Unmanned aerial vehicle group path planning method, device and equipment and storage medium

Abstract

The invention discloses an unmanned aerial vehicle group path planning method, which comprises the following steps: acquiring environment data of an environment in which an unmanned aerial vehicle group is located within a preset time period, and constructing a training set according to the environment data; inputting the training set into a pre-trained environment prediction model, and outputting an environment prediction data set; obtaining a flight path of a single unmanned aerial vehicle according to the environment prediction data set by adopting a black-wing algorithm; according to the flight paths of all the unmanned aerial vehicles in the unmanned aerial vehicle group and the environment prediction data set, performing path adjustment by adopting an MAPF path planning algorithm to obtain an optimal path of the unmanned aerial vehicle group; according to the invention, the optimal planning of the path of the unmanned aerial vehicle can be rapidly completed, and the real-time performance and accuracy of path planning are improved.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle control technology, and in particular to a method, device, equipment and storage medium for unmanned aerial vehicle group path planning. Background Art

[0002] With the rapid development of drone technology, 5G communication technology, artificial intelligence and machine learning technology, as well as the wind power industry's demand for efficient and intelligent inspection methods, drone path planning in wind turbine inspection is particularly important. Through the combined application of reasonable path planning algorithms and drone technology, efficient and intelligent inspection of wind turbines can be achieved, thereby improving the operation and maintenance management level and economic benefits of wind farms.

[0003] Existing UAV path planning technologies mainly include path planning methods based on graph search, path planning methods based on sampling, and path planning methods based on optimization. These methods mainly build environmental models and then use various optimization algorithms for path planning. When building the model, it is often necessary to establish a global map. In order to improve the accuracy of the map, more grids will be used to represent the flight area of ​​the UAV, which will greatly increase the computer's hungry data storage and calculation amount, resulting in longer algorithm running time. In addition, the static environment cannot meet the requirements of the UAV to quickly respond to dynamic environmental changes and plan paths in real time, resulting in deficiencies in the algorithm in terms of real-time and practicality. Summary of the invention

[0004] The embodiment of the present invention provides a method for UAV swarm path planning, which can quickly complete the optimal planning of UAV paths and improve the real-time performance and accuracy of path planning.

[0005] In a first aspect, an embodiment of the present invention provides a method for planning a path of a drone group, comprising:

[0006] Obtain environmental data of the environment in which the drone swarm is located within a preset time period, and construct a training set based on the environmental data;

[0007] Inputting the training set into a pre-trained environment prediction model, and outputting an environment prediction data set;

[0008] Using the Black Kite algorithm, the flight path of a single UAV is obtained according to the environmental prediction data set;

[0009] According to the flight paths of all drones in the drone swarm and the environmental prediction data set, a MAPF path planning algorithm is used to adjust the path to obtain the optimal path of the drone swarm.

[0010] Furthermore, before inputting the training set into the pre-trained environment prediction model, the method further includes:

[0011] The missing values ​​in the training set are supplemented, and the supplemented training set is normalized.

[0012] Furthermore, the environment prediction model is a BiTCN-BiGRU-Attention network model, then, inputting the training set into a pre-trained environment prediction model and outputting an environment prediction data set includes:

[0013] Inputting the training set into the BiTCN-BiGRU-Attention network model;

[0014] The BiTCN layer is used to perform forward convolution calculation on the training set, extract forward data features, and send the extracted information to the BiGRU layer;

[0015] The BiGRU layer is used to perform forward GRU and reverse GRU processing on the data received from the BiTCN layer, and learn the dynamic changes of data from two directions;

[0016] The Attention layer assigns different weights to the input data by training weights, thereby increasing the accuracy of the prediction;

[0017] According to the prediction result output by the Attention layer, an environment prediction data set is obtained.

[0018] Furthermore, the black kite algorithm is used to obtain the flight path of a single UAV according to the environmental prediction data set, including:

[0019] Step 1: Initializing a black-winged kite population based on the environmental prediction data set to obtain a flight path of each black-winged kite individual in the black-winged kite population;

[0020] Step 2: Calculate the fitness of each black-winged kite individual according to a preset fitness function, select the black-winged kite individual with the best fitness as the initial leader, and set the maximum number of iterations;

[0021] Step 3: In each iteration, the attack behavior and migration behavior of the black-winged kite are simulated, and after one iteration, the fitness of all black-winged kite individuals is recalculated, and the black-winged kite individual with the best fitness is selected as the new leader; wherein the fitness can preferably be the maximum fitness or the minimum fitness;

[0022] Step 4: Determine whether the current number of iterations reaches the maximum number of iterations. If so, end the iteration and output the flight path and fitness of the new leader. If not, repeat step 3 until the current number of iterations reaches the maximum number of iterations.

