Manned aerial vehicle / unmanned aerial vehicle co-fusion topology communication method based on pigeon flock interaction intelligence

Abstract

The invention discloses a manned aerial vehicle / unmanned aerial vehicle co-fusion topology communication method based on pigeon flock interaction intelligence. The method comprises the implementation steps that 1, pigeon flock homing flight data instantaneous correlation modeling is carried out; 2, modeling the alignment degree of the homing flight data of the pigeon flock; 3, dove homing flight interaction topology modeling is carried out; 4, designing a full-connection neural network structure; 5, carrying out dynamics modeling on the manned aerial vehicle / unmanned aerial vehicle; 6, a manned aircraft / unmanned aerial vehicle acting force model based on pigeon flock intelligence is established; 7, manned aerial vehicle / unmanned aerial vehicle communication control based on event triggering; and 8, outputting a manned aerial vehicle / unmanned aerial vehicle dynamic communication topology based on pigeon flock interaction intelligence. The manned aerial vehicle / unmanned aerial vehicle variable topology communication method based on pigeon flock interaction intelligence improves the communication efficiency and stability of a manned aerial vehicle / unmanned aerial vehicle cluster task, thereby effectively improving the overall efficiency of the autonomous task of the manned aerial vehicle / unmanned aerial vehicle cluster.

Description

Technical Field

[0001] The present invention relates to a manned-machine / unmanned-machine co-integrated variable-topology communication method based on pigeon group interactive intelligence, and belongs to the field of autonomous control of unmanned-machine clusters. Background Art

[0002] In the context of manned / unmanned aerial vehicle swarms performing combat missions such as search, tracking, and positioning, when the swarm approaches the combat area, the manned / unmanned aerial vehicle swarm needs to conduct targeted search, encirclement, or attack on the target point, and make timely avoidance of enemy interference, weapon attacks, etc. in the combat environment. In a highly dynamic and confrontational combat environment, the comprehensive decision-making ability of manned aircraft should be brought into play, combined with the distributed autonomous control mode of the unmanned aerial vehicle swarm, to improve the combat effectiveness of manned / unmanned aerial vehicle swarms. In order to achieve efficient and stable communication between manned aircraft and unmanned aerial vehicles in a dynamic environment, it is necessary to construct a manned / unmanned aerial vehicle swarm formation communication method based on an effective and appropriate motion model to support the realization of manned / unmanned aerial vehicle swarm formation tasks in unknown and complex environments.

[0003] Faced with increasingly complex tasks and environments, manned / unmanned aerial vehicle clusters must have a flexible communication system to deal with emergencies. Existing manned / unmanned aerial vehicle cluster communication networks can be divided into two categories: centralized and distributed. Among them, centralized communication networks can realize global perception of cluster individuals and environmental situation information, which can effectively improve the decision-making effectiveness of manned and unmanned aerial vehicles. However, at the same time, it greatly increases the communication load of the cluster, posing a severe challenge to the robustness of the airborne communication terminal; the distributed local interaction topology realizes information transmission through the interaction between individuals in the cluster and neighboring individuals, which reduces the cluster's requirements for communication robustness, but local interaction has a large delay in the transmission of information, which may cause the cluster to be unable to effectively execute manned aircraft command messages. Therefore, inventing a communication method that can reduce the cluster communication load and ensure the effective transmission of instructions is a basic prerequisite for realizing manned / unmanned aerial vehicle cluster tasks.

[0004] The common size of pigeon flocks in nature ranges from a few to dozens of pigeons. It is a common and spectacular behavior for pigeons to circle in groups near pigeon nests. The direction of rotation changes regularly and spontaneously, and the flock moves in coordination through effective and robust interactions between individuals. Analysis of high-resolution flight data of pigeon flocks found that in small-scale bird flocks, individuals cooperate with different numbers of neighbors in the metric space and tend to interact with time-varying neighbors with close distances rather than with a fixed number of topological neighbors. Biological behavior analysis shows that there is an intermittent interaction mechanism in pigeon flocks, and individuals can form a jointly connected topological network structure without the need for intra-group information interaction at all times, which helps reduce communication or information processing requirements while still maintaining the group's responsiveness and stability.

[0005] Under the influence of natural evolutionary selection, the harsh living environment has prompted pigeon flocks to develop a self-organizing interaction mechanism with the advantages of adaptability, intermittency, and stability, which is consistent with the communication requirements of manned / unmanned aerial vehicle cluster tasks such as self-organization, low frequency, and robustness.

