Unmanned aerial vehicle formation inter-vehicle cooperative positioning method, system, device and medium

Abstract

The invention discloses an unmanned aerial vehicle formation inter-vehicle cooperative positioning method, system and device and a medium, and the method comprises the steps: carrying out the recognition of an array signal between a master unmanned aerial vehicle and a slave unmanned aerial vehicle, and obtaining a bidirectional LOS vector according to the recognition result and the navigation positioning information of the master unmanned aerial vehicle; in combination with the navigation positioning information of the main unmanned aerial vehicle and the bidirectional LOS vector, obtaining the pose of a slave unmanned aerial vehicle; predicting an optimal formation layout of the unmanned aerial vehicles according to the poses of the slave unmanned aerial vehicles and the navigation positioning information of the master unmanned aerial vehicle in combination with a graph neural network model and a deep reinforcement learning model; and performing coordinate conversion and error processing on the optimal formation layout to obtain a cooperative positioning result between unmanned aerial vehicle formations. According to the embodiment of the invention, the overall navigation precision of the unmanned aerial vehicle formation can be greatly improved, the optimal formation layout of the unmanned aerial vehicles is predicted in real time through the deep reinforcement learning model, positioning precision optimization in a complex dynamic environment is realized, and the method can be widely applied to the technical field of unmanned aerial vehicles.

Description

Technical Field

[0001] The present application relates to the field of unmanned aerial vehicle technology, and in particular to a method, system, device and medium for collaborative positioning between unmanned aerial vehicle formations. Background Art

[0002] At present, compared with single UAV operation, multi-UAV formation can effectively improve operation efficiency and system fault tolerance by forming a dynamic network during flight and exchanging information between multiple UAVs, enhance the system's ability to defend against assaults, and ensure the completion of tasks. In related technologies, the mainstream solution for UAV navigation is the combined navigation of the Inertial Navigation System (INS) and the Global Navigation Satellite System (GNSS). Although this solution can achieve accurate navigation of UAVs, GNSS has poor anti-electromagnetic interference capabilities. In complex environments, GNSS may be unavailable due to reasons such as the positioning signal being blocked, which in turn causes the combined navigation system to accumulate positioning errors during the time when the GNSS signal is certain, resulting in reduced positioning accuracy. Summary of the invention

[0003] In order to solve at least one of the above problems, the main purpose of the embodiments of the present application is to propose a method, system, equipment and medium for collaborative positioning between drone formations, aiming to achieve high-precision collaborative positioning between drone formations in complex environments.

[0004] Identify the array signal between the master UAV and the slave UAV, and obtain a bidirectional LOS vector according to the identification result and the navigation and positioning information of the master UAV;

[0005] Combining the navigation and positioning information of the master drone and the bidirectional LOS vector, obtaining the position and posture of the slave drone;

[0006] Combining the graph neural network model and the deep reinforcement learning model, the optimal formation layout of the drones is predicted according to the position of the slave drones and the navigation and positioning information of the master drone;

[0007] The optimal formation layout is subjected to coordinate transformation and error processing to obtain the collaborative positioning result between the UAV formations.

[0008] In some embodiments, the identifying of the array signal between the master drone and the slave drone, and obtaining a bidirectional LOS vector according to the identification result and the navigation and positioning information of the master drone, comprises the following steps:

[0009] Configure the main UAV with a global satellite navigation system and inertial equipment;

[0010] Obtaining navigation and positioning information of the master drone through the global satellite navigation system and the inertial device;

[0011] Identify the array signal broadcasted from the drone and obtain the positioning information of the drone;

[0012] A bidirectional LOS vector between the master drone and the slave drone is calculated according to the master drone navigation positioning information and the slave drone positioning information.

[0013] In some embodiments, combining the navigation positioning information of the master drone and the bidirectional LOS vector to obtain the position and posture of the slave drone comprises the following steps:

[0014] Obtaining a first attitude rotation matrix between the main UAV carrier coordinate systems according to the main UAV navigation and positioning information;

[0015] According to the first attitude rotation matrix and the bidirectional LOS vector, a second attitude rotation matrix between the master UAV carrier coordinate system and the slave UAV carrier coordinate system and a third attitude rotation matrix between the slave UAV carrier coordinate systems are calculated;

[0016] Obtaining a relative attitude angle between the master UAV and the slave UAV according to the first attitude rotation matrix, the second attitude rotation matrix and the third attitude rotation matrix;

[0017] According to the relative attitude angle, the position and posture of the slave drone is obtained.

