Urban unmanned aerial vehicle distribution path optimization method based on multi-target A* algorithm

Abstract

The invention discloses an urban unmanned aerial vehicle distribution path optimization method based on a multi-target A * algorithm, and relates to the field of unmanned aerial vehicle distribution route planning in the urban low-altitude logistics field, in particular to the urban unmanned aerial vehicle distribution path optimization method based on the multi-target A * algorithm. The invention aims to solve the problems that the path search efficiency of a traditional A * algorithm is low, redundant nodes are easy to generate, and multi-target unmanned aerial vehicle path planning in a three-dimensional space is not fully considered. The method comprises the following steps: step 1, constructing an environment model based on a three-dimensional space; 2, constructing an environment model and a multi-target A * algorithm based on a three-dimensional space, and constructing a path length mathematical model, a conflict risk mathematical model and a steering characteristic mathematical model; step 3, based on a multi-target A * algorithm and the constructed path length mathematical model, conflict risk mathematical model and steering characteristic mathematical model, obtaining a total score of path quality; and selecting the path with the high total score of the path quality as the unmanned aerial vehicle distribution path.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned aerial vehicle (UAV) delivery route planning in urban low-altitude logistics, and in particular to a method for optimizing the delivery route of an urban UAV based on a multi-objective A * algorithm. BACKGROUND

[0002] With the acceleration of urbanization, transportation is facing unprecedented challenges, and urban low-altitude logistics using UAVs as the delivery method is gradually gaining attention. Since route planning is a must-consider object in the UAV delivery process, it is particularly necessary to study the impact of UAV route planning on urban low-altitude logistics. In recent years, UAV technology has developed rapidly and is widely used in many fields. With the increasing complexity of UAV tasks and the expansion of application scope, path planning, as one of the core technologies of UAV autonomous flight, has gradually become a research hotspot. Therefore, how to design an efficient and safe flight path has become a major challenge in current technology research.

[0003] Currently, in the field of UAV path planning technology, the traditional A * algorithm is widely used, but in practical applications, it has low path search efficiency and is prone to generate redundant nodes and turning points. Therefore, domestic and foreign scholars are constantly improving and optimizing the A * algorithm from various angles, and have achieved remarkable results. However, there is less research on UAV path planning in three-dimensional space. Therefore, it is necessary to design and implement a UAV path planning system for three-dimensional environments, focusing on solving various prominent problems such as effective representation and perception of obstacles and risk areas in three-dimensional environments, and generating optimal paths that balance distance and safety in three-dimensional environments. SUMMARY

[0004] The purpose of the present application is to solve the problem of low path search efficiency and the generation of redundant nodes in the traditional A * algorithm, and to insufficiently consider multi-objective UAV path planning in three-dimensional space. Therefore, a method for optimizing the delivery route of an urban UAV based on a multi-objective A * algorithm is proposed.

[0005] A method for optimizing the delivery route of an urban UAV based on a multi-objective A * algorithm includes the following specific process:

[0006] Step 1: Construct an environment model based on a three-dimensional space;

[0007] Step 2: Construct an environment model based on a three-dimensional space and a multi-objective A * algorithm, and construct a path length mathematical model, a conflict risk mathematical model, and a turning characteristic mathematical model.

[0008] Step three, based on multi-objective A * The algorithm and the constructed path length mathematical model, conflict risk mathematical model and turning characteristic mathematical model obtain a total path quality score;

[0009] Select a path with a high total path quality score as the unmanned aerial vehicle distribution path.

[0010] The beneficial effects of the present application are:

[0011] 1. In the unmanned aerial vehicle distribution path planning method in the low-altitude logistics distribution, the A * algorithm is widely applicable and has traditional limitations, and a three-dimensional unmanned aerial vehicle path planning method based on a multi-objective A * algorithm is designed and implemented, which can plan an optimal path balancing path length and safety for the unmanned aerial vehicle in a complex three-dimensional environment.

[0012] 2. The multi-objective A * algorithm is crucial for unmanned aerial vehicle distribution path planning, which can model the three-dimensional environment using a grid method, introduce an exponential decay risk evaluation mechanism to realize obstacle perception, and overcome the problems of low path search efficiency and easy generation of redundant nodes of the traditional A * algorithm, effectively improving the calculation performance.

[0013] 3. The multi-objective A * algorithm realizes key functions such as weight optimization, path scoring and turning characteristic analysis, and is superior to the traditional A * algorithm in terms of safety, especially in high-risk environments, which can provide an intuitive three-dimensional visualization interface and real-time interaction function, and is more suitable for complex urban environments. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is the path planning research technical roadmap of the unmanned aerial vehicle distribution path based on the multi-objective A * algorithm described in the present application;

[0015] Figure 2 is the path graph obtained by setting the distance weight to 0.5 and the risk weight to 0.5 during path optimization of the multi-objective A * algorithm described in the present application;

[0016] Figure 3 is the path graph obtained by setting the distance weight to 0.8 and the risk weight to 0.2 during path optimization of the multi-objective A * algorithm described in the present application;

[0017] Figure 4 is the path graph obtained by setting the distance weight to 0.5 and the risk weight to 0.5 during path optimization of the multi-objective A *Path data graph planned by different distance weight in algorithm path optimization

