An unmanned aerial vehicle cluster cooperative path planning method for urban logistics distribution

Abstract

The application discloses a kind of unmanned aerial vehicle cluster cooperative path planning methods for city logistics distribution, comprising: loading environment 3D model;According to the environment 3D model and task information, initialize path planning model;Path planning model is the problem model with the minimum comprehensive cost as the solving target when unmanned aerial vehicle cluster cooperative path is solved;Comprehensive cost comprehensively height cost, angle cost, avoid obstacle cost, collision cost, communication cost and path length cost;By improving wolf swarm algorithm, the path planning model is solved, and the optimal unmanned aerial vehicle cluster cooperative path is obtained;Improved search strategy is based on population vitality value and problem dimension selection search strategy to search.The application can realize accurate and efficient logistics distribution for dense building city.

Description

Technical Field

[0001] This invention belongs to the field of drone swarm path planning technology, specifically relating to a drone swarm collaborative path planning method for urban logistics delivery. Background Technology

[0002] Traditional ground logistics delivery models face problems such as traffic congestion, low delivery efficiency, and high costs. With the development of related technologies, drone delivery, with its advantages of flexibility, efficiency, and lack of ground traffic restrictions, has become an important research direction for the intelligent upgrading of urban logistics. Due to the widespread application of drones, using drone swarms for logistics delivery in urban environments is gradually becoming a reality. Path planning aims to calculate the optimal flight route for each drone to avoid collisions, optimize resource utilization, and ensure efficient and safe execution of logistics delivery. Accurate path planning plays a crucial role in improving the mission execution efficiency of drone swarms, reducing energy consumption, and ensuring flight safety.

[0003] Logistics delivery in urban environments typically faces more complex challenges, primarily manifested in two aspects: firstly, spatial constraints and high building density; and secondly, dynamic obstacles and airspace conflicts. Traditional single-drone or suburban path planning methods are ineffective in solving the path planning problem for drone swarms in urban environments. Furthermore, although significant progress has been made in existing research on drone swarm path planning, path planning for urban environments remains significantly limited due to the high complexity of the environment.

[0004] Specifically, current methods for UAV path planning include sample-based methods, graph-oriented algorithms, reinforcement learning-based methods, and swarm intelligence algorithms. Sample-based methods explore the environment through random sampling to construct feasible paths, including Probabilistic Route Mapping (PRM) algorithms and Rapid Random Tree Exploration (RRT) algorithms. While they do not require prior knowledge of the entire environment, they typically only find feasible solutions, not optimal ones, leading to imprecise path planning. Graph-oriented algorithms abstract the environment into a graph structure and find the optimal path by searching the graph. Examples include A* algorithms, Dijkstra's algorithm, Voronoi graph (VD) algorithms, and visibility graph (VG) algorithms. Although these algorithms can obtain optimal solutions, the number of nodes and edges in the graph increases exponentially with increasing dimensionality, making computation extremely difficult and failing to meet the high-efficiency requirements of practical applications. Reinforcement learning-based methods rely on extensive simulation training, and their policy transferability is severely limited by the uncertainties of the urban physics field. While swarm intelligence algorithms (such as PSO) have the advantage of parallel search, their premature convergence and lack of dynamic adaptability lead to a sharp decline in solution quality in high-dimensional time-varying urban environments, making it difficult to ensure the accuracy of path planning.

[0005] Therefore, there is an urgent need for a drone swarm collaborative path planning method to achieve accurate and efficient logistics delivery in densely populated urban areas. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this invention provides a method for collaborative path planning of drone swarms for urban logistics delivery.

[0007] The technical problem to be solved by this invention is achieved through the following technical solution:

[0008] A method for collaborative path planning of drone swarms for urban logistics delivery includes:

[0009] Load a 3D environmental model containing obstacle information and communication interference fields;

[0010] The path planning model is initialized based on the 3D environment model and task information; wherein, the task information includes at least the start and end point positions of each UAV in the UAV swarm; the path planning model is a problem model with the goal of minimizing the overall cost of the UAV swarm's cooperative path; the overall cost includes altitude cost, angle cost, obstacle avoidance cost, collision cost, communication cost, and path length cost;

[0011] The path planning model is solved by improving the wolf pack algorithm to obtain the optimal UAV swarm collaborative path; the improved wolf pack algorithm improves at least the search strategy of the wolf pack algorithm, and the improved search strategy selects a search strategy from multiple search strategies based on the population vitality value and the problem dimension.

[0012] This invention provides a collaborative path planning method for UAV swarms in urban logistics delivery. It calculates the flight path of a UAV swarm under multiple constraints while collaboratively performing logistics delivery tasks in densely built-up urban environments. Specifically, this method addresses the collaborative path planning scenario of UAV swarms in densely built-up 3D urban environments. It constructs a path planning model that integrates altitude cost, angle cost, obstacle avoidance cost, collision cost, communication cost, and path length cost. Combined with an improved wolf pack algorithm, it solves problems related to maintaining communication links and avoiding collisions among UAV swarms. The proposed algorithm framework is novel, with fast convergence and high efficiency. Especially for UAV swarm path planning in densely built-up urban environments, it can quickly solve the overall path planning scheme for the UAV swarm, improving task execution efficiency, reducing energy consumption, and ensuring flight safety.

