V, , , ; Calculate the query matrix The similarity with the key matrix K gives the attention score matrix A, ,in, is the scaling factor; the softmax function is used to normalize the attention score matrix A so that the sum of the elements in each row is 1, and the normalized matrix B is obtained and output. , the output after normalization is a weighted average value matrix, that is, each element in the sequence adjusts its representation according to its correlation with other elements; the normalized matrix B captures the important relationships within the output matrix X of hidden layer 2 and serves as the input of the next layer;

[0135] Hidden layer 3: Number of neurons: 128, activation function: ReLU, batch normalization;

[0136] Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection.

[0137] Hidden layer 4: number of neurons: 64, activation function: ReLU, batch normalization;

[0138] Hidden layer 5: number of neurons: 32, activation function: ReLU, batch normalization;

[0139] Hidden layer 6: number of neurons: 16, activation function: ReLU, batch normalization;

[0140] Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection;

[0141] Hidden layer 7: Number of neurons: 8, activation function: ReLU, batch normalization;

[0142] Output layer: Number of neurons: 1, outputs the predicted flight distance of the drone;

[0143] Specifically, when the UAV flight distance prediction model is applied, it only needs to input the battery capacity, motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity, and flight altitude data of the UAV to output the estimated flight distance of the UAV;

[0144] Compared with traditional mathematical models, neural network models have advantages in UAV flight distance prediction. First, neural network models can effectively capture and model the complex nonlinear relationship between input features, enabling neural network models to handle problems that are difficult to solve with traditional linear models. For example, the flight distance of a UAV is affected by many factors, and the interactions between these factors may be very complex. Neural networks can automatically learn and extract effective features from large amounts of data, reducing the need for manual selection and design of features. In the prediction of UAV flight distance, the network can automatically learn the key factors that determine the flight distance from the raw data without the need for specialized domain knowledge to guide feature engineering.

[0145] The present invention constructs a UAV flight distance prediction model based on a neural network, which mainly includes a hidden layer, a self-attention layer, a residual layer, etc., wherein the hidden layer can help the network capture the nonlinear relationship in the input data through a nonlinear activation function, and can be used to characterize the nonlinear influence of multiple factors on the flight distance of the UAV. Each hidden layer can be regarded as performing a feature conversion, gradually converting the original input into an internal representation with more predictive value for the output variable. In the above process, the deep network structure helps to extract more abstract high-level features; the self-attention layer enables the network to evaluate the importance of other features when processing each input feature, and dynamically adjusts the weight distribution by learning the contribution of different features to the prediction task, so as to more flexibly capture the interaction between the features and improve the accuracy of the prediction and the adaptability of the model; the residual layer ensures that the initial information can directly affect the final output by adding the features of the early layer to the features of the subsequent layer, which is very helpful for maintaining the information integrity of the network in the deep structure.

[0146] Furthermore, in said S4, a three-dimensional coordinate system is established for the logistics transportation area, the coordinates of the logistics transportation points are determined, and the first The coordinates of the logistics transport points are marked as , ,in, =1, 2, 3, ..., F, where F is the total number of logistics transportation points, For the The horizontal coordinate of the logistics transportation point, For the The vertical coordinate of the logistics transportation point, For the The vertical coordinates of the logistics transportation points are marked as the coordinates of the first drone landing pad. , ,in, is the horizontal coordinate of the drone landing pad, is the y value of the coordinate of the drone landing pad, is the z value of the coordinates of the drone landing pad;

[0147] For the first logistics transport point , , calculate the distance between the first logistics transportation point and the drone landing pad ;

[0148] The first distance between the first logistics transport point and the drone landing pad It is expressed as:

[0149] ;

[0150] The predicted flight distance of the first drone based on the output , compare the first distance The predicted flight distance of the first drone The relationship between When , it indicates that the distance between the first logistics transportation point and the drone landing pad exceeds the flight range of the first drone, and it is determined that the first logistics transportation point is not within the logistics distribution range of the first drone. When , it indicates that the first logistics transportation point is within the logistics distribution range of the first drone. Based on this, the distance between other logistics transportation points and the first logistics transportation point is calculated. ;

[0151] The distance between other logistics transportation points and the first logistics transportation point It is expressed as:

[0152] ;