[0018] Figure 5 The multi-objective A * Schematic diagram of complex scene in algorithm path optimization simulation environment

[0019] Figure 6 The multi-objective A * Schematic diagram of ordinary scene in algorithm path optimization simulation environment

[0020] Figure 7 The multi-objective A * Schematic diagram of simple scene in algorithm path optimization simulation environment

[0021] Figure 8 The multi-objective A * Algorithm unmanned aerial vehicle path planning graph

[0022] Figure 9 The multi-objective A * Algorithm unmanned aerial vehicle path planning graph DETAILED DESCRIPTION

[0023] Embodiment one: the embodiment is a kind of city unmanned aerial vehicle distribution path optimization method based on multi-objective A * Algorithm specific process is as follows:

[0024] Step one, based on three-dimensional space construction environment model;

[0025] Step two, based on three-dimensional space construction environment model and multi-objective A * Algorithm, construct path length mathematical model, conflict risk mathematical model, steering characteristic mathematical model;

[0026] Step three, based on multi-objective A * Algorithm and constructed path length mathematical model, conflict risk mathematical model, steering characteristic mathematical model, obtain path quality total score;

[0027] Select path quality total score high path as unmanned aerial vehicle distribution path.

[0028] Embodiment two: the embodiment is different from embodiment one, and the step one based on three-dimensional space construction environment model;

[0029] Specific process is as follows:

[0030] Step one, grid representation is carried out to three-dimensional space;Specific process is as follows:

[0031] The three-dimensional space environment is divided into N x N x N cubic cells, each of which is represented by coordinates (x, y, z) (coordinates (x, y, z) of the spatial rectangular coordinate system);

[0032] wherein the value of x is within the range of [0, N-1], the value of y is within the range of [0, N-1], and the value of z is within the range of [0, N-1];

[0033] This gridded modeling method allows the complex three-dimensional space to be discretized into a regular, easy-to-compute grid. By default, the system divides the space into 100 x 100 x 100 cells, i.e., 100 cells in each direction, so there are a total of 1,000,000 cells. Each cell represents a small area in three-dimensional space and is suitable for representing information such as obstacles, paths, and aircraft positions.

[0034] Step two, obstacles are mainly divided into two categories: buildings and no-fly zones, and the modeling method for each category of obstacles is different;

[0035] Buildings are modeled as cubes, and a building is defined by the center position (x, y, z) and the size (width, depth, height) of the cube. Specifically, the vertex range of the building is described by calculating the boundaries of the building in the x-axis, y-axis, and z-axis directions, and the width, depth, and height of the building affect the size and position of the building in three-dimensional space.

[0036] No-fly zones are modeled as spheres, and are determined by the center position (x, y, z) and the radius r. Any point within the no-fly zone satisfies the sphere geometry equation, i.e., the distance from the point to the center of the no-fly zone is less than or equal to the radius r;

[0037] First, the three-dimensional space is represented by a gridded method, dividing the space into N x N x N cubic cells, each of which represents a small area in three-dimensional space and is used to represent related information such as obstacles, paths, and aircraft positions. Then different methods are used to model different obstacles in the space environment, typically buildings are modeled as cubes and no-fly zones are modeled as spheres.

[0038] The other steps and parameters are the same as in the first embodiment.

[0039] Embodiment three: The difference between this embodiment and the first or second embodiment is that in step two, the environment model and multi-target A * algorithm, construct a path length mathematical model, a conflict risk mathematical model, and a turning characteristic mathematical model;

[0040] The specific process is as follows:

[0041] Step two one, construct a path length mathematical model; the specific process is:

[0042] The mathematical modeling of the path length of the unmanned aerial vehicle is mainly to optimize the total path length by establishing a path planning model combined with flight environment constraints and performance indicators. In constructing the model, the flight environment, such as the terrain and obstacles, needs to be clarified, followed by the path planning target, such as the principle of the shortest total path and obstacle avoidance priority. The three-dimensional space is divided into grids, each grid in the three-dimensional space represents a node, and the connection between the grids represents the feasible path. The path length is represented as the sum of the Euclidean distances between the nodes on the path:

[0043]

[0044] wherein,

[0045] length(S) represents the length of the path segment S;

[0046] x p+1 represents the coordinate of the p+1th node on the x-axis; x p represents the coordinate of the pth node on the x-axis;

[0047] y p+1 represents the coordinate of the p+1th node on the y-axis; y p represents the coordinate of the pth node on the y-axis;

[0048] z p+1 represents the coordinate of the p+1th node on the z-axis; z p represents the coordinate of the pth node on the z-axis;

[0049] p represents the pth node, p = 1, 2, …, n;

[0050] n represents the total number of nodes on the flight path; n is the minimum of 3 and the integer part of the length of the path segment divided by 0.5; n is determined by the length of the path length(S) and the step length;

[0051] Step two two, construct a conflict risk mathematical model, which includes building risk R b ( p ), no-fly zone risk R n(p) , point risk R(p), path segment risk R(S), continuity risk factor eff_risk(p), final path segment risk R final (S), and average path risk R avg (path);

[0052] Step two three, construct a turning characteristic mathematical model (the turning characteristic mathematical model is the turning radius).

[0053] Other steps and parameters are the same as embodiment one or two.