[0018] In some embodiments, the combined graph neural network model and the deep reinforcement learning model predict the optimal formation layout of the drones according to the posture of the slave drones and the navigation and positioning information of the master drone, including the following steps:

[0019] Initialize the drone formation according to the position and posture of the slave drone and the navigation and positioning information of the host drone; wherein the drone formation includes an initial position, a relative angle and a speed;

[0020] Determine the relative position and relative distance of the UAV according to the initial position;

[0021] Predicting the positioning error of the UAV formation layout through a graph neural network model according to the relative distance, the relative angle and the speed of each of the UAVs;

[0022] At each time step, construct a state vector according to the relative position, the relative angle, the angle and the positioning error;

[0023] Inputting the state vector into the policy network of the deep reinforcement learning model, performing the action of adjusting the relative position and the relative angle, and updating the drone formation and the positioning error;

[0024] The strategy network is rewarded or punished according to the reward function and the positioning error, and the strategy network is updated until the positioning error is less than a preset limit value, thereby obtaining the optimal formation layout of the UAVs.

[0025] In some embodiments, the method of predicting the positioning error of the drone formation by a graph neural network model based on the relative distance, the relative angle, and the speed of each drone comprises the following steps:

[0026] Using the relative distance, relative angle and speed of each drone as node attributes to generate a graph structure;

[0027] Perform convolution processing on the graph structure through multi-layer convolution operations to update the feature vector of each node;

[0028] The feature vector is fully connected to obtain the positioning error of the UAV formation.

[0029] In some embodiments, the coordinate transformation and error processing of the optimal formation layout to obtain the cooperative positioning result between the UAV formations includes the following steps:

[0030] Obtaining position and posture measurement information of slave UAVs of the optimal formation layout;

[0031] Performing coordinate system conversion on the position and attitude measurement information of the slave UAV to obtain absolute navigation information;

[0032] According to the slave inertial navigation error model and the array measurement error model, the absolute navigation information is optimally estimated to obtain the inertial navigation error and the measurement error;

[0033] According to the inertial navigation error and the measurement error, the absolute navigation information is error compensated to obtain a collaborative positioning result between UAV formations.

[0034] In some embodiments, the method further comprises the following steps:

[0035] Constructing and training the graph neural network model;

[0036] Build and train the deep reinforcement learning model.

[0037] To achieve the above purpose, another aspect of the embodiment of the present application proposes a cooperative positioning system between drone formations, the system comprising:

[0038] The first module is used to identify the array signal between the master drone and the slave drone, and obtain a bidirectional LOS vector according to the identification result and the navigation and positioning information of the master drone;

[0039] The second module is used to combine the navigation and positioning information of the master drone and the bidirectional LOS vector to obtain the position and posture of the slave drone;

[0040] The third module is used to combine the graph neural network model and the deep reinforcement learning model to predict the optimal formation layout of the drones according to the posture of the slave drones and the navigation and positioning information of the master drone;

[0041] The fourth module is used to perform coordinate transformation and error processing on the optimal formation layout to obtain the collaborative positioning result between the UAV formations.

[0042] In some embodiments, the system further comprises:

[0043] The fifth module is used to build and train the graph neural network model;

[0044] The sixth module is used to build and train the deep reinforcement learning model.

[0045] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above-mentioned method when executing the computer program.

[0046] To achieve the above objective, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described above is implemented.

[0047] The embodiments of the present application include at least the following beneficial effects: The present application provides a method, system, device and medium for collaborative positioning between drone formations. The scheme adopts a master-slave drone information interaction architecture, identifies a bidirectional LOS vector through array signals, and determines the position and posture of the slave drone based on the navigation and positioning information of the master drone and the bidirectional LOS vector. Then, combined with the graph neural network model and the deep reinforcement learning model, the optimal formation layout of the drones is predicted, and finally the collaborative positioning result between the drone formation is obtained. The drones of different levels and different array signal position information in the drone formation can be accurately positioned and dynamically adjusted, which greatly improves the overall navigation accuracy of the drone formation. The optimal formation layout of the drones is predicted in real time through the deep reinforcement learning model, and the positioning accuracy optimization in a complex dynamic environment is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0049] Figure 1 This is a flow chart of a method for collaborative positioning between drone formations provided in an embodiment of the present application;

[0050] Figure 2 is a schematic diagram of a specific implementation positioning model provided in an embodiment of the present application;

[0051] Figure 3 It is a schematic diagram of the space vector of the drone formation unit provided in the embodiment of the present application;

[0052] Figure 4 It is a module schematic diagram of a cooperative positioning system between drone formations provided in an embodiment of the present application;

[0053] Figure 5 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0055] Although the functional modules are divided in the system 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 system or the order in the flowchart. The terms "first / S100", "second / S200", 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.