[0013] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0014] Figure 1 This is a flowchart of a drone swarm collaborative path planning method for urban logistics delivery provided by the present invention;

[0015] Figure 2 The optimal values ​​and running times of the four search strategies involved in this invention when solving the Sphere function in 2D and 50D space are shown.

[0016] Figure 3 This is the dense building city model set up when conducting simulation experiments in this invention;

[0017] Figure 4 The results show the comparison of the costs of each UAV obtained by using the method of this invention and five existing algorithms for UAV swarm path planning in Experiment 1.

[0018] Figure 5 The path results obtained in Experiment 1 using the method of this invention for UAV swarm path planning;

[0019] Figure 6 The comparison results of the flight altitudes of each UAV obtained by using the method of this invention and five existing algorithms for UAV swarm path planning in Experiment 1;

[0020] Figure 7 The results show the comparison of the costs of each UAV obtained by using the method of this invention and five existing algorithms for UAV swarm path planning in Experiment 2.

[0021] Figure 8 The path results obtained in Experiment 2 using the method of this invention for UAV swarm path planning;

[0022] Figure 9 The results show the comparison of flight altitudes of each UAV obtained from UAV swarm path planning using the method of this invention and five existing algorithms in Experiment 2. Detailed Implementation

[0023] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0024] To achieve accurate and efficient logistics delivery in densely populated urban areas, this invention provides a method for collaborative path planning of drone swarms for urban logistics delivery. (See also...) Figure 1 As shown, the method includes the following steps:

[0025] S10. Load the 3D environmental model containing obstacle information and communication interference field.

[0026] Here, the 3D environmental model is pre-built. Specifically, high-precision geometric data of the urban environment, including key data such as building height, outline, and roof structure, is acquired through satellite remote sensing maps or architectural CAD drawings, enabling environmental data collection and geometric parameter extraction. Then, based on the geometric parameters of the densely built-up urban environment, a continuous terrain elevation grid is generated by interpolating ground point cloud data, constructing a Digital Elevation Model (DEM). It is important to distinguish between "ground DEM" and "building DEM": the ground DEM represents natural terrain undulations, while the building DEM abstracts buildings as vertical cuboids. By integrating the DEM data and marking non-flying areas (such as lakes and no-fly zones), buildings are converted into 3D obstacles (simplified as vertical cuboids with height). Communication interference in the urban environment is abstracted as a communication interference field centered on buildings, gradually attenuating with increasing distance. This yields an environmental 3D model containing obstacle information and the communication interference field, which is saved for later loading.

[0027] S20. Initialize the path planning model based on the 3D environmental model and task information; wherein, the task information includes at least the start and end point positions of each UAV in the UAV swarm; the path planning model is a problem model with the goal of minimizing the collaborative path of the UAV swarm; the comprehensive cost includes altitude cost, angle cost, obstacle avoidance cost, collision cost, communication cost and path length cost.

[0028] Here, the initialization parameters include setting the drone swarm's origin and destination points, as well as the communication interference field parameters. Regarding the drone swarm's origin and destination point setting: it is assumed that in a 3D urban environment, each drone has a corresponding origin and destination point, represented by 3D spatial coordinates. Regarding the communication interference field parameters: the electromagnetic interference present in the city is abstracted as a bounded cylinder, determined by the coordinates of its center point, radius, and height.

[0029] In addition, the wolf pack algorithm will be used to solve the initialized path planning model. Therefore, the parameters related to the wolf pack algorithm can also be initialized in step S20, including the population size, maximum number of iterations, maximum number of walks, step size factor, distance decision factor, and update ratio factor.

[0030] In this path planning model, each UAV path is discretized into... The nth path point (including the start and end points) will be the nth path point. The first drone The coordinates of the path points are as follows When initializing the path planning model based on the 3D environment model and task information, task information is imported into the path planning model. This task information includes: The starting coordinates of the flight path of a drone and the The endpoint coordinates of the flight path of a drone It can also include the altitude requirements of the missions performed by the drone. This involves incorporating known parameters from the 3D environmental model (such as average building height). Optimal flight altitude range Describes the geometric parameters of obstacles and the maximum effective range of the communication interference field. (etc.) are substituted into the path planning model for initialization.

