A GEO-assisted UAV cluster grouping method

Abstract

The present invention proposes a GEO-assisted drone cluster grouping method, which introduces the GEO concept into the exploration process of drone cluster grouping. By continuously exploring the distribution of the entire node drone, more effective node drone grouping and the corresponding relay drone position are found, which reduces the sum of the distances from the node drone to its relay drone, thereby reducing the size of the energy loss function of drone cluster communication, thereby achieving lower drone cluster energy loss.

Description

Technical Field

[0001] The present invention relates to the technical field of drone clusters, and in particular to a GEO-assisted drone cluster grouping method. Background Art

[0002] In recent years, drone swarm technology has been widely used in many fields, especially in complex and dynamic environments. Through networking and communication command between drones, drone swarms can achieve collaborative cooperation and complete more challenging tasks. Its applications cover rescue, environmental monitoring, agriculture, logistics and transportation, etc. However, due to the large number of drones in a drone swarm and the wide flight area, how to maintain efficient communication within the swarm has become a key challenge for drone swarm technology.

[0003] In the environment of drone swarm communication, the centralized communication mode (i.e. all drones communicate directly with the command center) will face problems such as channel congestion, excessive energy consumption, and large communication delay. Therefore, dynamic networking technology is currently widely used to reasonably group drone swarms and add drones for relay communication as relay nodes to solve problems such as channel congestion.

[0004] In dynamic networking, the entire drone cluster is divided into several local communication groups, and the location of the relay communication drone is determined during the grouping process. Each group communicates with the command center through the relay drone. In the entire communication network, the grouping of the drone cluster and the location of the relay drone determine the communication burden and energy consumption of the entire drone cluster. Unreasonable grouping methods and the location of relay drones will cause some groups to have a serious communication burden and increased energy consumption, affecting the working life of the entire network. However, the existing grouping method does not fully consider the distribution of drones, and the overall distribution of drones is not explored enough, resulting in serious communication load and increased energy consumption. Summary of the invention

[0005] In order to solve the problem that the existing drone grouping methods are insufficient in exploring the distribution of drones in a cluster and achieve lower energy consumption and more efficient communication for drone clusters, the present invention provides a drone cluster grouping method assisted by GEO (Golden Eagle Optimizer). This method can effectively explore the distribution of drones in the entire cluster, thereby obtaining a more complete drone cluster grouping method and reducing the energy consumption of drone clusters.

[0006] The technical solution of the present invention is:

[0007] The GEO-assisted drone cluster grouping method comprises the following steps:

[0008] Step 1: Initialize the initial coordinate space of the drone cluster and the node drone coordinate position set .in is the node drone position set, For the The location of the node drone, , is the number of node drones.

[0009] Step 2: Initialize the number of groups K and the initial position set of K relay drones , For the The initial position of the relay drones, , and use it as the initial value of the best memory position set of the relay drone , For the The initial value of the best memory position of each relay drone; the superscript 0 in the upper right corner indicates the 0th exploration process of drone grouping, that is, the initial state. The superscript in the upper right corner in the following formula also indicates the number of exploration iterations of drone grouping. The following formulas have the same meaning unless otherwise specified.

[0010] Step 3: Set the number of exploration iterations T for the node drone distribution and determine the attack parameters for exploration With cruise parameters .

[0011] Step 4: The best memory position set of the relay drone obtained after the t-1th iteration Perform random swap operations on the elements in the set to obtain The target position set of the iteration , For the In the iteration The target location of the relay drones, is the best memory position set after the t-1th iteration A function that randomly arranges the elements in a collection, generating a set of randomly arranged values, for example: , among which , , is the position coordinate, then The possible results are , randomly arrange the order of the position coordinates without changing their coordinate values; For the In the iteration, the set of target locations that the relay drone needs to explore.