[0006] Based on the above-mentioned prior art, the present invention proposes a manned-machine / unmanned aerial vehicle integrated variable-topology communication method based on pigeon flock interactive intelligence to solve the challenges of high communication frequency, variable-topology communication interruption, etc. in the existing manned-machine / unmanned aerial vehicle cluster missions, promote the improvement of communication efficiency and stability of manned-machine / unmanned aerial vehicle cluster missions, and effectively improve the overall performance of manned-machine / unmanned aerial vehicle cluster autonomous missions. Summary of the invention

[0007] 1. Purpose of the invention:

[0008] The purpose of the present invention is to provide a manned-machine / unmanned-machine co-integrated variable-topology communication method based on pigeon flock interactive intelligence, aiming to improve the communication efficiency and stability of manned-machine / unmanned-machine cluster tasks, thereby effectively improving the overall effectiveness of manned-machine / unmanned-machine cluster autonomous tasks.

[0009] 2. Technical solution:

[0010] In view of the problems of high communication frequency and topology-changing communication interruption of autonomous tasks of manned / unmanned aerial vehicle clusters, the present invention invents a manned / unmanned aerial vehicle co-integrated topology-changing communication method based on pigeon flock interactive intelligence. The implementation steps are as follows:

[0011] Step 1: Modeling the instantaneous correlation of pigeon flock homing flight data

[0012] First, we select the T collective turning stage data (here T is 20 seconds) of the pigeons circling around the pigeon nest in an orderly manner. The data sampling time t s The speed fluctuation of a single pigeon is 0.2 seconds. The influence of the speed fluctuation of a single pigeon on other pigeons is studied. The speed fluctuation (taking both amplitude and direction into account) is defined as the difference between the speed of pigeon i and the average speed of the pigeon group:

[0013]

[0014] Among them, N is the size of the pigeon group, k is the neighbor number of pigeon i, and the instantaneous correlation of the speed fluctuations of pigeon i and pigeon j can be obtained:

[0015]

[0016] Among them, τ∈[-1,1] is the delay size.

[0017] Step 2: Modeling the alignment of homing flight data

[0018] Based on the instantaneous correlation, the alignment between the two pigeons under the time delay condition can be quantified as:

[0019]

[0020] Usually, when the alignment degree P ij ≥P * = 0.92, it can be considered that there is a strong speed fluctuation correlation between pigeon i and pigeon j. In this case, pigeon i and pigeon j are considered to be aligned with each other, that is, pigeon j is a neighbor of pigeon i.

[0021] Step 3: Topological modeling of the interactive flight of pigeon flocks

[0022] Based on the alignment analysis between two pigeons, we can obtain all the interactive neighbors of pigeon i at time t, namely N i ={1,2,...,m},m<N, define the communication interaction topology network of the pigeon group at time t as Μ N×N , from which we can get:

[0023]

[0024] Thus, the communication interaction topological sequence of the pigeon flock within the homing flight time T can be obtained.

[0025] Step 4: Design of fully connected neural network structure

[0026] According to the interactive topological sequence of the homing flight of pigeons, a fully connected neural network (FCNN) structure is designed, which consists of five fully connected layers (Linear Layers). The topological network of the pigeon flock at each moment is spliced ​​into N 2 dimensional vector, input the pigeon flock topology sequence into the first fully connected first hidden layer, output 512 features after linear transformation and ReLU activation function, input the 512 features into the second fully connected layer, output 256 features after linear transformation and ReLU activation function, input the 256 features into the third fully connected layer, output 64 features after linear transformation and ReLU activation function, and input the 512 features output by the first fully connected layer into the fourth fully connected layer, output 64 features after linear transformation and ReLU activation function, concatenate the features of the third fully connected layer and the fourth fully connected layer to integrate feature information of different levels, input the fifth fully connected layer, and map the concatenated features to the original data dimension after linear transformation and ReLU activation function. The network architecture is as follows: Figure 1 shown.

[0027] Step 5: Dynamics modeling of manned / unmanned aircraft

[0028] In the process of manned / UAV cluster coordinated motion, the second-order UAV kinematic model is considered as follows:

[0029]

[0030] Among them, i∈{1,2,...,N}, N is the size of the drone cluster, u i represents the control input of drone i, p i 、v i , are the position vector, velocity vector and track angle of UAV i, v 0 is the speed of the drone. Furthermore, the communication topology of N individuals in the pigeon flock hierarchical network is mapped to the manned / drone cluster interaction topology.