[0023] Furthermore, the expression for simulating the attack behavior of the black-winged kite is as follows:

[0024]

[0025] in, and They represent the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, r is a random number ranging from 0 to 1, p is a constant with a value of 0.9, T is the maximum number of iterations, and t is the current number of iterations.

[0026] Furthermore, the expression for simulating the migration behavior of black-winged kites is as follows:

[0027]

[0028] m = 2 × sin (r + π / 2);

[0029] in, represents the latest leader of the j-th black-winged kite at the t-th iteration so far, and denote the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, respectively. i Represents the fitness value of any individual in the current population, F ri represents the fitness value of the random population, r is a random number ranging from 0 to 1, and C(0,1) represents the Cauchy mutation.

[0030] Furthermore, the path adjustment is performed using a MAPF path planning algorithm according to the flight paths of all drones in the drone swarm and the environmental prediction data set to obtain the optimal path of the drone swarm, including:

[0031] For each of the drones, determining a starting position and a target position of the flight according to the flight path;

[0032] Generate a conflict-free path for each of the drones using a depth-first search algorithm according to the environmental prediction data set, the starting position, and the target position;

[0033] According to the conflict-free paths of all the drones, the optimal path of the drone group is obtained.

[0034] In a second aspect, an embodiment of the present invention provides a drone group path planning device, comprising:

[0035] An environmental data acquisition module is used to acquire environmental data of the environment in which the drone group is located within a preset time period and to construct a training set based on the environmental data;

[0036] An environmental data prediction module, used to input the training set into a pre-trained environmental prediction model and output an environmental prediction data set;

[0037] A black kite algorithm module, used to obtain a flight path of a single UAV based on the environmental prediction data set using a black kite algorithm;

[0038] The path planning module is used to adjust the path according to the flight paths of all drones in the drone group and the environmental prediction data set by using the MAPF path planning algorithm to obtain the optimal path of the drone group.

[0039] In a third aspect, an embodiment of the present invention provides an electronic device, including:

[0040] Memory for storing computer programs;

[0041] A processor, configured to execute the computer program;

[0042] Wherein, when the processor executes the computer program, it implements the drone group path planning method described in any one of the first aspects above.

[0043] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed, the drone swarm path planning method described in any one of the first aspects above is implemented.

[0044] Compared with the prior art, the method for path planning of a drone swarm provided by an embodiment of the present invention has the following beneficial effects: by acquiring environmental data of the environment in which the drone swarm is located within a preset time period, a training set is constructed according to the environmental data; the training set is input into a pre-trained environmental prediction model, and an environmental prediction data set is output; a black kite algorithm is used to obtain the flight path of a single drone according to the environmental prediction data set; a MAPF path planning algorithm is used to adjust the path according to the flight paths of all drones in the drone swarm and the environmental prediction data set, so as to obtain the optimal path of the drone swarm; the present invention can quickly complete the optimal planning of the drone path, and improve the real-time and accuracy of path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical features of the embodiments of the present invention, the drawings required for use in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1It is a flowchart of an embodiment of a method for planning a path for a drone group provided by the present invention;

[0047] Figure 2 It is a schematic diagram of a depth-first search algorithm of an embodiment of a drone swarm path planning method provided by the present invention;

[0048] Figure 3 It is a structural schematic diagram of an embodiment of a UAV swarm path planning device provided by the present invention;

[0049] Figure 4 It is a structural schematic diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0053] In a first aspect, an embodiment of the present invention provides a method for planning a path for a drone group, see Figure 1 , which is a flow chart of an embodiment of a UAV swarm path planning method provided by the present invention.

[0054] like Figure 1 As shown, the method comprises the following steps:

[0055] S1: Obtain environmental data of the environment in which the drone group is located within a preset time period, and construct a training set based on the environmental data;

[0056] S2: inputting the training set into a pre-trained environment prediction model, and outputting an environment prediction data set;

[0057] S3: using the black kite algorithm to obtain the flight path of a single UAV according to the environmental prediction data set;

[0058] S4: According to the flight paths of all drones in the drone swarm and the environmental prediction data set, a MAPF path planning algorithm is used to adjust the path to obtain the optimal path of the drone swarm.