[0012] Step 5: The relay drone is in The attack vector in the iteration is expressed as A q t → = [ a q , 1 t , a q , 2 t , ⋯ , a q , j t , ⋯ , a q , n t ] , Attack vector The jth element in is the coordinate dimension of the drone position, such as: if the space where the drone cluster is located is a two-dimensional space, then is 2, and in three dimensions is 3; according to The current location of the relay drone, i.e. The position of the relay drone after the t-1th iteration and The target location that the relay drone is about to explore , calculate the Attack vectors for relay drones:

[0013]

[0014] In the formula For the The attack vector of the relay drone is given by The current location of the relay drone points to The target location explored by the relay drone is calculated. The attack vectors of the remaining relay drones are also calculated in the same way.

[0015] Step 6: Since the cruise vector exists in the hyperplane with the relay drone attack vector as the normal and containing the current position of the relay drone, The relay drone is in The hyperplane equation where the cruise vector in the iteration can be expressed as:

[0016]

[0017] In the formula is the normal vector of the hyperplane, i.e., the corresponding component of the relay drone attack vector, is the variable of the hyperplane equation. As a constant, by The current location of the relay drones Substituting the variables into the hyperplane equation into the following formula yields:

[0018]

[0019] So far, The hyperplane equation of the cruise vector of the relay drone is obtained. The next step is to solve the first The cruise vector of a relay drone.

[0020] Step 7: The cruise vector of a relay UAV is expressed as , For cruise vector The jth element in is the coordinate dimension of the drone's position; arbitrarily select an attack vector component that is not 0 Corresponding is the quantity to be determined, and the other components of the cruise vector are randomly selected to obtain The cruise vector of the relay drone; According to the formula

[0021]

[0022] get.

[0023] Step 8: According to The attack vector and cruise vector calculation of the relay drone The moving step length of the relay drone .

[0024] Step 9: According to the moving step length , update Relay drone locations:

[0025]

[0026] Step 10: Update the best memory position set of the relay drone using the following steps:

[0027] According to The best memory position set of relay drones obtained after iterations , and the relay drone is currently The position set at the iteration , get the initial value of the preset position set of the relay drone ,in No. The element value is , For the After the iteration, the The best memory position of the relay drone, For the Relay drones are currently The initial value of the preset position set has a total of possibilities.

[0028] For each preset position set initial value , through the following process to complete a node drone grouping, and calculate the local optimal relay drone position set under the initial value of the current preset position set, and calculate the size of the loss function, the process is as follows:

[0029] (a) For each node drone, assign it to the preset position closest to the current preset position set:

[0030]

[0031] in For the The location of the node drone, For the The node drone position is the closest to the preset position. Indicates the preset position The corresponding grouping, Represents the remaining preset positions in the preset position set;

[0032] The obtained grouping result is recorded as .in Record No. The node drone number within a group.

[0033] (b) For each group, recalculate the preset position corresponding to the group:

[0034]

[0035] For the The location of each node drone;

[0036] (c) Repeat (a) and (b) until the preset position set converges, and the obtained preset position set convergence value is used as the initial value based on a certain preset position set The local optimal relay drone location set , for The elements; Indicates the number of repetitions of process a and b. And calculate The energy loss function is defined as:

[0037]

[0038] for The above process is repeated for each of the preset position sets to obtain The local optimal relay drone position set is updated to the best memory position set of the relay drone: A local optimal relay drone location set with the minimum energy loss function among the local optimal relay drone location sets , that is, The best memory position set of relay drones after iterations , and save the corresponding node drone grouping. The best memory position set of the relay drones in the iteration is updated, and the corresponding drone cluster grouping is obtained.

[0039] Step 11: Repeat steps 4 to 10 according to the set number of iterations, and output the best memory position set of relay drones and the corresponding node drone grouping obtained in the last iteration to realize drone cluster grouping.

[0040] Furthermore, in step 2, the initial position of the relay drone Follow these steps to determine:

[0041] Step 2.1: Randomly select a node drone initial position as the first relay drone initial position ;

[0042] Step 2.2: For each node drone initial position, calculate the sum of its distances to the initial positions of all currently determined relay drones , For the The location of the node drone, is the number of initial positions of relay drones currently determined, symbol is the 2-norm; the initial position of the node drone corresponding to the maximum sum of distances is selected as the initial position of the next relay drone;

[0043] Step 2.3: Repeat step 2.2 until the initial positions of K relay drones are determined.