[0031] Step 6: Manned / UAV force model based on pigeon swarm intelligence

[0032] Based on the pigeon flock force model, the control input of the UAV is expressed as the sum of weighted coefficients of the force, which is defined as follows:

[0033]

[0034] Among them, μ 1 , μ 2 The first term on the right side of formula (6) represents the clustering force of individual drones on the position information of neighboring individuals within the communication distance, the second term on the right side represents the clustering force of individual drones on the velocity information of neighboring individuals, and the third term on the right side represents the attraction of the desired position traction command to the individual drone. The desired command acceleration is set to zero, that is,

[0035] Step 7: Event-triggered manned / unmanned aerial vehicle communication control

[0036] An event-triggered communication mechanism is used to reduce the communication load of the unmanned cluster. The controller is designed as follows:

[0037]

[0038] in, Indicates the time when the most recent event of drone i was triggered. In order to describe the tracking error of individual drones to the expected command, it is defined as:

[0039]

[0040] The position error and velocity error after the event trigger mechanism is introduced are defined as:

[0041]

[0042] Therefore, the error vector and state vector can be expressed as

[0043] The trigger conditions are designed as follows:

[0044]

[0045] Among them, c 1 ,c 2 ,c 3 ,c 4 is a constant. Based on models (5), (6) and (7), the system dynamic equation is as follows:

[0046]

[0047] Among them, the interactive neighbor set N i It can be obtained from the output of the fully connected neural network in step 4.

[0048] Step 8: Output the manned / unmanned aerial vehicle dynamic communication topology based on pigeon flock interaction intelligence

[0049] The task flow chart of manned / unmanned aerial vehicle topology-changing communication based on pigeon flock interaction intelligence is as shown in the attached figure. Figure 2 As shown in the figure, during the execution of the mission, the manned / unmanned aerial vehicle cluster determines whether the communication is triggered at the current moment through the trigger condition. If the communication is triggered, the topological network structure generated by the fully connected neural network is input. The drones exchange information according to the interactive neighbor set in the topological network and update their respective positions and speeds based on the system dynamic equation. When the time constraint is met, the mission ends and the manned / unmanned aerial vehicle dynamic communication topological sequence based on the pigeon flock interactive intelligence is output.

[0050] 3. Advantages and effects:

[0051] The present invention discloses a manned-machine / unmanned-machine co-integrated variable topology communication method based on pigeon flock interactive intelligence. The method takes into account the communication limitation problem in complex environment, studies the intermittent interactive topological network structure of pigeon flocks in the process of circling while homing, realizes the generation of joint connected communication interactive topology through a fully connected neural network, applies the intermittent interactive topological network based on pigeon flock intelligence to the manned-machine / unmanned-machine cluster communication topology, adopts an event triggering mechanism to reduce the communication load, thereby reducing the computational cost of manned-machine / unmanned-machine cluster communication and information processing, and achieving a balance between cluster collaborative efficiency and communication cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a fully connected neural network structure diagram.

[0053] Figure 2It is a simulation flow chart of variable topology communication between manned and unmanned aircraft based on pigeon flock interactive intelligence.

[0054] Figure 3 It is the cost function curve of pigeon flock communication topology prediction based on fully connected neural network.

[0055] Figure 4 It is a statistical chart of the number of times the obstacle avoidance communication of manned / unmanned aerial vehicle clusters is triggered.

[0056] The numbers and symbols in the figure are explained as follows:

[0057] t s ——Pigeon flock data sampling time;

[0058] N – cluster size;

[0059] U ij - instantaneous correlation;

[0060] P ij - alignment;

[0061] Μ——communication topology network;

[0062] u i ——Control input of drone i;

[0063] p i ——position vector of drone i;

[0064] v i ——velocity vector of drone i;

[0065] ——track angle of drone i;

[0066] ——Neighbor location information clustering power;

[0067] ——Neighborhood speed information clustering power;

[0068] — expected location attractiveness; DETAILED DESCRIPTION

[0069] The effectiveness of the communication method proposed in the present invention is verified by a specific example of manned / unmanned aerial vehicle integrated cluster formation obstacle avoidance in an obstacle environment. The experimental computer is configured with an Intel Core processor, 3.60Ghz main frequency, 32.0GB memory, and the software is MATLAB 2021 version.