[0059] In a specific implementation, firstly, the environmental data of the environment in which the drone swarm is located within a preset time period is obtained. By way of example, the environmental data includes but is not limited to temperature, humidity, air pressure, wind speed, wind direction, rainfall, etc., as well as data such as the fan height, fan spacing and fan blade length of the drone. A training set is constructed based on the environmental data, and the environmental data is predicted based on the training set. The time span and data granularity of data collection will affect the prediction results. The specific values ​​are determined according to the prediction requirements. For example, the time span is generally more than one year and includes all seasons and holidays. If it is necessary to predict environmental data at the hourly level, the data granularity should also be at the hourly level.

[0060] The above training set is input into the pre-trained environment prediction model, the environment prediction model predicts the environment and outputs the environment prediction data set. The flight path of a single UAV is obtained by using the Black Kite algorithm in combination with the environment prediction data set. According to the flight paths of all UAVs in the UAV swarm and the environment prediction data set, the MAPF path planning algorithm is used to adjust the path conflicts and obtain the optimal path of the UAV swarm.

[0061] In summary, the present invention obtains environmental data of the environment in which the drone group is located within a preset time period, and constructs a training set according to the environmental data; the training set is input into a pre-trained environmental prediction model, and an environmental prediction data set is output; the flight path of a single drone is obtained according to the environmental prediction data set by using the black kite algorithm; the path is adjusted according to the flight paths of all drones in the drone group and the environmental prediction data set by using the MAPF path planning algorithm to obtain the optimal path of the drone group; the present invention adopts the optimized BiTCN-BiGRU-Attention network model, which can realize the prediction of complex environments without relying on the global map, thereby improving the processing efficiency and accuracy, and realizes global exploration and dynamic optimization by simulating animal behavior through the improved black kite optimization algorithm. This simulation not only enhances the robustness of the algorithm, but also enables it to more effectively cope with the constantly changing optimization environment, introduces the Cauchy mutation strategy, helps the algorithm to jump out of the local optimal solution, and increases the probability of finding a better solution in the global search space, and further improves the versatility and adaptability of the algorithm through the adjustable parameters in the algorithm, and can quickly complete the optimal planning of the drone path, and improve the real-time and accuracy of the path planning.

[0062] In an optional implementation, before inputting the training set into a pre-trained environment prediction model, the method further includes:

[0063] The missing values ​​in the training set are supplemented, and the supplemented training set is normalized.

[0064] Specifically, before the training set is input into the pre-trained environmental prediction model, the data in the training set needs to be preprocessed, including the completion and normalization of missing data. Missing data refers to the problem of invalid data at some time due to the absence of discontinuous granularity or continuous data of multiple granularities. The optional methods to solve this problem include mean filling, median filling, interpolation, KNN (K-Nearest Neighbors) algorithm, etc. Exemplarily, the embodiment of the present invention adopts the KNN algorithm and uses the Y values ​​of K nearest neighbor samples to estimate the missing values. The formula is as follows:

[0065] Y i =(Y i1 +Y i 2+...+Y i k) / K;

[0066] Among them, Y i is the estimated value of the missing value in the target sample, Y i1 , Y i 2. ..., Y i k is the Y value of the corresponding feature among the K nearest neighbors to the target sample, and K is the number of nearest neighbors.

[0067] This formula means that for missing values ​​in the target sample, the Y values ​​of the corresponding features in its K nearest neighbors are added and then divided by K to obtain the estimated value of the missing value.

[0068] After the missing values ​​are supplemented, normalization is performed. Normalization is used to scale the data to the range of [0, 1], which can effectively improve the accuracy and performance of the model and ensure that all features are on the same scale. The normalization method can adopt Z-Score normalization, decimal calibration normalization or minimum-maximum normalization and other methods. Exemplarily, the embodiment of the present invention adopts the minimum-maximum normalization method, which scales the data to the range of [0, 1]. The specific steps are: find the minimum and maximum values ​​in the data, subtract the minimum value from each data point, and then divide it by the difference between the maximum value and the minimum value, repeat the above steps, and normalize the training set. The formula is as follows:

[0069] x * =(xx min ) / (xmax -x min );

[0070] Among them, x * is the normalized value, x is the original data, and x min is the minimum value in the data set, x max is the maximum value in the data set.