[0031] Among them, the optimal altitude flight zone The altitude can be determined based on the mission requirements of the drone. A hierarchical planning strategy can be employed here, using quantified hierarchical thresholds to discretize the continuous three-dimensional space into a low-altitude region that flies close to the ground but requires frequent obstacle avoidance, and a high-altitude region with straighter paths but is more susceptible to airflow. The formula for calculating the hierarchical threshold is:

[0032] ;

[0033] In this formula, It is the average height of buildings; It is the drone's maximum climb rate, determined by the drone's average speed and maximum elevation angle between buildings; This is a safety margin set to cope with sudden airflow or positioning errors. After obtaining the tiered thresholds, the flight altitude requirements of different tasks are abstracted into an optimal flight altitude range by combining them with actual mission requirements. :

[0034] ,

[0035] ;

[0036] in, and The flight altitude is determined by the requirements of the mission, and the two are not necessarily equal. The value is entirely determined by the mission requirements, and this invention does not limit it.

[0037] In this invention, the path planning model mainly includes its objective function, which is used to calculate the comprehensive cost. This comprehensive cost integrates the cost of height, angle, obstacle avoidance, collision, communication, and path length. The models for each cost are described below.

[0038] (1) Path length cost

[0039] The total path length is the sum of the Euclidean distances between all adjacent path points; minimizing this length reduces flight time and energy consumption. In this invention, the path length cost can be calculated using the following formula:

[0040] ;

[0041] Among them, the The first drone's flight path The coordinates of the path points are , No. The starting coordinates of the flight path of each drone are , No. The coordinates of the endpoint of the flight path of the drone are , The number of waypoints on the flight path. Indicates the first The path length cost of a single drone.

[0042] It is understandable that in the model expression for the path length cost, the denominator calculates the first... The actual path length of the first UAV is the sum of the Euclidean distances between all adjacent path points; the numerator is the length of the first UAV. The straight-line distance from the starting point to the destination of each drone.

[0043] (2) High cost

[0044] Drones often need to constantly adjust their altitude during flight, resulting in greater energy consumption. Therefore, we can abstract the drone's flight space into an optimal altitude flight range. The calculation method for altitude cost differs at different altitude levels. The specific calculation method is as follows:

[0045] ;

[0046] in, Indicates the first The drone in its flight path Flight altitude at each waypoint It is a preset obstacle influence factor, with a value range of [0,1]. It is a preset environmental airflow influence factor, with a value range of [0,1]. Indicates the first The drone in its flight path The elevation cost corresponding to each path point.

[0047] (3) Angle cost

[0048] Drones also incur energy costs when maneuvering to adjust their flight angles; the smoother the path, the better its stability. Therefore, this invention measures path stability by the magnitude of the turning and climb angles and the frequency of rotation, thereby defining an angle cost, which is calculated as follows:

[0049] ;

[0050] in, Indicates the first The drone in its flight path The angle cost corresponding to each path point. This is the maximum turning angle of the drone. This is the maximum climb angle for the drone. For the first The turning angle of a drone For the first The climb angle of a drone, and The following formula is used to calculate:

[0051] ,

[0052] ;

[0053] Among them, the The first drone's flight path The coordinates of the path points are .

[0054] (4) Cost of obstacle avoidance

[0055] Given the characteristics of densely built-up urban environments, drones need to frequently avoid building obstacles during flight, making this a crucial condition for path planning. Therefore, this invention treats each waypoint as a risk assessment point, and each path is assessed... - There are 2 path points (excluding the start and end points). The cost of avoiding obstacles can be calculated using the following formula:

[0056] ;

[0057] Among them, the The first drone's flight path The coordinates of the path points are , It is a set of geometric parameters describing the cuboid space in which the obstacle is located, including parameters describing the extent of the horizontal plane covered by the cuboid space. and parameters describing the height range covered by the cuboid space containing the obstacle. . It is the first The first drone's flight path The Euclidean distance of each path point from the center of the obstacle. It is the farthest effective range of the communication interference field. It is a preset communication impact factor with a value of 0.5, but it is not limited to this value. Indicates the first The drone in its flight path The cost of avoiding obstacles at each path point.

[0058] Understandably, in the model expression for obstacle avoidance cost, if there is a possibility of collision on a path, the cost of the path is set to infinity in order to avoid choosing that path.

[0059] (5) Collision cost

[0060] It is understandable that drone swarms cannot collide with each other while performing tasks; this is a necessary constraint. To ensure that the final path planning result meets this constraint, this invention will... The path point and the first The path segments between each path point are also evenly divided into... The distance assessment point is required to be the nth distance assessment point. The drone and the first The drone (different from the first) The drone of the drone) The distance between each distance assessment point must be greater than the safe distance; therefore, the collision cost is defined as follows:

[0061] ;

[0062] in, For the first The drone and the first The first drone on the same path segment The distance to each distance assessment point It refers to the number of drones in the cluster. The radius of the drone. This refers to the minimum safe flight distance between drones to avoid collisions. Indicates the first The drone in the first The collision cost corresponding to each distance evaluation point.

[0063] It is understandable that in the model expression for collision cost, the collision cost of paths that do not satisfy the collision constraints is set to infinity, otherwise it is recorded as 0.