[0056] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0057] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0058] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0059] 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 this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0060] In the related technologies, the design method of the communication and navigation module is not designed for small UAV formations, and does not reflect the mutual enhancement of communication technology and navigation technology. At present, the mainstream solution for UAV navigation is the combined navigation of the Inertial Navigation System (INS) and the Global Navigation Satellite System (GNSS). Since the navigation solution of INS integrates the errors of angular velocity and acceleration, the position, velocity, and attitude errors all accumulate and diverge over the working time, while the positioning and velocity measurement errors of GNSS do not diverge over time. Therefore, using GNSS signals to correct the accumulated errors of INS can give full play to the characteristics of INS's strong anti-interference ability, strong autonomy, and good dynamic performance, and realize the precise navigation of UAVs. However, GNSS has poor anti-electromagnetic interference ability. In complex environments, GNSS may be unavailable due to reasons such as the shielding of the positioning signal, and then the positioning error of the combined navigation system accumulates during the time when the GNSS signal is missing. Therefore, the collaborative navigation of UAV formation requires the design of an inter-machine communication system and a relative measurement system to obtain the relative position and relative attitude information between UAVs. The measured relative position and attitude information is combined with the navigation information of each UAV to achieve high-precision collaborative navigation of the UAV formation and improve the fault tolerance capability of formation navigation.

[0061] However, many positioning solutions require the use of communication devices, relative navigation devices, inertial navigation devices and GPS positioning devices, which makes the entire body structure bloated. The small UAV has limited payload, which puts more burden on the UAV.

[0062] In view of this, a method, system, equipment and medium for collaborative positioning between drone formations are provided in an embodiment of the present application. The scheme proposes a method for optimizing the positioning accuracy of drone formations based on combined navigation of master-slave drone communication array signal detection and inertial system and introduces a graph neural network and deep reinforcement learning. Inertial navigation devices and GPS positioning devices are only installed on the master drone, which reduces the burden on the drone formation. The graph neural network is used to model the relative distances and relative angles between each drone. At the same time, deep reinforcement learning is used to dynamically optimize the position layout of each drone, and drones of different levels and different array signal posture information in the drone formation are accurately positioned, greatly improving the overall navigation accuracy of the drone formation.

[0063] Figure 1 is an optional flow chart of a method for cooperative positioning between drone formations provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S100 to S400.

[0064] Step S100, identifying array signals between drones to obtain a bidirectional LOS vector; wherein the drones include a master drone and a slave drone.

[0065] Step S200, combining the navigation and positioning information of the master drone and the bidirectional LOS vector to determine the position and posture of the slave drone.

[0066] Step S300, using a graph neural network model to predict the positioning error of the drone formation based on the posture of the slave drone and the navigation positioning information of the master drone.

[0067] Step S400, determining the optimal formation layout using a deep reinforcement learning model according to the position and posture of the slave drone, the navigation and positioning information of the master drone, and the positioning error.

[0068] Steps S100 to S400 shown in the embodiment of the present application adopt a master-slave UAV information interaction architecture, identify the bidirectional LOS vector through array signals, determine the slave UAV posture according to the navigation positioning information of the master UAV and the bidirectional LOS vector, and then combine the graph neural network model and the deep reinforcement learning model to predict the optimal formation layout of the UAVs, and finally obtain the collaborative positioning result between the UAV formations. It can accurately locate and dynamically adjust UAVs of different levels and different array signal posture information in the UAV formation, greatly improve the overall navigation accuracy of the UAV formation, and use the deep reinforcement learning model to predict the optimal formation layout of the UAVs in real time, so as to achieve optimization of positioning accuracy in complex dynamic environments.

[0069] In some embodiments, step S100 may include but is not limited to steps S110 to S140:

[0070] Step S110, configuring a global satellite navigation system and an inertial device on the master drone;

[0071] Step S120, obtaining navigation and positioning information of the master drone through the global satellite navigation system and the inertial device;

[0072] Step S130, identifying the array signal broadcasted by the slave drone to obtain the positioning information of the slave drone;

[0073] Step S140: Calculate the bidirectional LOS vector between the master drone and the slave drone according to the navigation positioning information of the master drone and the positioning information of the slave drone.

[0074] In some embodiments, step S200 may include but is not limited to the following steps S210 to S240:

[0075] Step S210, obtaining a first attitude rotation matrix between the main UAV carrier coordinate systems according to the navigation and positioning information of the main UAV;

[0076] Step S220, calculating a second attitude rotation matrix between the master UAV carrier coordinate system and the slave UAV carrier coordinate system, and a third attitude rotation matrix between each of the slave UAV carrier coordinate systems according to the first attitude rotation matrix and the bidirectional LOS vector;

[0077] Step S230, obtaining a relative attitude angle between the master UAV and the slave UAV according to the first attitude rotation matrix, the second attitude rotation matrix and the third attitude rotation matrix;

[0078] Step S240, obtaining the posture of the slave drone according to the relative posture angle.