[0071] In an optional implementation, the environment prediction model is a BiTCN-BiGRU-Attention network model, then, inputting the training set into a pre-trained environment prediction model and outputting an environment prediction data set includes:

[0072] Inputting the training set into the BiTCN-BiGRU-Attention network model;

[0073] The BiTCN layer is used to perform forward convolution calculation on the training set, extract forward data features, and send the extracted information to the BiGRU layer;

[0074] The BiGRU layer is used to perform forward GRU and reverse GRU processing on the data received from the BiTCN layer, and learn the dynamic changes of data from two directions;

[0075] The Attention layer assigns different weights to the input data by training weights, thereby increasing the accuracy of the prediction;

[0076] According to the prediction result output by the Attention layer, an environment prediction data set is obtained.

[0077] Specifically, the preprocessed training set is input into a pre-trained environment prediction model, and the model outputs an environment prediction result. In an embodiment of the present invention, the environment prediction model adopts a BiTCN-BiGRU-Attention network model, which is a hybrid deep learning model that combines a bidirectional temporal convolutional network (BiTCN), a bidirectional gated recurrent unit (BiGRU), and an attention mechanism.

[0078] First, after the training set is input into the BiTCN-BiGRU-Attention network model, the BiTCN layer first performs forward convolution calculation on the input sequence to extract the forward data features. After each layer of convolution of the input sequence, the dilation factor increases exponentially. With the increase of convolution layers, the receptive field of BiTCN gradually increases. The increase of the receptive field of BiTCN may lead to problems such as gradient vanishing and slow convergence. In order to solve the problems of gradient vanishing and slow convergence, the residual block is introduced into the BiTCN layer. The residual block helps the network maintain the gradient signal during training by adding an identity mapping, thereby accelerating convergence and improving the performance of the model. BiTCN adopts a bidirectional temporal convolution structure, that is, it processes forward and reverse sequence data at the same time. This structure enables the network to capture the hidden features of the forward and backward directions in the sequence, so as to better understand the contextual relationship of the sequence. After being processed by the BiTCN layer, the extracted information is sent to the BiGRU layer for further processing.

[0079] The BiGRU layer consists of two independent GRUs, one for processing the forward sequence (from left to right) and the other for processing the reverse sequence (from right to left). The model can use the information of the previous and the following context at the same time to more fully understand the dynamic changes of the data. GRU can effectively capture the long-term dependencies in the sequence through its internal mechanism. Compared with LSTM, GRU has a simpler structure and fewer parameters, but it can achieve comparable or even better performance than LSTM in many tasks. This structural simplification makes GRU more efficient during training and can achieve powerful sequence modeling capabilities at a lower computational cost. BiTCN captures local temporal features through convolution operations, while BiGRU can learn more global and complex patterns based on these local features. This combination enables the model to more accurately capture the dynamic changes of data during prediction, thereby improving prediction accuracy.

[0080] Furthermore, the Attention mechanism can focus on the most important part of the sequence by calculating the correlation score between each element in the sequence and the prediction target, which helps the model not lose important information when processing long sequences, thereby improving the prediction accuracy of the model. The Attention layer introduces the weight parameter ω, which can strengthen the interaction between the internal features of the model, help the model learn more complex and abstract feature representations, and improve the generalization ability of the model.

[0081] In an optional implementation, the use of the Black Kite algorithm to obtain the flight path of a single UAV according to the environmental prediction data set includes:

[0082] Step 1: Initializing a black-winged kite population based on the environmental prediction data set to obtain a flight path of each black-winged kite individual in the black-winged kite population;

[0083] Step 2: Calculate the fitness of each black-winged kite individual according to a preset fitness function, select the black-winged kite individual with the best fitness as the initial leader, and set the maximum number of iterations;

[0084] Step 3: In each iteration, the attack behavior and migration behavior of the black-winged kite are simulated, and after one iteration, the fitness of all black-winged kite individuals is recalculated, and the black-winged kite individual with the best fitness is selected as the new leader; wherein the fitness can preferably be the maximum fitness or the minimum fitness;

[0085] Step 4: Determine whether the current number of iterations reaches the maximum number of iterations. If so, end the iteration and output the flight path and fitness of the new leader. If not, repeat step 3 until the current number of iterations reaches the maximum number of iterations.