[0044] Furthermore, in step 3, the attack parameters With cruise parameters According to the formula

[0045]

[0046] Calculated, represents the first iteration of the relay drone exploration Iterations; is the initial value of the iteration of the attack coefficient, is the final value of the iteration of the attack coefficient, is the iterative initial value of the cruise coefficient, is the iterative final value of the cruise coefficient, , , , All are set values, which means that during the entire exploration process of the node UAV distribution, the tendency gradually changes from the cruise tendency to the attack tendency, and finally completes the exploration.

[0047] Furthermore, in step 8, The moving step length of the relay drone is calculated according to the formula

[0048]

[0049] Calculated, where: and for [ 0 , 1 ] The random length between .

[0050] Beneficial Effects

[0051] The present invention introduces the GEO concept into the exploration process of drone cluster grouping. By continuously exploring the distribution of the entire node drone, more effective node drone grouping and the corresponding relay drone position are found, which reduces the sum of the distances from the node drone to its relay drone, thereby reducing the size of the energy loss function of drone cluster communication, thereby achieving lower drone cluster energy loss.

[0052] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0054] Figure 1 It is a flowchart of the present invention, which shows the process of the whole method. First, the drone cluster is initialized, including establishing the node drone coordinate space and position (step 1), the number and position of the relay drone (step 2), and the number of explorations and exploration parameters of the relay drone (step 3). Then the relay drone performs random optimal memory exchange (step 4), and calculates the attack vector and cruise vector of the relay drone (step 5)~(step 7), and calculates the moving step of the relay drone according to the attack vector and the cruise vector and then updates the position of the relay drone (step 8)~(step 9), and updates the best memory of the relay drone according to the explored position (step 10), obtains the node drone grouping of this round and the position of the relay drone in each group, and then enters the exploration process of the next round of grouping. After the whole exploration process is completed, the final node drone grouping and the position result of the relay drone in each group are output.

[0055] Figure 2 This is a schematic diagram of the initial point selection of the relay drone of the present invention. Taking the 2D drone cluster distribution as an example, the node drone position is represented by a blue □, and the initial position selection of the relay drone refers to the selection step in (step 2). The initial position of the relay drone is represented by a green express.

[0056] Figure 3 is a schematic diagram of the relay drone exploration of the present invention, based on Figure 1 The initial position of the relay drone, referring to the steps in (step 4) to (step 10), shows the exploration process and movement of the relay drone. The node drone position is represented by a blue □, and the current position of the relay drone is represented by a green Indicates that the attack vector of the relay drone The relay drone's current position points to the target position to be explored, with a red arrow indicating its direction. Another cruise vector of the relay drone can be solved by the hyperplane equation, with a pink arrow indicating its direction in the figure. The relay drone's movement compensation is based on its attack vector and cruise vector, with a blue arrow indicating its movement direction and step size in the figure. Finally, the position to which the relay drone moves is indicated by a blue arrow. express.

[0057] Figure 4 It is a general schematic diagram of the Monte Carlo simulation of the present invention, which shows the comparison of the drone cluster energy loss obtained by the present invention and the drone cluster energy loss function value obtained by the K-Means++ method based on the same node drone position and initial value selection. In the experiment, the performance of the two different methods was tested by randomly generating 1000 groups of different node drone positions. The x-axis shows the number of Monte Carlo simulations, and the y-axis shows the energy loss function value of the drone cluster grouping result. The red curve in the figure is the energy loss function value brought by the method of the present invention, and the blue curve is the energy loss function value brought by the K-Means++ method.