[0079] In some embodiments, step S300 may include but is not limited to the following steps S310 to S360:

[0080] Step S310, initializing a drone formation according to the position and posture of the slave drone and the navigation and positioning information of the host drone; wherein the drone formation includes an initial position, a relative angle and a speed;

[0081] Step S320, determining the relative position and relative distance of the UAV according to the initial position;

[0082] Step S330, predicting the positioning error of the UAV formation layout through a graph neural network model according to the relative distance, the relative angle and the speed of each of the UAVs;

[0083] Step S340, constructing a state vector at each time step according to the relative position, the relative angle, the angle and the positioning error;

[0084] Step S350, inputting the state vector into the policy network of the deep reinforcement learning model, performing the action of adjusting the relative position and the relative angle, and updating the drone formation and the positioning error;

[0085] Step S360, rewarding or punishing the policy network according to the reward function and the positioning error, updating the policy network until the positioning error is less than a preset limit, and obtaining the optimal formation layout of the UAVs.

[0086] In some embodiments, step S320 may include but is not limited to the following steps S321 to S323:

[0087] Step S321, taking the relative distance, relative angle and speed of each drone as node attributes to generate a graph structure;

[0088] Step S322, performing convolution processing on the graph structure through multi-layer convolution operations to update the feature vector of each node;

[0089] Step S323, performing full connection processing on the feature vector to obtain the positioning error of the UAV formation.

[0090] In some embodiments, step S400 may include but is not limited to the following steps S410 to S440:

[0091] Step S410, obtaining the position and posture measurement information of the slave drones of the optimal formation layout;

[0092] Step S420, performing coordinate system conversion on the position and attitude measurement information of the slave drone to obtain absolute navigation information;

[0093] Step S430, performing optimal estimation on the absolute navigation information according to the slave inertial navigation error model and the array measurement error model to obtain an inertial navigation error and a measurement error;

[0094] Step S440: performing error compensation on the absolute navigation information according to the inertial navigation error and the measurement error to obtain a collaborative positioning result between UAV formations.

[0095] In addition, in some embodiments, the method further includes steps S500 to S600:

[0096] Step S500, construct and train the graph neural network model.

[0097] Step S600, constructing and training the deep reinforcement learning model.

[0098] Below, in conjunction with a specific application example of cooperative positioning between drone formations, the solution of the embodiment of the present application is described in detail and explained:

[0099] In an embodiment of the present application, a method for collaborative positioning between drone formations is provided, and the drone formation adopts a master-slave information interaction architecture. The master drone (LV) is equipped with GPS and high-precision inertial equipment to ensure its own high-precision navigation information through combined navigation. In order to reduce the cost of the drone formation, each slave drone (FV) uses the high-precision positioning information of the master drone and the relative position and attitude determination method of the array antenna between the machines to estimate the real-time navigation speed, attitude, and position error, and compensate for the posture information calculated by the low-precision inertial navigation equipment, so as to achieve the navigation application requirements of the drone formation with low cost, low load, high precision, and high formation configuration adaptability.

[0100] All drones in the drone formation are equipped with array antennas composed of multiple antenna elements and installed on the surface of the drones. The bidirectional LOS vector between the master and slave drones is obtained through array signal detection. After obtaining the navigation parameter output from the inertial measurement unit IMU (Inertial Measurement Unit), inertial navigation is solved to obtain navigation parameters such as attitude, speed, and position. At the same time, information such as the angle of arrival and the angle of departure between drones is obtained from the antenna array for solution to obtain the LOS vector between drones and the relative attitude angle and relative position between drones.

[0101] Taking into account the installation error of the inertial navigation device, the arm error, etc., a system state space error model is established based on the mechanical arrangement of INS, and the navigation parameter error is predicted using this model. Through the accurate prediction of the navigation positioning error, the navigation parameter error obtained by the state space model and the measurement model is optimally estimated to correct the INS navigation parameter error, thereby indirectly achieving the optimal estimation of the navigation parameters such as the system attitude, velocity, and position.