[0086] Specifically, the environmental prediction dataset predicted by the BiTCN-BiGRU-Attention network model is input into the Black Kite Optimization Algorithm (BKA) to simulate its high adaptability to environmental changes and target locations, and to derive the optimal flight route for a single UAV.

[0087] First, based on the environmental prediction dataset, a black kite population is initialized. The location of the black kite is used as the solution. Each black kite individual represents a potential flight path. The initialization example is as follows:

[0088] X i =BK lb +rand(BK ub -BK lb );

[0089] Where i is an integer between 1 and pop, pop is the number of potential solutions, BK lb and BK ub are the lower and upper bounds of the i-th black-winged kite in the j-th dimension, and rand is a randomly selected value between 0 and 1.

[0090] According to the preset fitness function, the fitness of each black kite individual is calculated, and the individual with the best fitness is selected as the initial leader, which is considered to be the optimal position of the black kite. At the same time, a maximum number of iterations is set to control the termination condition of the algorithm. The mathematical representation of the initial leader is as follows, where the individual with the smallest fitness is taken as an example:

[0091] f best =min(f(X i ));

[0092] X L =X(find(f best= = f(X i )));

[0093] Among them, X L represents the initial leader, f(X i ) represents fitness.

[0094] In each iteration, the attack behavior and migration behavior of the black-winged kite are simulated to update the flight path of each black-winged kite individual. The attack behavior involves approaching the current leader (optimal flight path) or searching for a better solution near it. The migration behavior involves exploring new flight paths in the entire search space. After one iteration, the fitness of all black-winged kite individuals is recalculated, and the individual with the best fitness is selected as the new leader. It is determined whether the current number of iterations has reached the maximum number of iterations. If the maximum number of iterations is reached, the iteration process is terminated, and the flight path and fitness of the new leader are output. This flight path is considered to be the optimal flight path found by the algorithm. If the maximum number of iterations is not reached, the steps of simulating the attack behavior and migration behavior of the black-winged kite are repeated until the current number of iterations reaches the maximum number of iterations.

[0095] It should be noted that the fitness function is a key tool for evaluating the pros and cons of each individual, and usually corresponds to the objective function of the problem. If the goal of the problem is to maximize a certain objective function, the fitness function can be directly defined as follows:

[0096] Where λ≥0 and λ+f(x)≥0;

[0097] If the goal of the problem is to minimize a certain objective function, the fitness function can be defined as follows:

[0098] Where λ≥0 and λ-f(x)≥0;

[0099] The introduction of an adjustable parameter λ into the fitness function can control the fitness range. The selection of λ needs to be adjusted according to the specific problem to achieve the best optimization effect.

[0100] For each black kite solution in the population, the fitness value is calculated using the defined fitness function. This value reflects the performance of the solution in the optimization problem. When the goal is to minimize, the smaller the fitness value, the better the solution. When the goal is to maximize, the larger the fitness value, the better the solution.

[0101] In an optional embodiment, the expression for simulating the attack behavior of a black-winged kite is as follows:

[0102]

[0103] in, and They represent the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, r is a random number ranging from 0 to 1, p is a constant with a value of 0.9, T is the maximum number of iterations, and t is the current number of iterations.

[0104] Specifically, the attack behavior refers to the behavior of simulating the black-winged kite to catch small mammals and insects. During the flight, it adjusts the angle of its wings and tail according to the wind speed, hovers quietly to observe the prey, and then quickly dives to attack. This strategy includes different attack behaviors, such as hovering in the air, hovering in the air, etc., which are used for global exploration and search. The mathematical model of its attack behavior is as follows:

[0105]

[0106] in, and They represent the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, r is a random number ranging from 0 to 1, p is a constant with a value of 0.9, T is the maximum number of iterations, and t is the current number of iterations.

[0107] In an optional embodiment, the expression for simulating the migration behavior of black-winged kites is as follows:

[0108]

[0109] m = 2 × sin (r + π / 2);

[0110] in, represents the latest leader of the j-th black-winged kite at the t-th iteration so far, and denote the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, respectively. i Represents the fitness value of any individual in the current population, F ri represents the fitness value of the random population, r is a random number ranging from 0 to 1, and C(0,1) represents the Cauchy mutation.