[0153] The distance between all other logistics transportation points and the first logistics transportation point Sort from small to large and number the logistics transport points in the above order. The coordinates of the logistics transport points are , ,in, After the number The horizontal coordinate of the logistics transportation point, After the number The vertical coordinate of the logistics transportation point, After the number The vertical coordinates of the logistics transportation points, the coordinates of the first logistics transportation point after numbering are , ,in, is the horizontal coordinate of the first logistics transportation point after numbering, is the vertical coordinate of the first logistics transportation point after numbering, is the vertical coordinate of the first logistics transportation point after numbering, and the first logistics transportation point is calculated The first logistics transport point after the number The first center point coordinates between , ,in, is the horizontal coordinate of the first center point, , is the ordinate of the first center point, , is the vertical coordinate of the first center point, , calculate the distance between the first center point and the first drone landing pad ;

[0154] The second distance between the first center point and the first drone landing pad It is expressed as:

[0155] ;

[0156] Compare the second distance The predicted flight distance of the first drone The distance between When it is, it indicates the first logistics transportation point after the number Not within the range of the first drone logistics delivery, when When it is, it indicates the first logistics transportation point after the number In the first drone logistics delivery range, Based on this, the coordinates of the second logistics transportation point are , ,in, is the horizontal coordinate of the second logistics transportation point after numbering, is the ordinate of the second logistics transport point after numbering. The vertical coordinate of the second logistics transportation point after numbering is calculated to obtain the first logistics transportation point The first logistics transport point after the number And the second logistics transport point after the number The coordinates of the second center point between , ,in, is the horizontal coordinate of the second midpoint, , is the ordinate of the second midpoint, , is the vertical coordinate of the second midpoint, , calculate the distance between the second center point and the first drone landing pad ;

[0157] The third distance between the second center point and the first drone landing pad It is expressed as:

[0158] ;

[0159] Compare the third distance The predicted flight distance of the first drone The distance between When it is, it indicates the second logistics transportation point after the number Not within the range of the first drone logistics delivery, when When it is, it indicates the second logistics transportation point after the number In the logistics delivery range of the first drone, all logistics transportation points within the logistics delivery range of the first drone have been determined;

[0160] Repeat the above steps to determine all logistics transportation points within the range of all drone logistics distribution, that is, to identify the logistics transportation points after numbering , , …, The predicted flight distance of the first drone The distance relationship between them and the numbered logistics transport points , , …, Is it within the logistics delivery range of the first drone?

[0161] Furthermore, in S5, all logistics transportation points within the logistics distribution range of the first drone are numbered, and all logistics transportation points within the logistics distribution range of the first drone are recorded as , , …, ,in, =The number of all logistics transport points within the logistics distribution range of the first drone, and the number of the first drone landing point Defined as the starting point of the transportation route, for the remaining logistics transportation points except the first logistics transportation point within the logistics delivery range of the first drone Random sorting is performed, and the remaining logistics transportation points except the first logistics transportation point within the logistics distribution range of the first drone after random sorting are recorded as , the random sorting order is that the drone starts from the first logistics transportation point within the logistics delivery range of the first drone The order of logistics delivery after takeoff, and the total flight distance of the drone in the randomly sorted order ;

[0162] Total flight distance of the drone in the randomly sorted order It is expressed as:

[0163] ;

[0164] in, It is expressed as the flight distance between the first drone landing point and the first logistics transportation point within the first drone logistics distribution range. It represents the flight distance between the first logistics transportation point within the logistics delivery range of the first drone and the second logistics transportation point within the logistics delivery range of the first drone after random sorting. Indicates the second to last logistics transport point within the logistics delivery range of the first drone after random sorting To the last logistics transport point within the logistics delivery range of the first drone after random sorting The flight distance between

[0165] All logistics transport points within the first drone's logistics delivery range As the first logistics delivery point of the drone, it will deliver the remaining logistics transportation points within the logistics delivery range of the first drone except the first logistics transportation point. Perform GM random sorts, and the order after all random sorts is different. Calculate the total flight distance of all randomly sorted drones , , …, ,in, The first logistics transportation point for the drone from the first drone logistics delivery range After takeoff, the total flight distance of the drone required for flight is calculated according to the random sorting order obtained in the GMth time, and the total flight distance matrix of the drone is established. , ;