[0058] Figure 5 is a detailed schematic diagram of the Monte Carlo simulation of the present invention, which shows Figure 4 The detailed results of the 500th to 550th experiments, a total of 50 experiments, are compared. The x-axis is the number of experiments in these 50 groups of experiments, and the y-axis is the loss function of the results. The red curve in the figure is the energy loss function value brought by the method of the present invention, and the blue curve is the energy loss function value brought by the K-Means++ method. DETAILED DESCRIPTION

[0059] Embodiments of the present invention are described in detail below. The embodiments are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0060] The GEO-assisted UAV cluster grouping method proposed in this invention combines the powerful optimization advantages of the GEO algorithm to explore the UAV cluster. Figure 1 As shown in the figure, combined with the core idea of ​​GEO, the attack vector, cruise vector and random memory exchange are integrated into the group exploration, which reduces the energy loss function value of the drone cluster and improves the exploration ability.

[0061] In the following description, the drone cluster is divided into two parts. One part is the relay drone, which acts as a relay node in each group to communicate with the command center; the rest are node drones, which are divided into different groups to communicate with the relay drones.

[0062] The specific steps in this embodiment are:

[0063] Step 1: Initialize the initial coordinate space of the drone cluster and the node drone position set ;in is the node drone position set, For the The location of the node drone, , is the number of node drones.

[0064] like Figure 2 As shown, in this embodiment, the drone cluster distribution is a two-dimensional spatial distribution, so the spatial dimension , the number of node drones is 20, initialize the initial coordinate space of the drone cluster and the node drone coordinate position set , which is located in Figure 2 Indicated by blue □.

[0065] Step 2: Initialize the number of groups K = 5 and the initial position set of 5 relay drones , and use it as the initial value of the best memory position set of the relay drone , which is located in Figure 2 Green The upper right corner indicates the number of iterations of exploring drone groups.

[0066] The initial position of the relay drone Follow these steps to determine:

[0067] Step 2.1: Randomly select a node drone initial position as the first relay drone initial position ;

[0068] Step 2.2: For each node drone initial position, calculate the sum of its distances to the initial positions of all currently determined relay drones , For the The location of the node drone, is the number of initial positions of relay drones currently determined, symbol is the 2-norm; the initial position of the node drone corresponding to the maximum sum of distances is selected as the initial position of the next relay drone;

[0069] Step 2.3: Repeat step 2.2 until the initial positions of the five relay drones are determined.

[0070] Step 3: Set the number of exploration iterations of the node drone distribution to T = 50, and determine the attack parameters and cruise parameters for each iteration;

[0071] Among them, Attack parameters for iterations With cruise parameters According to the formula

[0072]

[0073] Calculated; is the initial value of the iteration of the attack coefficient, is the final value of the iteration of the attack coefficient, is the iterative initial value of the cruise coefficient, is the iterative final value of the cruise coefficient, , , , All are set values. [ p a 0 , p a T ] = [ 0 . 25 , 1 ] , [ p c 0 , p c T ] = [ 0 . 5 , 0 . 25 ] , which indicates that the tendency in the entire exploration process of the node UAV distribution gradually changes from a cruising tendency to an attack tendency, and finally completes the exploration.

[0074] Step 4: The best memory position set of the relay drone obtained after the t-1th iteration Perform random swap operations on the elements in the set to obtain The target position set of the iteration , is the best memory position set after the t-1th iteration A function that randomly arranges the elements in a collection, generating a set of randomly arranged values, for example: , among which , , , , is the position coordinate, then The possible results are , randomly arrange the order of the position coordinates without changing their coordinate values; For the In the iteration, the set of target locations that the relay drone needs to explore.

[0075] Step 5: In In the iteration, for the A relay drone, according to Relay drone current location and Target location of relay drones , calculate the Attack vectors for relay drones :

[0076]

[0077] No. Attack vectors for relay drones Expressed as A q t → = [ a q , 1 t , a q , 2 t ] , whose direction is determined by The current location of the relay drone points to The target location explored by the relay drones is Figure 3 Indicated by red arrows.

[0078] Step 6: In the iteration, according to the formula

[0079]

[0080] calculate ; For the first The attack vector of the relay drone is normal and contains the The constant of the hyperplane equation of the current position of the relay drone, For the Relay drone current location The j-th element of .