[0064] (6) Communication cost

[0065] Communication between drones can be achieved through direct links or relay communication via other drone nodes. Therefore, determining whether the entire drone swarm meets communication requirements only requires randomly checking whether one drone can communicate with all other drones. At each moment, the... The drone and the first There are two scenarios for the drones: a line-of-sight (LoS) environment with no obstructions and a non-line-of-sight (NLoS) environment with obstructions. Understanding this, if there are obstacles between the planned path points for these two drones, it's an NLoS environment; otherwise, it's a LoS environment. The communication cost differs in these two scenarios. The specific calculation method includes: traversing each path point; at each path point, randomly selecting a drone and calculating the signal-to-noise ratio (SNR) of its communication with other drones; and determining whether there are any two drones whose SNR is less than a threshold. If so, the communication cost of that path point is set to infinity; otherwise, it's set to 0.

[0066] Before calculating the signal-to-noise ratio (SNR) for communication between the selected UAV and other UAVs, the path point channel coefficients can be calculated first. Then, the SNR can be calculated based on these channel coefficients, and finally, the communication cost can be calculated based on the SNR. The specific calculation process is as follows:

[0067] ,

[0068] ,

[0069] ;

[0070] in, It refers to the signal transmission power of the drone. This is the noise power (in practice, it can be taken as -110dBm, but it is not limited to this). This refers to the path loss at a distance of 1 meter under line-of-sight conditions. , It's the speed of light. It is the frequency of the signals transmitted by the drone; It is the first The distance between drones at each waypoint This is a factor used to measure the impact of distance on the signal; its value can be 2. It is the channel coefficient. It is the path loss factor in an NLoS environment. , It is the path length the signal travels inside the obstacle. This is the influence coefficient of obstacles on the signal. Its value can vary depending on factors such as the material and temperature of the obstacle. For example, when the obstacle is a concrete structure... 0.4 can be taken. For signal-to-noise ratio, For the first The communication cost at each path point.

[0071] Based on the above communication cost, height cost, angle cost, obstacle avoidance cost, collision cost, and path length cost, the objective function of the path planning model in this invention is:

[0072] .

[0073] Understandably, the ultimate goal of common UAV path planning problems is to find the feasible route with the shortest path length for the entire UAV swarm, i.e., the path with the lowest length cost. In this invention, the path must also satisfy five constraints under given conditions: path altitude constraint, turning angle constraint, obstacle collision avoidance constraint, UAV collision avoidance constraint, and communication maintenance constraint. By designing these constraint-based penalty mechanisms, the "risk" of each path is quantified. Furthermore, based on the objective function above, appropriate weights can be assigned to each cost item to further balance path length cost and path risk.

[0074] S30. The path planning model is solved by improving the wolf pack algorithm to obtain the optimal UAV swarm collaborative path. The improved wolf pack algorithm improves at least the search strategy of the wolf pack algorithm. The improved search strategy selects a search strategy from multiple search strategies based on the population vitality value and the problem dimension.

[0075] In the wolf pack algorithm, each individual in the population represents a possible solution for the path of all drones in the drone swarm.

[0076] In this invention, several improvements have been made to the wolf pack algorithm. The specific improvements of the wolf pack algorithm proposed in this invention will be explained in detail below, in conjunction with the execution flow of the wolf pack algorithm.

[0077] First, in the wolf pack initialization phase, the improved wolf pack algorithm proposed in this invention improves the population initialization method by combining the optimal point set method with LHS (Latin Hypercube Sampling) to form a hybrid initialization strategy, specifically including:

[0078] (1) Generate LHS samples and best point set samples.

[0079] Specifically, the process of generating LHS samples is as follows:

[0080] It has Each dimension has an interval that is divided into several dimensions. Let there be equal probability subintervals. Indicates the sample number. Represents the dimension. For each dimension... First, randomly generate a permutation. , making Indicates the first The sample at the th The interval numbering of the dimension. Then, for each sample, at the dimensional interval numbering... The sampling points of a dimension can be calculated by the following formula:

[0081] ;

[0082] in, From uniform distribution Random numbers obtained from sampling. By analyzing its population generation formula, it can be determined that the LHS method ensures that there is at least one sample in each interval by randomly selecting a position within the interval where the sample is located, while also achieving randomness within each sub-interval.

[0083] The process of generating the optimal point set samples is as follows:

[0084] Generate a sample set of optimal points using prime numbers, and use modular arithmetic to ensure that the sample points are uniformly distributed in the solution space. Assume... yes m Euclidean geometry of 3D, if ,So .in, Indicates the number of samples. This represents the set of best points, while r This refers to the optimal point, and its generation formula is as follows:

[0085] ;

[0086] in, K It is to satisfy The smallest prime number, It is a random number.

[0087] Then, the steps for constructing the initial population using the optimal point set are as follows: Step 1: According to r The generation formula is generated ;

[0088] Step 2: Construct the number of... N Best spots ;

[0089] Step 3: Collect the best spots Mapped to the value space domain of the population:

[0090] ;

[0091] in, Representing dimensions (This could be the upper limit of the value for path segment length, vertical angle, or horizontal angle, etc.) Representing dimensions The lower limit of the value.