[0054] Embodiment four: the difference between this embodiment and one of the embodiments one to three is that the conflict risk mathematical model is constructed in step two, and the conflict risk mathematical model includes building risk R b(p) , no-fly zone risk R n(p) , point risk R(p), path segment risk R(S), continuity risk factor eff_risk(p), path segment final risk R final (S) and average risk of path R avg (path).

[0055] The specific process is as follows:

[0056] The construction of the unmanned aerial vehicle conflict risk mathematical model needs to comprehensively consider the aircraft dynamics, environmental perception and obstacle avoidance strategy and other factors, mainly by integrating environmental data and flight parameters, including the position, speed and acceleration and other state parameters of the unmanned aerial vehicle, combined with GPS error, crosswind disturbance and other uncertainty factors to evaluate various risks. The conflict risks considered in the present application mainly include seven kinds.

[0057] Step two one, calculate the building risk R b(p) ; expressed as:

[0058]

[0059] When dist(p, b) < 0.1, R b(p) = 1.0 (maximum risk);

[0060] dist(p, b) represents the distance from node p to the building;

[0061] The calculation formula of dist(p, b) is:

[0062]

[0063] B = (max{b x -p x , 0, p x -(b x + width)}) 2

[0064] C = (max{b y -p y , 0, p y -(b y + depth)}) 2

[0065] D = (max{b z -pz ,0,p z -(b z +height)}) 2

[0066] where,

[0067] (p x ,p y ,p z ) is the coordinate of node p; (b x ,b y ,b z ) is the coordinate of the building (usually the minimum coordinate point);

[0068] width represents the size of the building in the x-axis direction; depth represents the size of the building in the y-axis direction; height represents the size of the building in the z-axis direction;

[0069] B, C, D represent intermediate variables;

[0070] Step two two, calculate the risk R of the no-fly zone n(p) ; is expressed as:

[0071]

[0072] When dist(p, n) < 0.1, R n(p) = 1.0 (maximum risk);

[0073] dist(p, n) represents the distance from node p to the no-fly zone;

[0074] The calculation formula of dist(p, n) is:

[0075]

[0076] where,

[0077] (p x ,p y ,p z ) is the coordinate of node p; (n x ,n y ,n z ) is the coordinate of the no-fly zone (usually the minimum coordinate point);

[0078] R represents the radius of the no-fly zone;

[0079] Step two three, calculate the point risk R(p); is expressed as:

[0080] R(p) = max(R b(p) ,R n(p) ) (4)

[0081] wherein,

[0082] R b(p) is a building risk, representing the risk of the aircraft when approaching a building on the route; R n(p) is a no-fly zone risk, representing the risk of the aircraft when entering a no-fly zone; R(p) is a point risk, taking the maximum value of the building risk and the no-fly zone risk;

[0083] Step two, calculate the continuity risk factor eff_risk(p); represented as:

[0084] eff_risk(p) = min{1.0, R(S p )+ acc_risk(p-1) x 0.1} (5)

[0085] wherein,

[0086] R(S p ) represents the risk of node p of path segment S,

[0087] S p represents the node p of path segment S; p1 and p2 are the starting point and the ending point of the path segment, respectively, representing the two ends of the path; p is the index of the node, p = 1, 2, …, n;

[0088] 0.1 is a weighting coefficient, representing the influence of the cumulative risk of the previous point on the current point;

[0089] acc_risk(p-1) represents the cumulative risk of the (p-1)th node, represented as:

[0090] acc_risk(p) = acc_risk(p-1) x 0.5 + R(S p ) x 0.2

[0091] Initially, the cumulative risk acc_risk(0) is determined by the risk R(S1) of the first sampling point:

[0092] acc_risk(0) = R(S1) x 0.5

[0093] S1 represents the sampling point 1 of path segment S;

[0094] Step two, calculate the path segment risk R(S); represented as:

[0095]

[0096] Note: The risk calculation of path segment S is realized by averaging a plurality of sampling points;

[0097] wherein,

[0098] R(S) represents the risk of path segment S;

[0099] R(S p ) represents the risk value of path segment S; max-risk

[0100] n is minimum 3 and maximum is the length of path segment divided by 0.5 and rounded; n is determined by the length of path length(S) and step; length(S) represents the length of path segment S;

[0101] Step two twenty-six, calculate the final risk R final (S) of path segment;

[0102] The specific process is:

[0103] If the risk value R max-risk of path segment is greater than 0.9, the final risk R final (S) of path segment is represented as:

[0104] R final (S) = R(S) x 0.85 + R max-risk x 0.15 (7)

[0105] If the risk value R max-risk of path segment is less than or equal to 0.9, the final risk R final (S) of path segment is 0;

[0106] Step two twenty-seven, calculate the average risk R avg (path) of path; which is represented as:

[0107]

[0108] Wherein,

[0109] R(S p ) represents the risk of node p of path segment S;

[0110] length(S p ) represents the length of node p of path segment S;

[0111] The other steps and parameters are the same as one of the first to third embodiments.

[0112] The fifth embodiment is different from one of the first to fourth embodiments in that the step two three constructs a turning characteristic mathematical model (the turning characteristic mathematical model is turning radius);

[0113] The specific process is:

[0114] ​First, according to the two flight speeds of the unmanned aerial vehicle before and after turning, the included angle is calculated, then the radian value of turning is calculated according to the included angle, finally, according to the radian or the included angle, it is judged whether the unmanned aerial vehicle turns or not, if it turns, the turning radius is calculated according to the existing formula.