[0070] Step 1: Modeling the instantaneous correlation of pigeon flock homing flight data

[0071] The data of the collective turning stage when the pigeons circled the pigeon nest in an orderly manner for a period of time T = 20s were selected to study the impact of the speed fluctuation of a single pigeon on other pigeons, and the difference between the speed of pigeon i and the average speed of the pigeon group, that is, the speed fluctuation u i (t), and further obtain the instantaneous correlation U of the speed fluctuations of pigeon i and pigeon j ij (t,τ).

[0072] Step 2: Modeling the alignment of homing flight data

[0073] Quantifying the alignment degree P between two pigeons under time delay conditions based on instantaneous correlation ij (t), when the alignment degree P ij ≥P * When =0.92, it is considered that there is a strong speed fluctuation correlation between pigeon i and pigeon j, and pigeon i and pigeon j are considered to be aligned with each other, that is, pigeon j is the neighbor of pigeon i.

[0074] Step 3: Topological modeling of the interactive flight of pigeon flocks

[0075] Based on the alignment analysis between two pigeons, we can obtain all the interactive neighbors of pigeon i at time t, namely N i ={1,2,...,m},m<N, so the communication interaction topology network of the entire pigeon group at time t is M N×N .

[0076] Step 4: Design of fully connected neural network structure

[0077] Design a fully connected neural network (FCNN) structure consisting of five fully connected layers (Linear Layers). Splice the pigeon flock topology network at each moment into N 2 dimensional vector, input the pigeon flock topology sequence into the first hidden layer of the first full connection, output 512 features after linear transformation and ReLU activation function, input the 512 features into the second full connection layer, output 256 features after linear transformation and ReLU activation function, input the 256 features into the third full connection layer, output 64 features after linear transformation and ReLU activation function, and input the 512 features output by the first full connection layer into the fourth full connection layer, output 64 features after linear transformation and ReLU activation function, concatenate the features of the third and fourth full connection layers to integrate feature information of different levels, input the fifth full connection layer, and map the concatenated features to the original data dimension after linear transformation and ReLU activation function, thereby generating the communication topology sequence at the next moment.

[0078] Step 5: Dynamics modeling of manned / unmanned aircraft

[0079] In this example, a second-order UAV kinematic model is considered during the coordinated motion of a manned / unmanned aerial vehicle cluster.

[0080] Step 6: Manned / UAV force model based on pigeon swarm intelligence

[0081] In this example, based on the pigeon flock force model, the control input of the UAV is expressed as the weighted coefficient sum of the force, where the parameter is defined as μ 1 =0.2, μ 2 =0.6.

[0082] Step 7: Event-triggered manned / unmanned aerial vehicle communication control

[0083] An event-triggered communication mechanism is used to reduce the communication load of unmanned clusters. The trigger conditions and controller are designed, where the parameters are defined as c 1 =0.25, c 2 =0.25, c 3 =1, c 4 =1.

[0084] Step 7: Output the manned / unmanned aerial vehicle dynamic communication topology based on pigeon flock interaction intelligence

[0085] In this example, the sampling time dt = 0.1s, the maximum running time T max =250s, 2 manned aircraft and 9 unmanned aerial vehicles are used to perform patrol missions in mountainous areas. Eight peaks are set in the mission environment, located at [10,20], [30,5], [30,80], [50,50], [65,25], [70,90], [80,60], [95,80]. During the mission, the manned aircraft has a larger radar perception range and can detect obstacles ahead earlier. It shares information with the unmanned aerial vehicle through local interaction. At the same time, the manned aircraft adjusts the flight speed forward to guide the unmanned aerial vehicle cluster to avoid obstacles accurately. The unmanned aerial vehicle adopts a communication interaction mode based on an event-triggered mechanism, and uses the flight data of the pigeon flock to train the fully connected neural network to output a joint connected communication interaction topology. The cost function value evolution curve is shown in the attached figure. Figure 3 As shown in the figure, it can be seen that the error between the topological network output by training and the real network decreases with each generation, which can effectively simulate the intermittent interaction mechanism of the pigeon flock. The number of communications of the drone within 2500 calculation cycles is statistically analyzed as shown in the attached figure. Figure 4 As shown, it can be seen that the number of communications does not exceed 700 times, so this mode effectively reduces the communication load of the cluster and ensures the effective transmission of man-machine instructions.