[0111] Specifically, migration behavior refers to the behavior of simulating birds migrating from north to south in winter in order to adapt to seasonal changes. Migration is usually led by a leader, and the leader's navigation skills are crucial to the success of the team. In an embodiment of the present invention, if the fitness value of the current population is less than the fitness value of the random population, this indicates that it is not suitable for leading the population forward, and the leader will give up leadership and join the migrating population. On the contrary, if the fitness value of the current population is greater than the fitness value of the random population, the leader will guide the population to the destination. This strategy can dynamically select excellent leaders to ensure the success of migration. The following is a mathematical model of black-winged kite migration behavior:

[0112]

[0113] m = 2 × sin (r + π / 2);

[0114] in, represents the latest leader of the j-th black-winged kite at the t-th iteration so far, and denote the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, respectively. i Represents the fitness value of any individual in the current population, F ri represents the fitness value of the random population, r is a random number ranging from 0 to 1, and C(0,1) represents the Cauchy mutation, which is defined as follows:

[0115] The one-dimensional Cauchy distribution is a continuous probability distribution with two parameters. The probability density function of the one-dimensional Cauchy distribution is:

[0116]

[0117] When δ=1,μ=0, its probability density function becomes the standard form, and the exact formula is as follows:

[0118]

[0119] In an optional implementation, the step of adjusting the path using a MAPF path planning algorithm according to the flight paths of all drones in the drone swarm and the environmental prediction data set to obtain the optimal path of the drone swarm includes:

[0120] For each of the drones, determining a starting position and a target position of the flight according to the flight path;

[0121] Generate a conflict-free path for each of the drones using a depth-first search algorithm according to the environmental prediction data set, the starting position, and the target position;

[0122] According to the conflict-free paths of all the drones, the optimal path of the drone group is obtained.

[0123] Specifically, the embodiment of the present invention combines the MAPF (Multi-Agent Path Finding) strategy with the LaCAM (Local Collision Avoidance Mechanism), and utilizes technologies such as PIBT (Predictive Information Based Tree) and LaCAM to perform real-time path adjustment and collision avoidance during the flight of a drone swarm, thereby achieving path optimization and coordination between drones and ensuring collision-free flight.

[0124] The flight path of a single UAV output by the black kite algorithm is obtained. The starting position and target position of each UAV are determined according to the flight path, as well as the environmental prediction data set output by the environmental prediction model. The path of each UAV is initialized to empty. The combination of the starting position and the target position is recorded as a joint configuration space. A depth-first search (DFS) is performed in the joint configuration space. During the search process, a lazy constraint addition method is adopted, that is, necessary constraints are added only when conflicts are encountered. For an exemplary schematic diagram of the depth-first search, see Figure 2 As shown in Figure 2, each node represents a drone and the initial source point of DFS is v 0 , using depth-first search, first visit v 0 -v 1 -v 2 -v 5 , to v 5 If there is no node behind, backtrack to v 1 , that is, the nearest node v that is connected to the visited node 1 , then from v 1 Departure, visit v 1 -v 4 -v 6 -v 3 , at this time with v 3 Two connected nodes v 0 With v 6 All have been visited, backtrack to v with unvisited nodes 6 , and then from v 6 Departure, visit v 6 -v 7 , to v 7 There is no node behind, so backtrack to the source point v 0 , v 0 There are no unvisited nodes and the depth-first search process ends.

[0125] Based on the current search status, a path is generated for each drone and it is verified whether this path conflicts with the paths of other drones and does not collide with obstacles in the environment. If a path is proven to be conflicting and cannot be resolved by adding new constraints, LaCAM may backtrack to the previous node in the search tree and try to explore other possible paths. The above steps are iterated until conflict-free paths are found for all drones and the optimal path for the drone swarm is output.

[0126] In a second aspect, an embodiment of the present invention provides a drone group path planning device, see Figure 3 , which is a structural schematic diagram of an embodiment of a drone swarm path planning device provided by the present invention.

[0127] like Figure 3 As shown, the device comprises:

[0128] An environmental data acquisition module 31 is used to acquire environmental data of the environment in which the drone group is located within a preset time period, and to construct a training set based on the environmental data;

[0129] An environmental data prediction module 32, used to input the training set into a pre-trained environmental prediction model and output an environmental prediction data set;

[0130] A black kite algorithm module 33, configured to use a black kite algorithm to obtain a flight path of a single UAV according to the environmental prediction data set;

[0131] The path planning module 34 is used to adjust the path using the MAPF path planning algorithm according to the flight paths of all the drones in the drone group and the environmental prediction data set to obtain the optimal path of the drone group.