[0115] Step two three one, the included angle θ is calculated based on the two speed vectors v1 and v2 of the unmanned aerial vehicle, cosθ=v1·v2 / (|v1||v2|);

[0116] Wherein,

[0117] · represents the dot product (length) of the vector;

[0118] |v1| represents the modulus (length) of the speed vector v1; |v2| represents the modulus (length) of the speed vector v2;

[0119] |v1||v2| represents the product of the modulus |v1| of the vector v1 and the modulus |v2| of the vector v2;

[0120] Step two three two, the radian value of the included angle θ (the inverse cosine value of the cosine value) is calculated;

[0121] Step two three three, when θ≤0.087 radian (about 5 degrees), it is determined that the unmanned aerial vehicle does not turn, and the turning radius R=0;

[0122] When θ>0.087 radian (about 5 degrees), it is determined that the unmanned aerial vehicle turns, and the turning radius R is calculated; the process is as follows:

[0123] The turning radius of the unmanned aerial vehicle is a key parameter to ensure flight safety and task efficiency. The unmanned aerial vehicle needs to follow the principle of minimum turning radius when turning, and in complex terrain, reasonable planning of turning radius can shorten the distance and reduce energy consumption, therefore, the calculation of turning radius plays an important role in the flight of unmanned aerial vehicle; the calculation formula of turning radius is as follows:

[0124]

[0125] a, b and c are the lengths of the three sides of the triangle formed when the unmanned aerial vehicle turns;

[0126] Z area is the area of the triangle, Z area The calculation formula is

[0127] s is the semi-perimeter of the triangle,

[0128] Firstly, when constructing the path length mathematical model, the flight environment and path planning target need to be determined, and the flight environment constraints and performance indicators need to be considered; secondly, when constructing the conflict risk mathematical model, multiple factors affecting the flight of the unmanned aerial vehicle need to be considered, including the aircraft, the flight environment, and other uncertain factors, and the risks involved need to be considered as comprehensively as possible; finally, when establishing the turning characteristic mathematical model, the parameters for calculating the turning radius and the defining conditions for turning need to be determined, and then the calculation process can be solved.

[0129] The other steps and parameters are the same as one of the first to fourth embodiments.

[0130] The sixth embodiment is different from the first to fifth embodiments in that, in step three, the distance score distance_score is obtained based on the path length mathematical model constructed in step two. * The algorithm and the constructed path length mathematical model, conflict risk mathematical model, and turning characteristic mathematical model obtain the total path quality score; and a path with a high total path quality score is selected as the unmanned aerial vehicle delivery path.

[0131] The specific process is as follows:

[0132] Step three one, the distance score distance_score is obtained based on the path length mathematical model constructed in step two one;

[0133] Step three two, the average risk value avg_risk and the maximum risk value max_risk are obtained based on the conflict risk mathematical model constructed in step two two; and the risk score risk_score is obtained based on the average risk value avg_risk and the maximum risk value max_risk;

[0134] Step three three, the turning radius score radius_score is obtained based on the turning characteristic mathematical model constructed in step two three;

[0135] Step three four, the total path quality score is obtained based on the distance score distance_score, the risk score risk_score, and the turning radius score radius_score.

[0136] The other steps and parameters are the same as one of the first to fifth embodiments.

[0137] The seventh embodiment is different from the first to sixth embodiments in that, in step three one, the distance score distance_score is obtained based on the path length mathematical model constructed in step two one; and the specific process is as follows:

[0138] For applications that require saving flight time and energy, such as long-time UAV transportation or real-time tasks, distance score is a very critical indicator. It helps to optimize flight routes, reduce unnecessary detours, and thus improve flight efficiency and prolong flight time;

[0139] Distance score is used to evaluate the impact of path length on overall score, and the formula is as follows:

[0140]

[0141] Where length represents the total length of the flight path, which is calculated according to formula (1);

[0142] The distance score is inversely proportional, the longer the path, the lower the score. Adding a constant 10 can avoid division by zero error when the path length is zero.

[0143] The other steps and parameters are the same as one of the first six embodiments.

[0144] Embodiment eight: Different from one of the first seven embodiments, in step three two, based on the conflict risk mathematical model constructed in step two two, the average risk value avg_risk and the maximum risk value max_risk are obtained; based on the average risk value avg_risk and the maximum risk value max_risk, the risk score risk_score is obtained;

[0145] The specific process is:

[0146] And in the UAV task that requires high safety or flies in complex environment, the risk score is particularly important. By choosing a path with a lower risk score, the UAV can avoid dangerous areas, reduce the possibility of flight accidents, and ensure the safety and success rate of the task.

[0147] Step three two one, based on the conflict risk mathematical model constructed in step two two, the average risk value avg_risk and the maximum risk value max_risk are obtained;

[0148] The average risk value avg_risk is:

[0149]

[0150] The maximum risk value max_risk is the conflict risk mathematical model including building risk R b(p) , no-fly zone risk R n(p) , point risk R(p), path segment risk R(S), continuity risk factor eff_risk(p), path segment final risk R final (S), average risk of path R avgthe maximum value of the seven risks of formula (2), (3), (4), (5), (6), (7), (8);

[0151] The average risk value avg_risk is the average value of the seven risks of formula (2), (3), (4), (5), (6), (7), (8);

[0152] The maximum risk value max_risk is the maximum value of the seven risks of formula (2), (3), (4), (5), (6), (7), (8).