[0166] Filter and obtain the total flight distance matrix of the drone The smallest element in That is, the first logistics transportation point within the logistics distribution range of the first drone As the minimum logistics transportation route distance corresponding to the first logistics delivery point of the drone, ;

[0167] Repeat the above steps to select the remaining logistics transport points except the first logistics transport point within the logistics delivery range of the first drone. As the first logistics delivery point of the drone, calculate the corresponding minimum logistics transportation route distance , , …, , establish the minimum logistics transportation route distance matrix , ;

[0168] Screening to obtain the minimum logistics transportation route distance matrix The smallest element in That is, the minimum flight distance of the first drone in the multi-point logistics transportation process, and its corresponding flight route is the shortest logistics transportation route;

[0169] According to the total number of drones WN, repeat the above steps WN times to obtain the shortest logistics transportation route for all drones.

[0170] Furthermore, in S6, the total number of the same logistics transportation points of the shortest logistics transportation route of the first drone and the shortest logistics transportation route of the second drone is recorded as , the same logistics transportation points of the shortest logistics transportation route of the first drone and the shortest logistics transportation route of the second drone are numbered 1, 2, ..., ;

[0171] According to the collected flight speed of the first drone and the flight speed of the second drone , traverse the logistics transportation points where the logistics transportation route of the first drone is the same as that of the second drone, and calculate the time it takes for the first drone to fly to the The time for the same logistics transport point and the second drone flew to The time for the same logistics transport point ,in, =1, 2, …, ;

[0172] The first drone flew to The time for the same logistics transport point It is expressed as:

[0173] ;

[0174] The second drone flew to The time for the same logistics transport point It is expressed as:

[0175] ;

[0176] in, Fly the first drone to The total flight distance of drones between the same logistics transportation points, For the second drone to fly to The total flight distance of drones between the same logistics transportation points;

[0177] when When , it indicates that there is a risk of collision between the two drones. For the safe time range, When , it indicates that there is no risk of collision between the two drones;

[0178] When there is a risk of collision between two drones, the data of all logistics transportation points of the first drone are collected, and the transfer probability of the first drone flying from the current position to the collision direction is calculated at each logistics transportation point. , that is, the transfer probability of the first UAV flying from the current position to the position with collision risk, and the probability of the first UAV flying from the current position to the collision position is calculated ;

[0179] The probability of the first drone flying from its current position to the collision position It is expressed as:

[0180] ;

[0181] Among them, KM is the number of logistics transportation points between the initial position of the first UAV and the position with collision risk;

[0182] Collect data from all logistics transport points of the second drone, and calculate the transfer probability of the second drone flying from its current position to the collision direction at each logistics transport point , that is, the transfer probability of the second UAV flying from the current position to the position with collision risk, and the probability of the second UAV flying from the current position to the collision position is calculated ;

[0183] The probability of the second drone flying from its current position to the collision position It is expressed as:

[0184] ;

[0185] in, is the number of logistics transport points between the initial position of the second UAV and the position with collision risk;

[0186] Calculate the risk probability of collision between the first UAV and the second UAV , set the risk probability threshold , compare the risk probability of collision between the first drone and the second drone Risk probability threshold The relationship between When , it indicates that the probability of collision between the two drones is high and an early warning is needed. , it indicates that the probability of a collision between the two drones is low and no warning is needed.

[0187] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is selected primarily for readability and teaching purposes, rather than for explaining or defining the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the present invention is illustrative, not restrictive, with respect to the scope of the present invention, which is defined by the appended claims.