[0081] Step 7: In the iteration, The cruise vector of a relay UAV is expressed as C q t → = [ c q , 1 t , c q , 2 t ] , arbitrarily select a non-zero attack vector component Corresponding is the quantity to be determined, and the other components of the cruise vector are randomly selected to obtain The cruise vector of the relay drone is Figure 3 Indicated by pink arrows; According to the formula

[0082]

[0083] get.

[0084] Step 8: In the iteration, according to The attack vector and cruise vector calculation of the relay drone The moving step length of the relay drone :

[0085]

[0086] Calculated, and Both [ 0 , 1 ] The random length between .

[0087] Step 9: In the iteration, according to the moving step length , update The location of the relay drones, Figure 3 Blue express:

[0088]

[0089] Step 10: In the iteration, the following steps are used to update the best memory position set of the relay drone:

[0090] According to The best memory position set of relay drones obtained after iterations , and the relay drone is currently The position set at the iteration , get the initial value of the preset position set of the relay drone ,in No. The element value is ; The initial value of the preset position set has a total of possibilities;

[0091] For each preset position set initial value , the node drone grouping is completed once through the following process, and the local optimal relay drone position set under the initial value of the current preset position set is calculated, as well as the size of the loss function:

[0092] a. For each node drone, assign it to the nearest preset position in the current preset position set:

[0093]

[0094] in For the The location of the node drone, For the The node drone position is the closest to the preset position. Indicates the preset position The corresponding grouping, represents the remaining preset positions in the preset position set; the obtained grouping result is recorded as ,in Record No. The node drone number within each group;

[0095] b. For each group, recalculate the preset position corresponding to the group:

[0096]

[0097] For the The location of each node drone;

[0098] c. Repeat steps a and b until the preset position set converges, and the obtained preset position set convergence value is used as the initial value based on a certain preset position set. The local optimal relay drone location set , for The elements; Indicates the number of repetitions of process a and b; and according to the formula

[0099]

[0100] calculate The energy loss function ;

[0101] for The initial value of the preset position set is obtained The local optimal relay drone position set is updated to the best memory position set of the relay drone: A local optimal relay drone location set with the minimum energy loss function among the local optimal relay drone location sets , get the The best memory position set of relay drones after iterations .

[0102] Step 11: Repeat steps 4 to 10 according to the set number of iterations, and output the best memory position set of relay drones and the corresponding node drone grouping obtained in the last iteration to realize drone cluster grouping.

[0103] Performance Analysis:

[0104] In the above example, a randomly generated 2D drone cluster is used for testing. The attack parameters and cruise parameters are selected as [ p a 0 , p a T ] = [ 0 . 25 , 1 ] , [ p c 0 , p c T ] = [ 0 . 5 , 0 . 25 ] , the number of groups is specified as 5, the number of node drones is 20, and the total number of Monte Carlo simulations is 1000. Figure 4 , Figure 5 As shown in the figure, from the overall Monte Carlo simulation experiment results, the result achieved by the grouping method proposed in the present invention has a lower UAV cluster energy loss function value. From the more detailed Monte Carlo simulation experiment results, if the K-Means++ grouping method has achieved the best deployment effect, the grouping method proposed in the present invention can achieve the same result; if not, the UAV grouping result of the present invention can achieve a lower energy loss function value, have stronger exploration capabilities and lower UAV cluster energy loss.

[0105] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and purpose of the present invention.