[0102] In order to solve the problem of error variation in relative distance and relative angle between drones in a drone formation, a Graph Neural Network (GNN) is used to process graph structure data, which can handle scenes with topological relationships such as drone formations. A graph neural network (GNN) is used to build a relative position and relative angle relationship model between drones, and the GNN is trained based on the training data set to predict positioning errors; Deep Reinforcement Learning (DRL) provides a dynamic optimization mechanism based on long-term rewards, which can be used to dynamically adjust the relative position and relative angle of drones to optimize positioning accuracy. The drone formation layout is dynamically adjusted through the policy network to optimize positioning accuracy; the relative position and relative angle of drones are monitored and adjusted in real time to ensure high positioning accuracy in complex dynamic environments.

[0103] Specifically, the embodiment of the present application includes the following steps:

[0104] (1) The array signals between the master UAV and the slave UAV are identified, and the bidirectional LOS vector is obtained based on the identification results and the navigation and positioning information of the master UAV.

[0105] like Figure 2 As shown in the figure, the square represents the drone, the connecting line is the line of sight between the drones, and the arrow is the body coordinate system. In a specific implementation of the positioning model, the XY plane of the drone carrier coordinate system coincides with the array antenna plane, and the positive direction of the z axis is the sky direction. n To the main drone LV m The signal transmission path wave departure angle AoD relative to FV when broadcasting signalsn The carrier coordinate system can be expressed as (δ m,n ,ε m,n ), where δ m,n It is the azimuth, which means the angle from the positive direction of the X-axis counterclockwise to the projection of the vector on the XY plane in the XY plane. m,n is the elevation angle, which represents the angle from the positive direction of the Y axis downward to the LOS diameter. n From FV in the carrier coordinate system n To LV m The unit LOS vector is represented by f n,m , in LV m In the carrier coordinate system, from LV m To FV n The unit LOS vector is represented as l n,m , LV m The unit LOS vector between is denoted as g n,m .

[0106] When all slave drones FVs transmit array signals synchronously, the master drone LV uses CDMA technology to identify different FVs. When all N FVs broadcast positioning information at the same time, the signal received by LV is the superposition of the signals sent by all FVs.

[0107] Based on this, LV m The signal received by the hth antenna element is:

[0108]

[0109] y n,i,m,h (t) = r m,n p n,i (δ m,n ,ε m,n )q m,h (δ' m,n ,ε' m,n ) n (tt n,m );

[0110] In the formula, r m,n is the free space path loss, which is determined by the distance between the master and slave drones. n (tt n,m ) is the signal transmission delay t n,m The subsequent spreading sequence. n,i (δ m,n ,ε m,n ) is the complex gain of the i-th element of the transmitting array antenna, which includes the antenna at the wave departure angle (δ m,n ,ε m,n) and the phase shift caused by its position relative to the reference antenna. Similarly, q m,h (δ' m,n ,ε' m,n ) is the complex gain of the hth element of the receiving antenna array, which includes the antenna at the angle of arrival (δ' m,n ,ε' m,n ) and its phase shift relative to the reference antenna position. n,i,m,h (t) is the signal from the ith element of the nth slave spacecraft antenna array, w m,h (t) is the noise of the master UAV and the hth receiving link of the slave UAV.

[0111] Then, multiple drones are detected and identified based on the orthogonal characteristics of the spread spectrum sequence. m Using the spread spectrum signal a distributed to the kth slave spacecraft n (t) for the received signal y m,h (t) Perform sliding correlation and spatial channel parameter estimation to obtain FV n To LV m The wave departure angle and wave arrival angle of the LOS path.

[0112] (2) Determine the position and posture of the slave drone by combining the navigation and positioning information of the master drone and the bidirectional LOS vector.

[0113] The schematic diagram of the space vector of the UAV formation unit is as follows Figure 3 As shown in the figure, since LV1 and LV2 are equipped with high-precision navigation equipment, the distance d between LV1 and LV2 1,2 is known high-precision information, from which the distance value d can be obtained 1,k and d 2,k The estimated value of .

[0114] In the drone relative posture measurement solution proposed in the embodiment of the present application, the LOS vector is measured by the array antenna signal, which is a bidirectional vector estimated simultaneously by both the transmitting and receiving ends of the same signal. There is no error caused by the non-parallelism of the vectors, so parallel calibration is no longer required. and They are the attitude rotation matrices from the drone carrier coordinate system FVn to the main drone carrier coordinate system BLV1, FVn to BLV2, and BLV1 to BLV2. It is a known quantity that can be calculated from the precise position data between the main drones. The LOS vector is also measured by the array antenna, so it can be calculated using the TRAID algorithm. and make Right now:

[0115]

[0116] vector and An orthogonal rectangular coordinate system D can be determined 123 ,vector and The orthogonal rectangular coordinate system V can be determined 123 , the three-axis unit vectors of the two rectangular coordinate systems are:

[0117]

[0118] The vector satisfies: It can be written in matrix form:

[0119] Where T = [t1 t2 t3], D = [d1 d2 d3], since both matrix D and matrix T are orthogonal matrices, the attitude rotation matrix C from the FVn body coordinate system to the LVk body coordinate system is calculated by the following formula: Therefore, the attitude rotation matrix between the master and slave UAVs is obtained, and the relative attitude angles between the master and slave UAVs can be obtained by obtaining the attitude rotation matrix between the master and slave UAVs.