[0092] (2) The LHS samples and the best point set samples are linearly combined using dynamic weights. The dynamic weights are calculated as follows:

[0093] ;

[0094] in, For the dimensions of LHS, This indicates dynamic weights.

[0095] Here, through dynamic weights By linearly combining LHS samples and the best point set samples, the advantages of both methods are balanced, and the proportions of LHS and best point set in the initial population are balanced. The calculation takes into account the dimension of the problem, so that the weight of LHS increases relatively when the dimension is high, in order to make up for the shortcomings of the optimal point set in high-dimensional space.

[0096] (3) Perform boundary processing on the combined samples to ensure that all sample points are within the defined solution space.

[0097] Throughout the execution of the wolf pack algorithm, fitness needs to be calculated multiple times. The improved wolf pack algorithm proposed in this invention uses the objective function of the path planning model as the fitness function of the artificial wolf.

[0098] In the walking phase of the wolf pack algorithm, this invention improves the search strategy. The improved search strategy defines the wolf pack's walking behavior based on the population vitality value.

[0099] Specifically, first calculate the population vitality value H:

[0100] ;

[0101] in, t It refers to the number of times the alpha wolf changes. This is the current iteration number. It is the average fitness of the top 10% of individuals in the current population, ranked by fitness. It is the average fitness value of the middle 10% of individuals in the current population, ranked by fitness. It is the average fitness value of the bottom 10% of individuals in the current population, ranked by fitness. This is the fitness value of the current alpha wolf.

[0102] Then, the current state of the population is evaluated based on the population vitality value H. Specifically, four vitality value intervals are pre-defined: [0, 0.25], (0.25, 0.5], (0.5, 0.75], and (0.75, 1]. The current state of the population is determined by which interval the population vitality value H falls within, and then an appropriate search strategy is selected based on the population state. Specifically, when the problem dimension is less than 50, the four vitality value intervals, from smallest to largest, correspond to four search strategies: adaptive step, chaos theory, Levy flight, and random perturbation factor. When the problem dimension is 50 or greater, the four vitality value intervals, from smallest to largest, correspond to four search strategies: Levy flight, chaos theory, adaptive step, and random perturbation factor.

[0103] Here, by combining the population vitality value H and the problem dimension to select different improved search strategies, a balance can be struck between convergence speed, search accuracy, and running time, thereby maximizing the possibility of achieving the optimal result with minimal overhead. Taking 2D and 50D spaces as examples, the optimal value (BEST) and running time (time) when solving the Sphere function are as follows: Figure 2 As shown.

[0104] The search is conducted based on the selected search strategy to simulate the roaming behavior of a wolf pack. After the roaming ends, the wolf's location is determined to satisfy the following conditions:

[0105] ;

[0106] In the formula, For walking stride length; , The upper limit of the dimension; Because it simulates the different search and movement abilities of wolves in nature, The number of directions to explore is determined by random integers, generally... The larger the value, the more directions the search engine can explore, resulting in better global optimization, but it will also increase the search time. The location of the scout after the roaming ended. This is the location of the scout wolf before it moves out.

[0107] In the aggressive attack phase of the wolf pack algorithm, the improved wolf pack algorithm proposed in this invention also improves the attack step size. The improved attack step size is a distance-adaptive attack step size, and its calculation method is as follows:

[0108] ;

[0109] in, It is the first in the wolf pack The distance between an artificial wolf and its prey It is the average distance of all wolves from their prey. It is the furthest distance from the prey currently within the wolf pack. It is the closest distance between the wolf pack and its prey. and They are The upper and lower limits of the value, Step size factor A random number between 0 and 1 Indicates the stride length of a sprint. and The preset weights are 0.5 and 0.8 respectively.

[0110] The population is updated based on the stride length during the raid, and the positional changes during the raid are as follows:

[0111] ;

[0112] In the formula, For the first k After the nth iteration The location of the artificial wolf For the first k The position of the alpha wolf after the next iteration.

[0113] After the raid ends, the siege action begins. First, the siege step length is calculated:

[0114] ;

[0115] in, It is the maximum number of iterations. It is the first The distance between each artificial wolf and the alpha wolf. It's the wolf's adaptability. It is the lowest fitness in the population. It's the alpha wolf's adaptability. This is the distance required to enter the siege. As the number of iterations increases, the closer an individual is to the alpha wolf, the smaller the siege step size, and the more refined the search becomes. The positional changes of the artificial wolf during the siege are shown in the following formula:

[0116] ;

[0117] It is a random number that takes the value between (0, 1).

[0118] During the population renewal phase, the probability of an individual generating offspring is first calculated using its fitness as a reference indicator:

[0119] ;

[0120] in, It refers to the size of the population. b It is the individual's fitness ranking in the entire population (from high to low, i.e., the alpha wolf is ranked 1).