[0132] In an optional implementation, before inputting the training set into a pre-trained environment prediction model, the method further includes:

[0133] The missing values ​​in the training set are supplemented, and the supplemented training set is normalized.

[0134] In an optional implementation, the environment prediction model is a BiTCN-BiGRU-Attention network model, then the environment data prediction module 32 is further used for:

[0135] Inputting the training set into the BiTCN-BiGRU-Attention network model;

[0136] The BiTCN layer is used to perform forward convolution calculation on the training set, extract forward data features, and send the extracted information to the BiGRU layer;

[0137] The BiGRU layer is used to perform forward GRU and reverse GRU processing on the data received from the BiTCN layer, and learn the dynamic changes of data from two directions;

[0138] The Attention layer assigns different weights to the input data by training weights, thereby increasing the accuracy of the prediction;

[0139] According to the prediction result output by the Attention layer, an environment prediction data set is obtained.

[0140] In an optional implementation, the black kite algorithm module 33 is further used to:

[0141] Step 1: Initializing a black-winged kite population based on the environmental prediction data set to obtain a flight path of each black-winged kite individual in the black-winged kite population;

[0142] Step 2: Calculate the fitness of each black-winged kite individual according to a preset fitness function, select the black-winged kite individual with the best fitness as the initial leader, and set the maximum number of iterations;

[0143] Step 3: In each iteration, the attack behavior and migration behavior of the black-winged kite are simulated, and after one iteration, the fitness of all black-winged kite individuals is recalculated, and the black-winged kite individual with the best fitness is selected as the new leader; wherein the fitness can preferably be the maximum fitness or the minimum fitness;

[0144] Step 4: Determine whether the current number of iterations reaches the maximum number of iterations. If so, end the iteration and output the flight path and fitness of the new leader. If not, repeat step 3 until the current number of iterations reaches the maximum number of iterations.

[0145] In an optional embodiment, the expression for simulating the attack behavior of a black-winged kite is as follows:

[0146]

[0147] in, and They represent the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, r is a random number ranging from 0 to 1, p is a constant with a value of 0.9, T is the maximum number of iterations, and t is the current number of iterations.

[0148] In an optional embodiment, the expression for simulating the migration behavior of black-winged kites is as follows:

[0149]

[0150] m = 2 × sin (r + π / 2);

[0151] in, represents the latest leader of the j-th black-winged kite at the t-th iteration so far, and denote the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, respectively. i Represents the fitness value of any individual in the current population, F ri represents the fitness value of the random population, r is a random number ranging from 0 to 1, and C(0,1) represents the Cauchy mutation.

[0152] In an optional implementation, the path planning module 34 is further configured to:

[0153] For each of the drones, determining a starting position and a target position of the flight according to the flight path;

[0154] Generate a conflict-free path for each of the drones using a depth-first search algorithm according to the environmental prediction data set, the starting position, and the target position;

[0155] According to the conflict-free paths of all the drones, the optimal path of the drone group is obtained.

[0156] In a third aspect, an embodiment of the present invention provides an electronic device, see Figure 4 , which is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention.

[0157] like Figure 3 As shown, the device includes:

[0158] A memory 31, used for storing computer programs;

[0159] A processor 32, configured to execute the computer program;

[0160] Wherein, when the processor 32 executes the computer program, the drone group path planning method as described in any of the above embodiments is implemented.

[0161] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 32 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0162] The processor 32 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0163] The memory 31 can be used to store the computer program and / or module, and the processor 32 realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory 31, and calling the data stored in the memory 31. The memory 31 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory 31 can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0164] It should be noted that the above electronic device includes, but is not limited to, a processor and a memory. Those skilled in the art can understand that Figure 4 The structural diagram is merely an example of the electronic device described above and does not constitute a limitation on the electronic device, and may include more components than shown in the figure, or a combination of certain components, or different components.

[0165] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed, the drone swarm path planning method described in any of the above embodiments is implemented.

[0166] It should be understood that the present invention implements all or part of the processes in the above-mentioned drone group path planning method, and can also be completed by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned drone group path planning method can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0167] The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. It should be pointed out that for those skilled in the art, several equivalent obvious variations and / or equivalent substitutions can be made without departing from the technical principles of the present invention. These obvious variations and / or equivalent substitutions should also be regarded as the protection scope of the present invention.