Claims

1. A multi-UAV logistics transportation collision prediction and warning method, characterized in that: The following steps are involved: S1. Collect data based on the evaluation indicators, build a quantitative analysis evaluation index system for the flight distance of UAVs, and obtain the element matrix corresponding to the evaluation indicators; S2. According to the element matrix corresponding to the evaluation index, identify the abnormal data of the UAV influencing factors and obtain the updated UAV evaluation index element matrix; S3. Build a UAV flight distance prediction model, input the updated UAV evaluation index element matrix to train the model, and output the predicted UAV flight distance; S4. Determine the coordinates of the logistics transportation points according to the drone flight area, and determine all logistics transportation points within the drone logistics distribution range by calculating the distance relationship between the numbered logistics transportation points and the predicted drone flight distance; S5. According to all the logistics transportation points within the UAV logistics distribution range, the minimum logistics transportation route distance is obtained by calculation, and the minimum element is selected to obtain the shortest logistics transportation route for all UAVs; S6. Select the area where the shortest logistics transportation routes of all drones intersect, analyze the risk probability of direct collision between any two drones, and obtain the drone warning result; In S4, a three-dimensional coordinate system is established for the logistics transportation area, the coordinates of the logistics transportation points are determined, and the coordinates of the i-th logistics transportation point are marked as C i , C i =[x i y i z i ], where i = 1, 2, 3, ..., F, F is the total number of logistics transportation points, x i is the horizontal coordinate of the i-th logistics transportation point, y i is the ordinate of the i-th logistics transport point, z i is the vertical coordinate of the i-th logistics transportation point, and the coordinate of the first drone landing pad is marked as C w , C w =[x w y w z w ], where x w is the horizontal coordinate of the drone landing pad, y w is the y value of the coordinate of the drone landing pad, z w is the z value of the coordinates of the drone landing pad; For the first logistics transportation point C1, C1 = [x1 y1 z1], the distance D between the first logistics transportation point and the drone landing pad is calculated 1-w ; The first distance D between the first logistics transportation point and the drone landing pad 1-w It is expressed as: According to the output predicted flight distance JL1 of the first UAV, compare the first distance D 1-w The relationship between the predicted flight distance of the first UAV JL1 and the predicted flight distance of the first UAV JL1 is 1-w ≥JL1, indicating that the distance between the first logistics transportation point and the drone landing apron exceeds the flight range of the first drone, and it is determined that the first logistics transportation point is not within the logistics distribution range of the first drone. 1-w When <JL1, it indicates that the first logistics transportation point is within the logistics distribution range of the first drone, in D 1-w <JL1, calculate the distance D between other logistics transportation points and the first logistics transportation point 1-i ; The distance D between other logistics transportation points and the first logistics transportation point 1-i It is expressed as: The distance D between all other logistics transportation points and the first logistics transportation point 1-i Sort from small to large, and number the logistics transportation points in the above order. The coordinates of the i-th logistics transportation point after numbering are E i , E i =[x ei y ei z ei ], where x ei is the horizontal coordinate of the numbered i-th logistics transportation point, y ei is the ordinate of the numbered i-th logistics transport point, z ei is the vertical coordinate of the numbered ith logistics transport point, the coordinate of the first logistics transport point after numbering is E1, E1 = [x e1 y e1 z e1 ], where x e1 is the horizontal coordinate of the first logistics transportation point after numbering, y e1 is the ordinate of the first logistics transport point after numbering, z e1 is the vertical coordinate of the first logistics transportation point after numbering, and the first center point coordinate CE1 between the first logistics transportation point C1 and the first logistics transportation point E1 after numbering is calculated, CE1 = [x ce1 y ce1 z ce1 ], where x ce1 is the horizontal coordinate of the first center point, y ce1 is the ordinate of the first center point, z ce1 is the vertical coordinate of the first center point, Calculate the distance D between the first center point and the first drone landing pad 1-CE1 ; The second distance D between the first center point and the first drone landing pad 1-CE1 It is expressed as: Compare the second distance D 1-CE1 The distance between the predicted first UAV flight distance JL1, when D 1-CE1 ≥JL1, indicating that the first logistics transport point E1 after the numbering is not within the logistics distribution range of the first drone. 1-CE1 When <JL1, it indicates that the first logistics transportation point E1 after the numbering is within the logistics distribution range of the first drone, in D 1-CE1 <JL1, the coordinates of the second logistics transportation point are E2, E2=[x e2 y e2 z e2 ], where x e2 is the horizontal coordinate of the second logistics transportation point after numbering, y e2 is the ordinate of the second logistics transport point after the numbering, z e2 is the vertical coordinate of the second logistics transportation point after numbering, and the second center point coordinate CE between the first logistics transportation point C1 and the first logistics transportation point E1 and the second logistics transportation point E2 after numbering is calculated. 12 , C.E. 12 =[x ce12 y ce12 z ce12 ], where x ce12 for The horizontal coordinate of the second midpoint, y ce12 is the ordinate of the second midpoint, z ce12 is the vertical coordinate of the second midpoint, Calculate the distance D between the second center point and the first drone landing pad 1-CE12 ; The third distance D between the second center point and the first drone landing pad 1-CE12 It is expressed as: Compare the third distance D 1-CE12 The distance between the predicted first UAV flight distance JL1, when D 1-CE12 ≥JL1, indicating that the second logistics transport point E2 after the numbering is not within the logistics distribution range of the first drone. 1-CE12 When <JL1, it indicates that the second logistics transportation point E2 after the numbering is within the logistics distribution range of the first drone; Repeat the above steps to determine all logistics transportation points within the entire drone logistics distribution range.