[0153] Step three, based on the average risk value avg_risk and the maximum risk value max_risk, obtain the risk score risk_score; the specific process is:

[0154] The calculation rule is as follows:

[0155] If avg_risk < 0.1: risk_score = 45 x (1-avg_risk 0.8 )

[0156] If 0.1≤avg_risk < 0.3: risk_score = 45 x (1-avg_risk 0.9 )

[0157] If 0.3≤avg_risk < 0.7: risk_score = 45 x (1-avg_risk)

[0158] If avg_risk≥0.7: risk_score = max_risk x (1-(avg_risk-0.7) x 0.3).

[0159] The other steps and parameters are the same as one of the first seven embodiments.

[0160] Specific embodiment nine: this embodiment is different from one of the first eight embodiments in that in step three, based on the steering characteristic mathematical model constructed in step two, obtain the steering radius score radius_score; the specific process is:

[0161] In tasks that require high flight stability and precise control (such as express delivery or shooting tasks), too many sharp turns can affect the smoothness of flight, increase flight time, and even cause the task to fail. Optimizing the steering score can help the UAV maintain a smooth flight path, ensure accurate landing, reduce energy consumption, and reduce the risk of the aircraft, ultimately improving task efficiency and safety. The steering radius score reflects the steering radius characteristics of the path, and the formula is as follows:

[0162] If the minimum steering radius R is 0: radius_score = 20

[0163] If the minimum turning radius R is greater than 0:

[0164] Wherein, R represents the minimum turning radius.

[0165] The turning radius score is used to reward larger turning radius, because larger turning is usually more gentle, and more suitable for safe flight.

[0166] The other steps and parameters are the same as one of embodiments one to eight.

[0167] Embodiment ten: different from one of embodiments one to nine, the step three four is based on the distance score distance_score, risk score risk_score, turning radius score radius_score, to obtain the total score of path quality;

[0168] total_score = 0.35 x distance_score + 0.45 x risk_score + 0.2 x radius_score

[0169] Wherein, total_score represents the total score of path quality.

[0170] The other steps and parameters are the same as one of embodiments one to nine.

[0171] The score is usually calculated according to the weighted average of the above several scores. The total score considers the distance, risk and turning radius of the path, and is a comprehensive evaluation index of the quality of the path.

[0172] Multi-objective A * The path planning method of the algorithm is as follows: first, the path generation strategy is proposed, the complex urban distribution environment is simulated accurately through multi-path planning, and the closed loop flight path is realized; then, the weight optimization mechanism is determined, the distance, risk and turning radius factors are considered comprehensively, and different weights are set.

[0173] The path generation strategy is proposed: the system can set the starting point and multiple intermediate nodes according to the actual demand, and can also randomly generate each node, which can accurately simulate the complex urban distribution environment. In the planning process, the system will use multi-objective A * The complete path from the starting point to each intermediate node is planned in turn, and then the complete path from the last node to the starting point is planned, so as to realize a closed loop flight path.

[0174] Determine the weight optimization mechanism: the system designs an automatic weight optimization mechanism, by iterating different weight combinations (from 0 to 1, step size 0.05) to calculate the path score, and select the highest score combination. The path score takes into account three factors, namely distance, risk and turning radius. According to existing research, risk is the most important consideration, with the highest weight, followed by distance, with a higher weight, and finally turning radius. Therefore, the usual score is the distance score (35 points), the risk score (45 points) and the turning radius score (20 points).

[0175] Embodiment:

[0176] In order to verify the effect of the present application, different numbers of buildings and no-fly zones are set to simulate real urban scenes, namely complex, ordinary and simple scenes, and the simulation test is completed according to the process, so that the unmanned aerial vehicle flies in three different environments, so that the experimental results are more realistic. Figure 1

[0177] 1. Construct an environment model based on three-dimensional space:

[0178] 1) Three-dimensional space representation based on gridding method:

[0179] The three-dimensional environment is modeled using the gridding method, which divides the space into N x N x N cubic cells, each cell is represented by coordinates (x, y, z), where x values are in the range [0, N-1], y values are in the range [0, N-1], and z values are in the range [0, N-1].

[0180] 2) According to the iteration of different distance weights used for planning the path each time, the highest evaluation path is found, and the optimal weight path is obtained. In addition, referring to Figure 4 , Figure 2 , Figure 3 , it can be seen that the paths planned under different weights are completely different, such as Figure 3 , the risk weight is only 0.2, so the planned path is closer to the area with more buildings, Figure 2 the risk weight 0.5 will avoid the area with obstacles.