Claims

1. A method for planning a path for a drone swarm, characterized in that: include: Obtain environmental data of the environment in which the drone swarm is located within a preset time period, and construct a training set based on the environmental data; Inputting the training set into a pre-trained environment prediction model, and outputting an environment prediction data set; Using the Black Kite algorithm, the flight path of a single UAV is obtained according to the environmental prediction data set; According to the flight paths of all drones in the drone swarm and the environmental prediction data set, a MAPF path planning algorithm is used to adjust the path to obtain the optimal path of the drone swarm.

2. The UAV swarm path planning method according to claim 1, characterized in that: Before inputting the training set into the pre-trained environment prediction model, the method further includes: The missing values ​​in the training set are supplemented, and the supplemented training set is normalized.

3. The UAV swarm path planning method according to claim 1, characterized in that: The environment prediction model is a BiTCN-BiGRU-Attention network model, then, inputting the training set into a pre-trained environment prediction model and outputting an environment prediction data set includes: Inputting the training set into the BiTCN-BiGRU-Attention network model; The BiTCN layer is used to perform forward convolution calculation on the training set, extract forward data features, and send the extracted information to the BiGRU layer; The BiGRU layer is used to perform forward GRU and reverse GRU processing on the data received from the BiTCN layer, and learn the dynamic changes of data from two directions; The Attention layer assigns different weights to the input data by training weights, thereby increasing the accuracy of the prediction; According to the prediction result output by the Attention layer, an environment prediction data set is obtained.

4. The UAV swarm path planning method according to claim 1, characterized in that: The method of using the Black Kite algorithm to obtain the flight path of a single UAV according to the environmental prediction data set includes: Step 1: Initializing a black-winged kite population based on the environmental prediction data set to obtain a flight path of each black-winged kite individual in the black-winged kite population; Step 2: Calculate the fitness of each black-winged kite individual according to a preset fitness function, select the black-winged kite individual with the best fitness as the initial leader, and set the maximum number of iterations; Step 3: In each iteration, the attack behavior and migration behavior of the black-winged kite are simulated, and after one iteration, the fitness of all black-winged kite individuals is recalculated, and the black-winged kite individual with the best fitness is selected as the new leader; wherein the fitness can preferably be the maximum fitness or the minimum fitness; Step 4: Determine whether the current number of iterations reaches the maximum number of iterations. If so, end the iteration and output the flight path and fitness of the new leader. If not, repeat step 3 until the current number of iterations reaches the maximum number of iterations.

5. The UAV swarm path planning method according to claim 4, characterized in that: The expression for simulating the attack behavior of the black-winged kite is as follows: in, and They represent the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, r is a random number ranging from 0 to 1, p is a constant with a value of 0.9, T is the maximum number of iterations, and t is the current number of iterations.

6. The UAV swarm path planning method according to claim 4, characterized in that: The expression for simulating the migration behavior of black-winged kites is as follows: m = 2 × sin (r + π / 2); in, represents the latest leader of the j-th black-winged kite at the t-th iteration so far, and denote the position of the i-th black-winged kite in the j-th dimension in the t-th and (t+1)-th iterations, respectively. i Represents the fitness value of any individual in the current population, F ri represents the fitness value of the random population, r is a random number ranging from 0 to 1, and C(0,1) represents the Cauchy mutation.

7. The UAV swarm path planning method according to claim 1, characterized in that: The method of adjusting the path using the MAPF path planning algorithm according to the flight paths of all drones in the drone group and the environmental prediction data set to obtain the optimal path of the drone group includes: For each of the drones, determining a starting position and a target position of the flight according to the flight path; Generate a conflict-free path for each of the drones using a depth-first search algorithm according to the environmental prediction data set, the starting position, and the target position; According to the conflict-free paths of all the drones, the optimal path of the drone group is obtained.

8. A drone group path planning device, characterized in that: include: An environmental data acquisition module is used to acquire environmental data of the environment in which the drone group is located within a preset time period, and to construct a training set based on the environmental data; An environmental data prediction module, used to input the training set into a pre-trained environmental prediction model and output an environmental prediction data set; A black kite algorithm module, used to obtain a flight path of a single UAV based on the environmental prediction data set using a black kite algorithm; The path planning module is used to adjust the path according to the flight paths of all drones in the drone group and the environmental prediction data set by using the MAPF path planning algorithm to obtain the optimal path of the drone group.

9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to execute the computer program; Wherein, when the processor executes the computer program, the drone swarm path planning method as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the drone swarm path planning method according to any one of claims 1 to 7 is implemented.

Images