[0120] The navigation output parameter error of the subsystem inertial navigation is selected as the state, that is, the IMU state error parameter is the attitude angle error φ = [φ E φ N φ U ] T , speed error δv n =[δv E δv N δv U ] T 、Position error δp n =[δp E δp N δp U ] T , gyro bias ε = [ε x ε y ε z ] T , Accelerometer bias Array antenna installation deviation angle μ = [μ x μ y μ z ] T .

[0121] Based on this, the state equation of the cooperative navigation system is: The calculation process of the state equation is as follows:

[0122] The inertial navigation error and the array antenna installation deviation angle error are combined to form the following state vector:

[0123]

[0124] Therefore, the state transfer matrix F is as follows:

[0125]

[0126] in, F SINS represents the state transfer matrix of the inertial navigation system, F aa 、F av 、F ap 、F va 、F vv 、F vp 、F pv 、F pp They are, respectively, the autocorrelation term of acceleration error, the influence of acceleration error on velocity error, the influence of acceleration error on position error, the influence of velocity error on acceleration error, the autocorrelation term of velocity error, the influence of velocity error on position error, the influence of position error on velocity error and the autocorrelation term of position error.

[0127]

[0128]

[0129] The process noise input matrix is:

[0130]

[0131] The state noise is

[0132]

[0133] in, and They are the gyroscope angular velocity measurement white noise and the accelerometer specific force measurement white noise respectively. ie is the angular velocity of the Earth's rotation, usually indicating the angular velocity of the Earth's rotation in an inertial system; L is the latitude, indicating the geographical latitude of the current location; R Nh R is the meridian curvature radius, which is usually used to represent the curvature radius of the earth in the navigation coordinate system, specifically the curvature radius of the meridian plane of the earth ellipsoid at that point; Mh v is the radius of curvature of the yoke, which is usually used to represent the radius of curvature of the yoke surface of the earth ellipsoid at that point; N , v E , v UThey represent the velocity components in the north, east and vertical directions in the navigation coordinate system (northeast celestial coordinate system) respectively; g0 represents the gravitational acceleration on the earth's surface; is the measurement error of the accelerometer assembly.

[0134] (3) Combining the graph neural network model and the deep reinforcement learning model, the optimal formation layout of the UAVs is predicted based on the posture of the slave UAVs and the navigation and positioning information of the master UAV.

[0135] The specific working steps are as follows 1-7:

[0136] 1. Initialize the environment and parameters. Initialize the drone formation, including the initial position, relative speed, angle and other parameters of each drone; initialize the weight parameters of the graph neural network (GNN) to predict the positioning error under the current formation layout; initialize the weight and learning rate of the deep reinforcement learning policy network; define the deep reinforcement learning algorithm Proximal Policy Optimization (PPO) used by the policy network; set the initial state space, action space and reward function.

[0137] 2. Construct state space. For each time step, construct the state vector s based on the relative position, relative angle, speed, and predicted current positioning error of the drone. t ; The state space S includes the relative distance, relative angle, speed, current positioning error, and topological structure represented by the adjacency matrix between drones. The state vector s t Feed it into the policy network as input.

[0138] 3. The policy network selects an action. The policy network selects an action based on the input state s t Perform forward pass and output the corresponding action a t Action a t Represents the adjustment suggestions for the position and angle of each drone. The action space A includes: adjusting the relative position of drones, adjusting the attitude angle of drones, selecting the main drone, etc. Each action can be represented by a vector, which is defined as the increment of the position or angle of each drone.

[0139] 4. Environment interaction and state update. After executing action a t After that, the position, attitude and relative distance of the UAV formation are updated. Through the graph neural network (GNN), the positioning error under the current formation layout is re-predicted and the state s is updated. t+1 The environment returns the new state s t+1 , and calculate the reward r generated after the action is executed t .

[0140] 5. Calculation of reward function. The reward function is set to be calculated according to the change of positioning error, and the goal is to minimize the positioning error of the entire formation: if the overall positioning error of the UAV formation is reduced, a positive reward r is given. t >0; if the positioning error increases, a negative reward r is given t <0. The form of the reward function is rt = -α×positioning error + β×energy saving efficiency + γ×formation stability, where α, β and γ are weight coefficients.