[0121] Then, R wolves are randomly selected based on probability to reproduce. This improves the initial wolf pack algorithm's mechanism of randomly generating R wolves for elimination and renewal. The position of offspring individuals is generated as follows:

[0122] ;

[0123] in, This is the current position of the alpha wolf. It is the first The positions of the artificial wolves are represented by three summation terms, reflecting the influence of the alpha wolf on offspring, the influence of the parent on offspring, and the random perturbations caused by environmental factors. Then, according to the set population update ratio factor, the R artificial wolves with the smallest objective function value are eliminated, and R new artificial wolves are randomly generated from the solution space to complete the population update. R is a random number whose value ranges from [value missing]. , To update the scaling factor.

[0124] By performing multiple population iterations and updates, after each update, it is determined whether the termination condition has been reached. If it has, the iteration ends, and the current optimal solution is calculated and output. Otherwise, the above walking, rushing, besieging and population update process is repeated until the maximum number of iterations is reached, at which point the algorithm ends, and the current optimal solution is calculated and output.

[0125] After obtaining the optimal solution, i.e. the optimal UAV swarm cooperative path scheme, since the scheme is a discretized representation of the path, the trajectory can be smoothed using B-spline curves, thereby obtaining a complete UAV path that meets the actual requirements.

[0126] This invention provides a collaborative path planning method for UAV swarms in urban logistics delivery. It calculates the flight path of a UAV swarm under multiple constraints while collaboratively performing logistics delivery tasks in densely built-up urban environments. Specifically, this method addresses the collaborative path planning scenario of UAV swarms in densely built-up 3D urban environments. It constructs a path planning model that integrates altitude cost, angle cost, obstacle avoidance cost, collision cost, communication cost, and path length cost. Combined with an improved wolf pack algorithm, it solves problems related to maintaining communication links and avoiding collisions among UAV swarms. The proposed algorithm framework is novel, with fast convergence and high efficiency. Especially for UAV swarm path planning in densely built-up urban environments, it can quickly solve the overall path planning scheme for the UAV swarm, improving task execution efficiency, reducing energy consumption, and ensuring flight safety.

[0127] To verify the effectiveness and superiority of the improved wolf pack algorithm proposed in this invention in solving the path planning problem of UAV swarms in dense urban building environments, the inventors conducted simulation experiments and comparative experiments, and set up three-dimensional simulated terrain and different numbers of communication interference fields, and the start and end points of each UAV in the UAV swarm were different.

[0128] Experiment 1:

[0129] In Experiment 1, a densely built-up city model was set up within a terrain area of ​​1045×879, such as... Figure 3 As shown, the actual height data has been enlarged for visual clarity, but this does not affect the experimental results. The light blue areas in the image represent buildings with an actual height greater than 40 meters, while the green areas represent buildings with an actual height less than 40 meters.

[0130] Based on the aforementioned city model, a path planning problem involving 6 drones is considered. Each individual in the population contains paths for all 6 drones, and the number of distance evaluation points is consistent with the number of path points. Furthermore, and All are specified as 45°. The value is taken as 50 meters, communication impact factor It is 0.3. Taking a distance of 10 meters, the radius of the UAV is designed to be 0.3 meters. The relevant parameter settings for calculating the communication cost are: c = 3 x 10. 8 m / s, = 2.4 GHz, = 26 dBm, = -110 dBm, = 0.1, =0.22, SNRt = 2.5.

[0131] The settings for the communication interference field are shown in Table 1:

[0132] Table 1

[0133]

[0134] Based on the above settings, the proposed Multi-strategy Improved Wolf Pack Algorithm (MSIWPA) and five existing algorithms were used for UAV swarm path planning, and the planning results were compared. The five existing algorithms included the Coontail Algorithm (COA), the Fishing Algorithm (CFOA), the Magnificent Swan-Wren Algorithm (SFOA), the Instance-Based Learning Algorithm (IBL), and the Tornado Algorithm (TOC). These algorithms were executed 30 times, with a maximum of 100 iterations, and a swarm size of 30. The starting and ending points of each UAV were obtained, as shown in Table 2.

[0135] Table 2

[0136]

[0137] Considering the randomness of the search process of each algorithm, the path corresponding to the experiment with the 15th highest final cost function value out of 30 experiments was selected for plotting. A comparison of the final costs of each UAV for the six algorithms is shown below. Figure 4 As shown in the diagram, since each algorithm found its optimal solution in the final result, the collision cost and communication cost are both 0. The sum of the costs of each drone in the diagram is the total cost of the drone swarm path. Figure 4 As can be seen, the rightmost red rectangle in the diagram has the lowest height. Compared to other algorithms, the multi-strategy improved wolf pack algorithm proposed in this invention achieves the minimum cost for each drone in the final result, reflecting that its obtained path is superior to other algorithms. The experiment included the number of path points. N Setting it to 12 is generally the correct way to set it. N The larger the number of path points, the more finely divided the path becomes, resulting in smaller path segments between path points. This increases the probability that all path segments will also satisfy the constraints, provided all path points satisfy them. However, this also leads to greater computational resource consumption and longer computation time. Therefore, the number of path points... N It should be appropriately adjusted according to specific practical needs.