Claims

1. A manned / unmanned aerial vehicle co-integrated variable topology communication method based on pigeon flock interactive intelligence, characterized by: The steps of this method are as follows: Step 1: Modeling the instantaneous correlation of pigeon flock homing flight data First, we selected the data of the T-group turning stage in which the pigeons circled the pigeon nest in an orderly manner to study the impact of the speed fluctuation of a single pigeon on other pigeons. The speed fluctuation was defined as the difference between the speed of pigeon i and the average speed of the pigeon group: Among them, N is the size of the pigeon group, k is the neighbor number of pigeon i, and the instantaneous correlation of the speed fluctuations of pigeon i and pigeon j is obtained: Among them, τ∈[-1,1] is the delay size; Step 2: Modeling the alignment of homing flight data Based on the instantaneous correlation, the alignment between the two pigeons under the time delay condition can be quantified as: When the alignment degree P ij ≥P * =0.92, it is considered that there is a strong speed fluctuation correlation between pigeon i and pigeon j; pigeon i and pigeon j are considered to be aligned with each other, that is, pigeon j is the neighbor of pigeon i; Step 3: Topological modeling of the interactive flight of pigeon flocks Based on the alignment analysis between two pigeons, we can obtain all the interactive neighbors of pigeon i at time t, namely N i ={1,2,...,m},m<N, define the communication interaction topology network of the pigeon group at time t as Μ N×N , from which we get: Thus, the communication interaction topology sequence within the homing flight time T of the pigeon flock is obtained; Step 4: Design of fully connected neural network structure Aiming at the interactive topological sequence of pigeon homing flight, a fully connected neural network FCNN structure is designed, which consists of five fully connected layers; Step 5: Dynamics modeling of manned / unmanned aircraft In the process of manned / UAV cluster coordinated motion, the second-order UAV kinematic model is considered as follows: Among them, i∈{1,2,...,N}, N is the size of the drone cluster, u i represents the control input of drone i, p i 、v i , are the position vector, velocity vector and track angle of UAV i, respectively, and v0 is the velocity of the UAV; further, the communication topology relationship of N individuals in the pigeon flock hierarchical network is mapped to the manned / UAV cluster interaction topology; Step 6: Manned / UAV force model based on pigeon swarm intelligence Based on the pigeon flock force model, the control input of the UAV is expressed as the sum of weighted coefficients of the force, which is defined as follows: Among them, μ1 and μ2 are constants. The first term on the right side of formula (6) represents the clustering force of individual drones on the position information of neighbor individuals within the communication distance, the second term on the right side represents the clustering force of individual drones on the velocity information of neighbor individuals, and the third term on the right side represents the attraction of the desired position traction command to the individual drone. The desired command acceleration is set to zero, that is, Step 7: Event-triggered manned / unmanned aerial vehicle communication control An event-triggered communication mechanism is used to reduce the communication load of the unmanned cluster. The controller is designed as follows: in, represents the time when the most recent event of UAV i was triggered; in order to describe the tracking error of individual UAVs to the expected command, it is defined as: The position error and velocity error after the event trigger mechanism is introduced are defined as: Therefore, the error vector and state vector are expressed as The trigger conditions are designed as follows: Where c1, c2, c3, c4 are constants. Based on models (5), (6) and (7), the system dynamic equation is as follows: Among them, the interactive neighbor set N i Obtained from the output of the fully connected neural network in step 4; Step 8: Output the manned / unmanned aerial vehicle dynamic communication topology based on pigeon flock interaction intelligence During the execution of the mission, the manned / unmanned aerial vehicle cluster determines whether communication is triggered at the current moment through the trigger conditions. If communication is triggered, the topological network structure generated by the fully connected neural network is input. The drones exchange information according to the set of interactive neighbors in the topological network and update their respective positions and speeds based on the system dynamic equations. When the time constraint conditions are met, the mission is completed and the manned / unmanned aerial vehicle dynamic communication topological sequence based on the pigeon flock interactive intelligence is output.

2. The method according to claim 1, characterized in that: The fourth step is to splice the pigeon flock topology network at each moment into N 2 dimensional vector, input the pigeon flock topology sequence into the first hidden layer of the first full connection, output 512 features after linear transformation and ReLU activation function, input the 512 features into the second full connection layer, output 256 features after linear transformation and ReLU activation function, input the 256 features into the third full connection layer, output 64 features after linear transformation and ReLU activation function, and input the 512 features output by the first full connection layer into the fourth full connection layer, output 64 features after linear transformation and ReLU activation function, concatenate the features of the third and fourth full connection layers to integrate feature information of different levels, input the fifth full connection layer, and map the concatenated features to the original data dimension after linear transformation and ReLU activation function.

Images