2. The multi-UAV logistics transportation collision prediction and warning method according to claim 1 is characterized in that: In S1, the evaluation indicators include battery capacity, motor efficiency, propeller diameter, fuselage mass, load, flight speed, flight mode, wind speed, wind direction, temperature, humidity, flight altitude and flight distance; The quantitative analysis and evaluation index system of UAV flight distance is specifically expressed as follows: The battery capacity element matrix corresponding to the battery capacity is RL, RL = [rl1 rl2 ... rl WN ], WN is the number of drones, rl WN is the battery capacity corresponding to the WNth drone; The motor efficiency element matrix corresponding to the motor efficiency is XL, XL = [xl1 xl2 ... xl WN ], xl WN is the motor efficiency corresponding to the WNth UAV; The propeller diameter element matrix corresponding to the propeller diameter is ZJ, ZJ = [zj1 zj2 ... zj WN ],zj WN is the propeller diameter corresponding to the WNth UAV; The fuselage mass element matrix corresponding to the fuselage mass is ZL, ZL = [zl1 zl2 ... zl WN ], zl WN is the fuselage mass corresponding to the WNth UAV; The load element matrix corresponding to the load is ZZ, ZZ = [zz1 zz2 ... zz WN ],zz WN is the load corresponding to the WNth UAV; The flight speed element matrix corresponding to the flight speed is SD, SD = [sd1 sd2 ... sd WN ], sd WN is the flight speed corresponding to the WNth UAV; The flight mode element matrix corresponding to the flight mode is MS, MS = [ms1 ms2...ms WN ],ms WN The flight mode corresponding to the WNth drone; The wind speed element matrix corresponding to the wind speed is FS, FS = [fs1 fs2...fs WN ], fs WN is the wind speed corresponding to the WNth UAV; The wind direction element matrix corresponding to the wind direction is FX, FX = [fx1 fx2...fx WN ], fx WN is the wind direction corresponding to the WNth UAV; The temperature element matrix corresponding to the temperature is QW, QW = [qw1 qw2...qw WN ], qw WN is the temperature corresponding to the WNth UAV; The humidity element matrix corresponding to humidity is PD, PD = [pd1 pd2...pd WN ], pd WN is the humidity corresponding to the WNth drone; The flight height element matrix corresponding to the flight height is GD, GD = [gd1 gd2 ... gd WN ], gd WN is the flight altitude corresponding to the WNth UAV; The flight distance element matrix corresponding to the flight distance is JL, JL = [jl1 jl2 ... jl WN ], jl WN is the flight distance corresponding to the WNth UAV.

3. The multi-UAV logistics transportation collision prediction and warning method according to claim 2 is characterized in that: In S2, the process of identifying abnormal battery capacity data is as follows: the average value of the battery capacity of WN drones is calculated. The standard deviation σ of the battery capacity of WN drones rl The relative distance D between any data and the mean value rl , compare the relative distance D between a single data point and the mean rl Relationship with 3, when D rl ≤3, the data rl for judging the battery capacity of the drone i It is normal data and is retained. rl >3, the data rl for judging the battery capacity of the drone i The above operation is performed on all data, and all abnormal data are deleted to obtain the updated battery capacity element matrix RL', RL' = [rl1'rl2'...rl WM '],rl WM 'The updated battery capacity data corresponding to the WNth drone; The average battery capacity of WN drones It is expressed as: The standard deviation σ of the battery capacity of WN drones rl It is expressed as: The relative distance D between any data and the mean value rl It is expressed as: According to the battery capacity abnormal data identification process, an updated motor efficiency element matrix XL', an updated propeller diameter element matrix ZJ', an updated fuselage mass element matrix ZL', an updated load element matrix ZZ', an updated flight speed element matrix SD', an updated flight mode element matrix MS', an updated wind speed element matrix FS', an updated wind direction element matrix FX', an updated temperature element matrix QW', an updated humidity element matrix PD', an updated flight altitude element matrix GD' and an updated flight distance element matrix JL' are established.