[0138] The path results obtained from the multi-strategy improved wolf pack algorithm of this invention are plotted in the model, such as... Figure 5As shown, the six line graphs represent the flight paths of the six drones, the red cylinders represent the distribution of communication interference fields, and the green cuboids represent buildings in the city. It can be seen that each drone in the swarm starts from its own starting point, avoids building obstacles, bypasses communication interference fields, and reaches its own destination while maintaining inter-swarm communication. The multi-strategy improved wolf pack algorithm proposed in this invention yields a relatively smooth final path result, with no large turning angles in the paths of the six drones. Furthermore, the flight paths of each drone can be more compact without collisions, better ensuring communication connections between drones, and the path length is also relatively short. From the top-down view, each drone flies over some buildings at a height higher than the buildings; however, combined with the 3D graph, it can be seen that each drone's path passes through parts of the buildings. This is because when evaluating obstacle avoidance costs, only the path points are evaluated; path segments may still pass through obstacles, therefore the overall objective function value is not infinite.

[0139] Then, the flight altitude of each drone is analyzed, such as... Figure 6 The left figure compares the average flight altitude of the six drones in the results of each algorithm. Each rectangular bar represents the average flight altitude of a drone across 12 path points. It can be seen that the multi-strategy improved wolf pack algorithm proposed in this invention has the smallest variation in the average flight altitude of the six drones, indicating that the flight altitudes of the drones within the swarm are relatively concentrated. The flight altitudes planned by other algorithms also show a concentrated trend, all relatively concentrated around 70 meters. Because the layered strategy divides the flight space into two layers, excessively high or low flight altitudes will lead to high altitude costs; therefore, the flight altitudes of the results from each algorithm strive to approach the optimal flight altitude range. The right figure compares the maximum altitude difference of the drone paths in the results of each algorithm. It can be seen that the multi-strategy improved wolf pack algorithm has the smallest relevant value, indicating that the drone flight altitude variation is minimal, and therefore its altitude cost is also lower.

[0140] Experiment 2:

[0141] Based on Experiment 1, the number of communication interference fields was increased to 8. The relevant settings for the 3 additional communication interference fields are shown in Table 3.

[0142] Table 3

[0143]

[0144] Similar to Experiment 1, compare the path costs of drone swarms using the results of the six algorithms (e.g., Figure 7 ), and plot the resulting path in the model, such as Figure 8 .from Figure 7 and Figure 8As can be seen, even with an increase in the number of communication interference fields, the multi-strategy improved wolf pack algorithm proposed in this invention can still obtain better path results.

[0145] Specifically, from Figure 7 As can be seen from the data, among the six rectangles corresponding to each drone, the rightmost red rectangle has the lowest height. This indicates that the multi-strategy improved wolf pack algorithm proposed in this invention can find a path with lower cost in the path planning of each drone, demonstrating its higher search accuracy.

[0146] Figure 8 In the diagram, six line graphs represent the flight paths of six drones, while red cylinders represent the distribution of communication interference fields, and green cuboids represent buildings in the city. Analyzing the top-view view of the final path results of this invention, compared to the path results in Experiment 1, even with the increased number of communication interference fields, each drone can still plan a path while maintaining communication. As can be seen from the 3D diagram, some path segments shown as passing through buildings in the top-view view actually do not pass through them, but rather pass over them. This invention avoids frequent turns between buildings by increasing the altitude cost of some paths, thus reducing the overall cost of the path.

[0147] Then, similar to Experiment 1, Experiment 2 also performed an analysis of the drone's flight altitude, such as... Figure 9 By comparing the average flight altitude and maximum altitude difference of each UAV, it can be seen that the multi-strategy improved wolf pack algorithm proposed in this invention still achieves the smoothest final path even after increasing the communication interference field in the environment, realizing dynamic consistency at the altitude level. Simultaneously, the hierarchical planning strategy enables individuals to automatically adjust to the optimal altitude range to minimize communication delay, energy consumption, or collision risk. Even though the maximum altitude difference on UAV5 is not the smallest, its overall performance still ranks first.

[0148] The above experiments demonstrate that for drone swarm path planning in densely built-up urban environments, existing path planning methods still suffer from convergence to local optima and significantly increased computation time due to the complex urban environment. In contrast, the multi-strategy improved wolf pack algorithm employed in this invention achieves the lowest path cost to the optimal solution compared to other algorithms. This invention enables drone swarms to collaboratively perform logistics delivery tasks in densely built-up urban environments, calculating flight paths that satisfy multiple constraints.

[0149] The method provided in this invention can be applied to electronic devices. Specifically, the electronic device can be a desktop computer, a portable computer, a smart mobile terminal, a server, etc. No limitation is made herein; any electronic device that can implement this invention falls within the protection scope of this invention.