[0181] 3) Propose different types of obstacle modeling methods:

[0182] Referring to Figure 5 , Figure 6 , Figure 7 ​Obstacles are mainly divided into two categories: buildings and no-fly zones, and the modeling methods of each category are different. Buildings are modeled as cubes, defined by their center position (x, y, z) and size (width, depth, height). Specifically, the range of the top of the building can be described by calculating its boundaries in the x-axis, y-axis, and z-axis directions, where the width, depth, and height affect the size and position of the building in three-dimensional space. No-fly zones are modeled as spheres, defined by their center position (x, y, z) and radius r. Any point within the no-fly zone satisfies the sphere geometry equation, i.e., the distance from the point to the center of the no-fly zone is less than or equal to the radius r.

[0183] 2. Multi-objective A * Path optimization modeling of the algorithm:

[0184] 1) Construct a mathematical model of path length:

[0185] After fully considering the flight environment of the UAV in complex, ordinary, and simple scenarios, and clearly defining the path planning goal, while considering the shortest total path, the principle of giving priority to obstacle avoidance is followed. According to formula (1), the flight path length can be obtained.

[0186] 2) Construct a conflict risk mathematical model:

[0187] After understanding the flight environment such as buildings and no-fly zones in different scenarios, the maximum risk value of the UAV flight can be calculated according to formula (2); the risk of the UAV touching the building during flight can be calculated according to formula (3); the risk of the UAV flying into the no-fly zone can be calculated according to formula (4); the continuity risk factor value can be calculated through formula (5); the path segment risk value of the UAV flight can be calculated according to formula (6); if the maximum risk exceeds 0.9, the path segment risk can be adjusted through formula (7) and the final path segment risk value can be obtained; in addition, the average risk of the path segment can be calculated through formula (8); by estimating and calculating various risks, the flight strategy can be further optimized and potential risks can be reduced.

[0188] 3) Construct a turning characteristic mathematical model:

[0189] The turning angle of the UAV during flight is determined by v1 and v2 to determine whether the UAV flight at this stage is turning. If it is turning, the turning radius also needs to be calculated, which can be calculated in turn by referring to formula (9).

[0190] 3. Multi-objective A * Evaluation of the path planning system of the algorithm:

[0191] 1) Distance evaluation:

[0192] The path score can be calculated from the total length of the flight path according to equation (10). The distance score is inversely proportional to the length of the path. A constant of 10 is added to avoid division by zero errors when the path length is zero.

[0193] 2) Risk evaluation:

[0194] The risk score is calculated from the average risk value avg_risk and the maximum risk value max_risk of the path. The average risk value avg_risk is raised to the power of 2 in the equation. By using different exponents, different sensitivity adjustments can be made to the risk score of the path. The calculation rules are as follows:

[0195] If avg_risk < 0.1: risk_score = 45 x (1 - avg_risk 0.8 )

[0196] If 0.1 ≤ avg_risk < 0.3: risk_score = 45 x (1 - avg_risk 0.9 )

[0197] If 0.3 ≤ avg_risk < 0.7: risk_score = 45 x (1 - avg_risk)

[0198] If avg_risk ≥ 0.7: risk_score = max_risk x (1 - (avg_risk - 0.7) x 0.3)

[0199] The risk score under different conditions can be calculated respectively.

[0200] 3) Turn radius evaluation:

[0201] The turn radius score reflects the turn radius characteristics of the path. If the minimum turn radius is 0, the turn radius score is 20. If the minimum turn radius is greater than 0, the minimum turn radius score is calculated by equation The turn radius score is used to reward larger turn radii, as larger turns are generally more gradual and more suitable for safe flight.

[0202] 4) Total path quality score based on A * algorithm:

[0203] Referring to Figure 4The score is usually calculated according to the weighted average of the above several scores. The total score takes into account the distance, risk and turning radius of the path, etc. It is an index for comprehensive evaluation of path quality. The total score of the quality of the UAV flight path is obtained from the formula total_score = 0.35 x distance_score + 0.45 x risk_score + 0.2 x radius_score.

[0204] 4. Experimental results and conclusions:

[0205] Referring to Figure 8 , Figure 9 , it can be seen that the multi-objective A * algorithm (optimal weight) is significantly better than the traditional A * algorithm in overall performance, especially in safety, path optimization efficiency and comprehensive performance in simple and ordinary scenes. However, in complex scenes, due to the presence of many buildings and no-fly zones, the optimal weight directly gives up the influence of risk and prioritizes reducing distance, which cannot take advantage of the optimization of multi-objective A * , so the planned path is roughly the same as that of the traditional A * algorithm and it is difficult for the traditional A * algorithm to make a difference. The detailed data is shown in Table 1:

[0206] Table 1: Comparison of paths planned by multi-objective A * and traditional A *

[0207]

[0208]

[0209] Therefore, in ordinary and simple scenes, the multi-objective A * algorithm reduces the average risk by 11.60% and 30.15% respectively, significantly improves the average risk, and greatly reduces the potential danger. The maximum risk is reduced by 9.78% and 28.57% respectively, which means that in extreme cases, it can also provide a safer path. Although the maximum risk is the same in complex scenes, the average risk is still slightly better (1.30%), indicating that the overall risk distribution is more balanced. The multi-objective A * reduces the number of nodes in the ordinary scene by 12.50% and in the simple scene by 17.14%, which means that the multi-objective A * algorithm has less calculation when searching for a path, and can converge to the optimal solution faster, which is suitable for systems with high real-time requirements. The multi-objective A * ​The path length is slightly longer (2.59%-4.24%), but this disadvantage is offset by the significant improvement in safety, stability, and overall score. In most practical applications (such as unmanned vehicles, robot navigation), safety and path quality are more important than absolute path length.