4. The multi-UAV logistics transportation collision prediction and warning method according to claim 3 is characterized in that: In S3, the UAV flight distance prediction model includes an input layer, 7 hidden layers, a self-attention layer, 2 residual layers and an output layer; The neural network structure of the established UAV flight distance prediction model is connected in the following order: Input layer: take the updated battery capacity element matrix RL', the updated motor efficiency element matrix XL', the updated propeller diameter element matrix ZJ', the updated fuselage mass element matrix ZL', the updated load element matrix ZZ', the updated flight speed element matrix SD', the updated flight mode element matrix MS', the updated wind speed element matrix FS', the updated wind direction element matrix FX', the updated temperature element matrix QW', the updated humidity element matrix PD', the updated flight altitude element matrix GD' and the updated flight distance element matrix JL' as input; Hidden layer 1: Number of neurons: 512, activation function: ReLU, batch normalization; Hidden layer 2: number of neurons: 256, activation function: ReLU, batch normalization; Self-attention layer: inputs the output matrix X of hidden layer 2; Initialize the weight matrix and build the initial weight query matrix W in turn Q , initial weight key matrix W K and the initial weight matrix W V , transform the output matrix X into the query matrix Q, key matrix K and value matrix V, Q = XW Q , K = XW K , V = XW V ; Calculate the similarity between the query matrix Q and the key matrix K to obtain the attention score matrix A. Among them, d k is the scaling factor; the softmax function is used to normalize the attention score matrix A so that the sum of the elements in each row is 1, and the normalized matrix B is obtained and output. Hidden layer 3: Number of neurons: 128, activation function: ReLU, batch normalization; Residual layer 1: The output of hidden layer 1 is connected to the output of hidden layer 3 through residual connection. Hidden layer 4: number of neurons: 64, activation function: ReLU, batch normalization; Hidden layer 5: number of neurons: 32, activation function: ReLU, batch normalization; Hidden layer 6: number of neurons: 16, activation function: ReLU, batch normalization; Residual layer 2: The output of hidden layer 4 is connected to the output of hidden layer 6 through residual connection; Hidden layer 7: Number of neurons: 8, activation function: ReLU, batch normalization; Output layer: Number of neurons: 1, outputs the predicted flight distance of the drone.

5. The multi-UAV logistics transportation collision prediction and warning method according to claim 4 is characterized in that: In S5, all logistics transportation points within the logistics distribution range of the first drone are numbered, and all logistics transportation points within the logistics distribution range of the first drone are recorded as G1, G2, ..., G hm , where hm is the number of all logistics transportation points within the logistics distribution range of the first drone, the first drone landing point T1 is defined as the starting point of the transportation route, and the remaining logistics transportation points G2-G hm Random sorting is performed, and the remaining logistics transportation points within the logistics distribution range of the first drone after random sorting, except for the first logistics transportation point, are recorded as H2-H hm The random sorting order is the logistics delivery order of the drone after it takes off from the first logistics transportation point G1 within the logistics delivery range of the first drone. The total flight distance DS of the drone under the random sorting order is calculated. 1-1 ; The total flight distance DS of the drone in the randomly sorted order 1-1 It is expressed as: in, It is expressed as the flight distance between the first drone landing point and the first logistics transportation point within the first drone logistics distribution range. It represents the flight distance between the first logistics transportation point within the logistics delivery range of the first drone and the second logistics transportation point within the logistics delivery range of the first drone after random sorting. Represents the second to last logistics transportation point H within the logistics delivery range of the first drone after random sorting hm-1 To the last logistics transport point H within the logistics delivery range of the first drone after random sorting hm The flight distance between All logistics transport points G1 within the logistics delivery range of the first drone are used as the first logistics delivery point of the drone, and the remaining logistics transport points G2-G hm Perform GM random sorting, and the order after all random sorting is different. Calculate the total flight distance DS of all randomly sorted drones 1-2 DS 1-3 ,…,DS 1-GM , where DS 1-GM The total flight distance of the drone after the drone takes off from the first logistics transportation point G1 within the logistics distribution range of the first drone, and flies according to the random sorting order obtained for the GMth time, and establishes the total flight distance matrix DST1 of the drone, DST1 = [DS 1-1 DS 1-2 DS 1-3 DS 1-GM ]; Screen and obtain the minimum element DS in the total flight distance matrix DST1 of the drone 1-min That is, the first logistics transportation point G1 within the logistics distribution range of the first drone is used as the minimum logistics transportation route distance corresponding to the first logistics delivery point of the drone, DS 1-min =minDST1; Repeat the above steps to use the remaining logistics transport points G2-G within the logistics delivery range of the first drone except the first logistics transport point hm As the first logistics delivery point of the drone, the corresponding minimum logistics transportation route distance DS is calculated 2-min DS 3-min ,…,DS hm-min , establish the minimum logistics transportation route distance matrix MDS1, MDS1 = [DS 1-min DS 2-min DS 3-min DS hm-min ], Screen and obtain the minimum element MDS in the minimum logistics transportation route distance matrix MDS1 1-min That is, the minimum flight distance of the first drone in the multi-point logistics transportation process, and its corresponding flight route is the shortest logistics transportation route; According to the total number of drones WN, repeat the above steps WN times to obtain the shortest logistics transportation route for all drones.