Claims

1. A GEO-assisted UAV cluster grouping method, characterized by: The following steps are involved: Step 1: Initialize the initial coordinate space of the drone cluster and the node drone position set ;in is the node drone position set, For the The location of the node drone, , is the number of node drones; Step 2: Initialize the number of groups K and the initial position set of K relay drones , For the The initial position of the relay drones, , and use it as the initial value of the best memory position set of the relay drone , For the The initial value of the best memory position of the relay drones. The upper right corner superscript indicates the number of iterations of exploring drone grouping. Step 3: Set the number of exploration iterations T for the node drone distribution and determine the attack parameters and cruise parameters for each iteration; Step 4: In In the iteration, the best memory position set obtained after the t-1th iteration Perform random swap operations on the elements in the set to obtain The target position set of the relay drones in the iteration , For the In the iteration The target location of the relay drone; Step 5: In In the iteration, for the A relay drone, according to Relay drone current location and Target location of relay drones , calculate the Attack vectors for relay drones : No. Attack vectors for relay drones Expressed as , is the coordinate dimension of the drone’s position; Attack vector The jth element in ; Step 6: In the iteration, according to the formula calculate ; For the first The attack vector of the relay drone is normal and contains the The constant of the hyperplane equation of the current position of the relay drone, For the Relay drone current location The jth element of ; Step 7: In the iteration, The cruise vector of a relay UAV is expressed as , For cruise vector The jth element in; arbitrarily select an attack vector component that is not 0 Corresponding is the quantity to be determined, and the other components of the cruise vector are randomly selected to obtain The cruise vector of the relay drone; According to the formula get; Step 8: In the iteration, according to The attack vector and cruise vector calculation of the relay drone The moving step length of the relay drone ; Step 9: In the iteration, according to the moving step length , update The location of the relay drone Step 10: In the iteration, the following steps are used to update the best memory position set of the relay drone: According to The best memory position set of relay drones obtained after iterations , and the relay drone is currently The position set at the iteration , get the initial value of the preset position set of the relay drone ,in No. The element value is , For the After the iteration, the The best memory position of the relay drone, For the Relay drones are currently The position at the time of the iteration; the initial value of the preset position set has a total of possibilities; For each preset position set initial value , the node drone grouping is completed once through the following process, and the local optimal relay drone position set under the initial value of the current preset position set is calculated, as well as the size of the loss function: a. For each node drone, assign it to the nearest preset position in the current preset position set: in For the The location of the node drone, For the The node drone position is the closest to the preset position. Indicates the preset position The corresponding grouping, Represents the remaining preset positions in the preset position set; The obtained grouping result is recorded as ,in Record No. The node drone number within each group; b. For each group, recalculate the preset position corresponding to the group: For the The location of each node drone; c. Repeat steps a and b until the preset position set converges, and the obtained preset position set convergence value is used as the initial value based on a certain preset position set. The local optimal relay drone location set , for The elements; Indicates the number of repetitions of processes a and b; And according to the formula calculate The energy loss function ; for The initial value of the preset position set is obtained The local optimal relay drone position set is updated to the best memory position set of the relay drone: A local optimal relay drone location set with the minimum energy loss function among the local optimal relay drone location sets , get the The best memory position set of the relay drone after iterations ; Step 11: Repeat steps 4 to 10 according to the set number of iterations, and output the best memory position set of relay drones and the corresponding node drone grouping obtained in the last iteration to realize drone cluster grouping.

2. A GEO-assisted drone cluster grouping method according to claim 1, characterized in that: In step 2, the initial position of the relay drone Follow these steps to determine: Step 2.1: Randomly select a node drone initial position as the first relay drone initial position ; Step 2.2: For each node drone initial position, calculate the sum of its distances to the initial positions of all currently determined relay drones , For the The location of the node drone, is the number of initial positions of relay drones currently determined, symbol is the 2-norm; the initial position of the node drone corresponding to the maximum sum of distances is selected as the initial position of the next relay drone; Step 2.3: Repeat step 2.2 until the initial positions of K relay drones are determined.

3. A GEO-assisted UAV cluster grouping method according to claim 1, characterized in that: In step 3, Attack parameters for iterations With cruise parameters According to the formula Calculated; is the initial value of the iteration of the attack coefficient, is the final value of the iteration of the attack coefficient, is the iterative initial value of the cruise coefficient, is the iterative final value of the cruise coefficient, , , , All are set values.

4. The GEO-assisted UAV cluster grouping method according to claim 1, characterized in that: In step 8, The moving step length of the relay drone is calculated according to the formula Calculated, and Both The random length between .

Images