[0210] Therefore, in most practical applications, multi-objective A * algorithm is a better choice, especially for scenarios with high safety and path quality requirements. In the future, the weight strategy can be further optimized to achieve a better balance between path length and risk.

[0211] The present application can also have other various embodiments, and those skilled in the art can make various corresponding changes and modifications according to the present application without departing from the spirit and essence of the present application, but these corresponding changes and modifications should all belong to the protection scope of the claims attached to the present application.

Claims

1. A method based on multi-objective A * The algorithm-based method for optimizing urban drone delivery routes is characterized by: The specific process of the method is as follows: Step 1: Construct an environment model based on three-dimensional space; Step 2: Construct an environment model and multi-objective A based on three-dimensional space. * Algorithms are used to construct mathematical models for path length, conflict risk, and turning characteristics. Step 3: Based on multi-objective A * The algorithm and the constructed mathematical models of path length, conflict risk, and turning characteristics are used to obtain the overall path quality score. The routes with the highest overall route quality score are selected as drone delivery routes.

2. A method based on multi-objective A according to claim 1 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In step one, an environment model is constructed based on a three-dimensional spatial structure. The specific process is as follows: Step 11: Represent the three-dimensional space using a mesh; the specific process is as follows: The three-dimensional space environment is divided into N×N×N cube cells, and each cube cell is represented by coordinates (x, y, z); The values ​​of x, y, and z are all within the range of [0, N-1]. Steps 1 and 2: Obstacles are mainly divided into two categories: buildings and no-fly zones; The building is modeled as a cube, defined by the cube's center position (x, y, z) and dimensions (width, depth, height); The no-fly zone is modeled as a sphere, determined by the center position (x, y, z) and radius r of the sphere.

3. A method based on multi-objective A according to claim 2 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In step two, an environment model and multi-objective A are constructed based on three-dimensional space. * Algorithms are used to construct mathematical models for path length, conflict risk, and turning characteristics. The specific process is as follows: Step 2.1: Construct a mathematical model for the path length; the specific process is as follows: The three-dimensional space is divided into a grid, where each grid represents a node, and the lines connecting the grids represent feasible paths. The path length is expressed as the sum of the Euclidean distances between the nodes on the path. in, length(S) represents the length of path segment S; x p+1 This represents the coordinate of the (p+1)th node on the x-axis; x p This represents the coordinates of the p-th node on the x-axis; y p+1 This represents the coordinate of the (p+1)th node on the y-axis; y p This represents the y-coordinate of the p-th node; z p+1 This represents the coordinate of the (p+1)th node on the z-axis; z p This represents the coordinates of the p-th node on the z-axis; p represents the p-th node, where p = 1, 2, ..., n; n represents the total number of nodes on the flight path S; length(S) represents the length of path segment S; Step 22: Construct a mathematical model of conflict risk. The mathematical model of conflict risk includes the building risk R. b(p) No-fly zone risk R n(p) Point risk R(p), path segment risk R(S), continuity risk factor eff_risk(p), and path segment final risk R final (S) and the average risk R of the path avg (path); Steps 2 and 3: Construct a mathematical model of steering characteristics.

4. A method based on multi-objective A according to claim 3 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In step two, a conflict risk mathematical model is constructed. The conflict risk mathematical model includes building risk R. b(p) No-fly zone risk R n(p) Point risk R(p), path segment risk R(S), continuity risk factor eff_risk(p), and path segment final risk R final (S) and the average risk R of the path avg (path); The specific process is as follows: Step 221: Calculate the building risk R b(p) ; indicates as: When dist(p,b) < 0.1, R b(p) =1.0; dist(p,b) represents the distance from node p to the building; The formula for calculating dist(p,b) is: B=(max{b x -p x ,0,p x -(b x +width)}) 2 C=(max{b y -p y ,0,p y -(b y +depth)}) 2 D=(max{b z -p z ,0,p z -(b z +height)}) 2 in, (p x ,p y ,p z (b) are the coordinates of node p; x ,b y ,b z () represents the coordinates of the building; Width represents the building's dimension along the x-axis; depth represents the building's dimension along the y-axis; height represents the building's dimension along the z-axis. B, C, and D represent intermediate variables; Step 222: Calculate the risk R of the no-fly zone n(p) ; indicates as: When dist(p,n)<0.1, R n(p) =1.0; dist(p,n) represents the distance from node p to the no-fly zone; The formula for calculating dist(p,n) is: in, (p x ,p y ,p z (n) are the coordinates of node p; x ,n y ,n z () are the coordinates of the no-fly zone; R represents the radius of the no-fly zone; Step 223: Calculate the point risk R(p); expressed as: R(p)=max(R b(p) ,R n(p) ) (4) in, R b(p) It is a building risk; R n(p) R(p) represents no-fly zone risk; R(p) represents point risk. Step 224: Calculate the continuous risk factor eff_risk(p); expressed as: eff_risk(p)=min{1.0,R(S p )+acc_risk(p-1)×0.1} (5) in, R(S p ) represents the risk of node p in path segment S. S p Let p represent the node of path segment S; p1 and p2 are the start and end points of the path segment, respectively; p is the index of the node, p = 1, 2, ..., n; 0.1 is a weighting coefficient; acc_risk(p-1) represents the cumulative risk of the (p-1)th node, expressed as: acc_risk(p)=acc_risk(p-1)×0.5+R(S p )×0.2 Initially, the cumulative risk acc_risk(0) is determined by the risk R(S1) of the first sampling point: acc_risk(0) = R(S1) × 0.5 S1 represents sampling point 1 of path segment S; Step 225: Calculate the path segment risk R(S); expressed as: in, R(S) represents the risk of path segment S; R(S p The maximum value of ) is denoted as the risk value R in the path segment. max-risk ; Step 226: Calculate the final risk R of the path segment. final (S); The specific process is as follows: If the risk value R in the path segment max-risk A value greater than 0.9 indicates a final risk R for the path segment. final (S) is represented as: R final (S)=R(S)×0.85+R max-risk ×0.15 (7) If the risk value R in the path segment max-risk Less than or equal to 0.9, the final risk R of the path segment final (S) = 0; Step 227: Calculate the average risk R of the path. avg 9path); represented as: in, R(S p ) represents the risk of node p in path segment S; length(S p ) represents the length of node p in path segment S.