6. The multi-UAV logistics transportation collision prediction and warning method according to claim 5 is characterized in that: In S6, the total number of the same logistics transportation points of the shortest logistics transportation route of the first drone and the shortest logistics transportation route of the second drone is recorded as jm, and the same logistics transportation points of the shortest logistics transportation route of the first drone and the shortest logistics transportation route of the second drone are numbered 1, 2, ..., jm in sequence; According to the collected flight speed v1 of the first UAV and the flight speed v2 of the second UAV, the logistics transportation points with the same logistics transportation route of the first UAV and the second UAV are traversed in turn, and the time t for the first UAV to fly to the jth identical logistics transportation point is calculated. 1-ji and the time t when the second drone flies to the same logistics transportation point j 2-ji , where ji = 1, 2, ..., jm; The time t when the first drone flies to the jth same logistics transportation point 1-ji It is expressed as: The time t when the second drone flies to the same logistics transportation point 2-ji It is expressed as: Among them, DS 1-ji is the total flight distance of the first drone to the jth same logistics transportation point, DS 2-ji The total flight distance of the drones from the second drone to the jth same logistics transportation point; When |t 1-ji -t 2-ji |≤t tr When , it indicates that there is a risk of collision between the two drones, t tr is the safe time range, when |t 1-ji -t 2-ji |>t tr When , it indicates that there is no risk of collision between the two drones; When there is a risk of collision between two drones, the data of all logistics transportation points of the first drone are collected, and the transfer probability of the first drone flying from the current position to the collision direction is calculated at each logistics transportation point. That is, the transfer probability of the first UAV flying from the current position to the position with collision risk, and the probability P of the first UAV flying from the current position to the collision position is calculated. 1 ; The probability P of the first drone flying from its current position to the collision position 1 It is expressed as: Among them, km is the number of logistics transportation points between the initial position of the first drone and the location with collision risk; Collect data from all logistics transport points of the second drone, and calculate the transfer probability of the second drone flying from its current position to the collision direction at each logistics transport point That is, the transfer probability of the second UAV flying from the current position to the position with collision risk, and the probability P of the second UAV flying from the current position to the collision position is calculated. 2 ; The probability P of the second drone flying from its current position to the collision position 2 It is expressed as: Among them, lm is the number of logistics transportation points between the initial position of the second UAV and the position with collision risk; The risk probability P of the first UAV colliding with the second UAV is calculated. 1-2 , set the risk probability critical value P tr , compare the risk probability P of the first UAV colliding with the second UAV 1-2 and the risk probability critical value P tr The relationship between 1-2 ≥P tr When P 1-2 <P tr When , it indicates that the probability of collision between the two drones is low and no warning is needed; Repeat the above steps to obtain the risk probability of collision between any two drones, and integrate them to obtain the drone warning result.

Images