[0141] 6. Policy update. Use the PPO algorithm to update the policy network. PPO balances exploration and exploitation by limiting the magnitude of each update to prevent excessive policy updates from causing unstable training. Based on the current policy, perform a series of actions a t , collect a batch of data (state, action, reward). Calculate the policy gradient and update the parameters of the policy network so that the selected action can maximize the cumulative reward.

[0142] 7. Iteration. After each action is executed, the system enters the next state and repeats steps 3 to 6 to continuously update the strategy. After multiple training rounds, the strategy network will gradually learn the optimal drone formation layout to minimize the overall positioning error.

[0143] The UAV formation is considered as a graph structure, where each UAV is considered as a node, and the relative position (distance) and angle between UAVs are considered as the weight of the edge. The relative distance and angle between UAVs are obtained through sensors or simulation systems. Each node represents a UAV, and the weight of the edge is the relative distance and angle between UAVs.

[0144] A graph neural network (GNN) is used to capture the topological structure and relationship between drones. The input of the neural network is the node features in the graph structure (the location, relative distance and relative angle of the drone), and the message passing mechanism is used to learn on the graph structure, update the embedding vector of the node, and finally output the predicted positioning error.

[0145] The graph neural network structure includes:

[0146] Input layer: Receive the relative position, relative angle and speed information of each drone.

[0147] Graph convolution layer: The graph structure is processed through multi-layer graph convolution operations to update the feature vector of each node.

[0148] Fully connected layer: fully connect the embedded vector after graph convolution and output the positioning error.

[0149] The policy gradient method (such as Proximal Policy Optimization, PPO) in deep reinforcement learning (DRL) is used to achieve dynamic layout optimization of drone formations. Through environmental interaction, the policy network learns the optimal strategy under different formation layouts. The structure of the deep reinforcement learning network includes:

[0150] State space: relative distance, relative angle and current positioning error between drones.

[0151] Action Space: Compare the position and relative angle of each drone to select the optimal formation layout.

[0152] Reward function: Set a reward function that minimizes the positioning error. When the overall positioning error decreases, give the policy network a positive reward.

[0153] In the simulation environment, the DRL algorithm is used to train the policy network. Through interaction with the environment, the policy network gradually learns the optimal UAV formation layout to minimize the overall positioning error. This process is completed through multiple iterations, and the policy network parameters are adjusted after each iteration to improve the robustness of the policy.

[0154] In practical applications, the positioning error under the current formation layout is predicted by the graph neural network, and the deep reinforcement learning network adjusts the selection of the main drone and its relative attitude angle according to the error value. This dynamic adjustment strategy ensures that the drone always maintains high positioning accuracy in complex environments.

[0155] This can be achieved by building graph neural network models and deep reinforcement learning models.

[0156] (4) Coordinate transformation and error processing are performed on the optimal formation layout to obtain the collaborative positioning result between the UAV formations.

[0157] Due to assembly errors and other reasons, there is often a deviation angle vector μ = [μ x μ y μ z ] T , let the deviation angle coordinate transformation matrix be:

[0158]

[0159] Considering the installation deviation angle, three-dimensional angle random error, and three-dimensional distance random error between the coordinate axes of the antenna array coordinate system (t system) and the ideal carrier coordinate system (b system), the AAN relative posture measurement model can be expressed as:

[0160]

[0161] In the formula Represents relative angle measurements, Represents the relative distance measurement value of the array antenna, and is the true value of the theoretical measured attitude and the theoretical position measured by AAN, μ is the installation error angle, δ φ is the random angle error, δ p Represents the random error of relative position.

[0162] In some embodiments, the "East-North-Sky" geographic coordinate system can be selected as the navigation reference coordinate system (n system) of the inertial navigation system, and the information solved is the absolute navigation information under the n system. In this case, it is necessary to convert the solved relative navigation information under the n system into absolute navigation information. The slave drone transmits the data link through the antenna array to receive the precise navigation information of the master drone, and adds it to its own relative information relative to the master drone to obtain the calculated absolute navigation information of the slave drone, and then subtracts the slave drone's own inertial navigation absolute navigation information from the calculated absolute navigation information to obtain the measurement model.

[0163] The position information calculated by the array antenna is the distance of the slave drone relative to the master drone in three directions. Since the distance between the drones in the formation is much smaller than the diameter of the earth, the straight-line distance between the drones is approximated as the arc length of the two positions on the earth's surface. The relative position in the navigation coordinate system is calculated as:

[0164]

[0165] Where R M and R N are the local meridian curvature radius and the meridian curvature radius respectively; p x , p y , p z is the relative distance between the master and slave UAVs measured by AAN; L L ,λ L ,h L Main UAV dimensional accuracy and altitude positioning information; L F ,λ F ,h F The longitude, latitude and altitude positioning information of the UAV is calculated.