5. A method based on multi-objective A according to claim 4 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In steps two and three, a mathematical model of steering characteristics is constructed. The specific process is as follows: Step 231: Calculate the included angle θ based on the two velocity vectors v1 and v2 of the UAV, cosθ=v1·v2 / (|v1||v2|); in, • Represents the dot product of vectors; |v1| represents the magnitude of velocity vector v1; |v2| represents the magnitude of velocity vector v2; |v1||v2| represents the product of the modulus of vector v1|v1 and the modulus of vector v2|v2; Step 2: Calculate the radian value of the included angle θ; Step 233: When θ≤0.087 radians, it is determined that the drone has not turned, and the turning radius R=0; When θ > 0.087 radians, the drone is determined to be turning, and the turning radius R is calculated; the process is as follows: a, b, and c are the side lengths of the triangle formed when the drone turns; Z area Z is the area of ​​the triangle. area The calculation formula is s is the semi-perimeter of the triangle.

6. A method based on multi-objective A according to claim 5 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: Step three is based on multi-objective A * The algorithm and the constructed mathematical models of path length, conflict risk, and turning characteristics are used to obtain the overall path quality score; the path with the highest overall path quality score is selected as the drone delivery path. The specific process is as follows: Step 31: Obtain the distance score (distance_score) based on the path length mathematical model constructed in Step 21; Step 32: Based on the conflict risk mathematical model constructed in Step 22, obtain the average risk value avg_risk and the maximum risk value max_risk; based on the average risk value avg_risk and the maximum risk value max_risk, obtain the risk score risk_score. Step 33: Based on the mathematical model of steering characteristics constructed in Step 23, obtain the steering radius score (radius_score). Steps 3 and 4: Based on the distance score (distance_score), risk score (risk_score), and turning radius score (radius_score), obtain the overall path quality score.

7. A method based on multi-objective A according to claim 6 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In step three-one, the distance score (distance_score) is obtained based on the path length mathematical model constructed in step two-one; the specific process is as follows: The formula is as follows: Where length represents the total length of the flight path.

8. A method based on multi-objective A according to claim 7 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In step three two, based on the conflict risk mathematical model constructed in step two two, the average risk value avg_risk and the maximum risk value max_risk are obtained; based on the average risk value avg_risk and the maximum risk value max_risk, the risk score risk_score is obtained. The specific process is as follows: Step 321: Based on the conflict risk mathematical model constructed in Step 22, obtain the average risk value avg_risk and the maximum risk value max_risk; The average risk value avg_risk is: The maximum risk value, max_risk, is the mathematical model for conflict risk, including building risk R. b(p) No-fly zone risk R n(p) Point risk R(p), path segment risk R(S), continuity risk factor eff_risk(p), and path segment final risk R final (S), average risk of the path R avg The maximum value in (path); Step 322: Based on the average risk value avg_risk and the maximum risk value max_risk, obtain the risk score risk_score; the specific process is as follows: The calculation rules are as follows: If avg_risk<0.1: risk_score=45×(1-avg_risk 0.8 ) If 0.1≤avg_risk<0.3: risk_score=45×(1-avg_risk 0.9 ) If 0.3≤avg_risk<0.7: risk_score=45×(1-avg_risk) If avg_risk≥0.7: risk_score=max_risk×(1-(avg_risk-0.7)×0.3).

9. A method based on multi-objective A according to claim 8 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In step three, based on the mathematical model of steering characteristics constructed in steps two and three, the steering radius score (radius_score) is obtained; the specific process is as follows: The formula is as follows: If the minimum turning radius R is 0: radius_score = 20 If the minimum turning radius R is greater than 0: Where R represents the minimum turning radius.

10. A method based on multi-objective A according to claim 9 * The algorithm-based method for optimizing urban drone delivery routes is characterized by: In steps three and four, the total path quality score is obtained based on the distance score, risk score, and turning radius score. total_score = 0.35 × distance_score + 0.45 × risk_score + 0.2 × radius_score, where total_score represents the total path quality score.

Images