[0150] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention.

[0151] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0152] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0153] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus (devices), or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects, all of which are collectively referred to herein as "modules" or "systems." Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The computer program may be stored / distributed in a suitable medium, provided with or as part of other hardware, or may take other distribution forms, such as via the Internet or other wired or wireless telecommunications systems.

[0154] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0155] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0156] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0157] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for collaborative path planning of drone swarms for urban logistics delivery, characterized in that, include: Load a 3D environmental model containing obstacle information and communication interference fields; The path planning model is initialized based on the 3D environment model and task information; wherein, the task information includes at least the start and end point positions of each UAV in the UAV swarm; the path planning model is a problem model with the goal of minimizing the overall cost of the UAV swarm's cooperative path; the overall cost includes altitude cost, angle cost, obstacle avoidance cost, collision cost, communication cost, and path length cost; The optimal UAV swarm collaborative path is obtained by solving the path planning model using an improved wolf pack algorithm. The improved wolf pack algorithm improves at least the search strategy of the wolf pack algorithm. The improved search strategy selects a search strategy from multiple search strategies based on the population vitality value and the problem dimension. The path length cost is calculated using the following formula: ； Among them, the The first drone's flight path The coordinates of the path points are , No. The starting coordinates of the flight path of each drone are , No. The coordinates of the endpoint of the flight path of the drone are , The number of waypoints on the flight path. Indicates the first The path length cost of a single drone.

2. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The altitude cost is calculated using the following formula: ； in, , It is the average height of buildings; It is the drone's maximum climb rate. It is a safety margin set to cope with sudden airflow or positioning errors; , , The optimal flight altitude range for drones. and The flight altitude is determined by the requirements of the mission. Indicates the first The drone in its flight path Flight altitude at each waypoint It is a preset obstacle influence factor. These are preset environmental airflow influencing factors. Indicates the first The drone in its flight path The elevation cost corresponding to each path point.

3. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The angle cost is calculated using the following formula: ； in, Indicates the first The drone in its flight path The angle cost corresponding to each path point. This is the maximum turning angle of the drone. This is the maximum climb angle for the drone. For the first The turning angle of a drone For the first The climb angle of a drone, and The following formula is used to calculate: ， ； Among them, the The first drone's flight path The coordinates of the path points are .

4. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The cost of avoiding obstacles is calculated using the following formula: ； Among them, the The first drone's flight path The coordinates of the path points are , It is a set of geometric parameters describing the cuboid space in which the obstacle is located. It is the first The first drone's flight path The Euclidean distance of each path point from the center of the obstacle. It is the farthest effective range of the communication interference field. It is a preset communication impact factor. Indicates the first The drone in its flight path The cost of avoiding obstacles at each path point.

5. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The collision cost is calculated using the following formula: ； in, For the first The drone and the first The first drone on the same path segment The distance to each distance assessment point It refers to the number of drones in the cluster. The radius of the drone. This refers to the minimum safe flight distance between drones to avoid collisions. Indicates the first The drone in the first The collision cost corresponding to each distance evaluation point.

6. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The communication cost is calculated in the following manner: At each path point, a drone is randomly selected to calculate the signal-to-noise ratio (SNR) of its communication with other drones. It is then determined whether there are two drones whose SNR is less than the SNR threshold. If so, the communication cost of that path point is set to infinity; otherwise, it is set to 0.

7. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The improved wolf pack algorithm also improves the population initialization method of the wolf pack algorithm, including: Generate LHS samples and optimal point set samples; The LHS samples and the best point set samples are linearly combined using dynamic weights; The dynamic weights are calculated as follows: ； in, For the dimensions of LHS, This indicates dynamic weights.

8. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 7, characterized in that, The improved wolf pack algorithm also improves the raid step size. The improved raid step size is a distance-adaptive raid step size, and its calculation method is as follows: ； in, It is the first in the wolf pack The distance between an artificial wolf and its prey It is the average distance of all wolves from their prey. It is the furthest distance from the prey currently within the wolf pack. It is the closest distance between the wolf pack and its prey. and They are The upper and lower limits of the value, Step size factor A random number between 0 and 1 Indicates the stride length of a sprint. and The preset weights.

9. The method for collaborative path planning of drone swarms for urban logistics delivery according to claim 1, characterized in that, The improved wolf pack algorithm uses the objective function of the path planning model as the fitness of the artificial wolves, and the objective function is: ； in, Indicates the first The path length cost of a single drone Indicates the first The drone in its flight path The height cost corresponding to each path point Indicates the first The drone in its flight path The angle cost corresponding to each path point. Indicates the first The drone in its flight path The cost of avoiding obstacles at each path point Indicates the first The drone in the first The collision cost corresponding to each distance evaluation point Indicates the first The communication cost of a drone The number of waypoints on the flight path. It refers to the number of drones in the cluster. For the fitness of artificial wolves.

Images