[0166] According to the AAN relative posture measurement model, the posture error and position error can be expressed as:

[0167]

[0168] The system observation can be expressed as:

[0169]

[0170] The observation matrix is:

[0171]

[0172]

[0173] V=[δ φ δ p ] T ;

[0174] Among them, δ φ and δ p They are AAN attitude measurement white noise and position measurement white noise respectively. According to the system state equation and measurement equation, the Kalman filter algorithm is used to estimate and correct the inter-machine AAN navigation measurement error and the inertial navigation error obtained from the UAV, and finally the cooperative positioning result between the UAV formation is obtained.

[0175] In summary, the embodiments of the present application at least include the following beneficial effects:

[0176] (1) In view of the problems such as limited payload of small UAVs, accumulation of positioning errors of inertial navigation systems over time, unavailability of global satellite navigation in shielded environments, and influence of relative positions and relative attitudes between UAVs on positioning errors, the present application provides a method, system, device and medium for collaborative positioning between UAV formations. The scheme adopts a master-slave UAV information interaction architecture, identifies a bidirectional LOS vector through array signals, determines the position and attitude of the slave UAV based on the navigation and positioning information of the master UAV and the bidirectional LOS vector, and then predicts the optimal formation layout of the UAVs by combining a graph neural network model and a deep reinforcement learning model. Finally, the collaborative positioning result between the UAV formation is obtained. The UAVs of different levels and different array signal position and attitude information in the UAV formation can be accurately positioned and dynamically adjusted, greatly improving the overall navigation accuracy of the UAV formation. The optimal formation layout of the UAVs is predicted in real time through a deep reinforcement learning model, and the positioning accuracy optimization in a complex dynamic environment is achieved.

[0177] (2) The inertial navigation system and GPS system are only installed on the main UAV, which can reduce the load on the UAV formation.

[0178] See also Figure 4 The embodiment of the present application also provides a UAV formation inter-machine collaborative positioning system, which can implement the above-mentioned UAV formation inter-machine collaborative positioning method, and the system includes:

[0179] The first module 101 is used to identify the array signal between the master drone and the slave drone, and obtain a bidirectional LOS vector according to the identification result and the navigation and positioning information of the master drone;

[0180] The second module 102 is used to obtain the position and posture of the slave drone by combining the navigation and positioning information of the master drone and the bidirectional LOS vector;

[0181] The third module 103 is used to combine the graph neural network model and the deep reinforcement learning model to predict the optimal formation layout of the drones according to the posture of the slave drones and the navigation and positioning information of the master drone;

[0182] The fourth module 104 is used to perform coordinate transformation and error processing on the optimal formation layout to obtain the collaborative positioning result between the UAV formations.

[0183] In some embodiments, the system further comprises:

[0184] The fifth module is used to build and train the graph neural network model;

[0185] The sixth module is used to build and train the deep reinforcement learning model.

[0186] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0187] The embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned cooperative positioning method between drone formations when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0188] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0189] See also Figure 5 , Figure 5 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0190] The processor 201 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0191] The memory 202 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 202 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 202, and the processor 201 calls and executes a method for cooperative positioning between drone formations in the embodiment of this application;

[0192] Input / output interface 203, used to implement information input and output;

[0193] The communication interface 204 is used to realize the communication interaction between the device and other devices. The communication can be realized through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.);

[0194] Bus 205 , which transmits information between various components of the device (e.g., processor 201 , memory 202 , input / output interface 203 , and communication interface 204 );

[0195] The processor 201 , the memory 202 , the input / output interface 203 and the communication interface 204 are connected to each other in communication within the device via the bus 205 .

[0196] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned method for collaborative positioning between drone formations.

[0197] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0198] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0199] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0200] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0201] The system embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0202] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0203] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0204] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0205] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0206] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0207] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0208] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0209] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for cooperative positioning between drone formations, characterized in that: The following steps are involved: Identify the array signal between the master UAV and the slave UAV, and obtain a bidirectional LOS vector according to the identification result and the navigation and positioning information of the master UAV; Combining the navigation and positioning information of the master drone and the bidirectional LOS vector, obtaining the position and posture of the slave drone; Combining the graph neural network model and the deep reinforcement learning model, the optimal formation layout of the drones is predicted according to the position of the slave drones and the navigation and positioning information of the master drone; The optimal formation layout is subjected to coordinate transformation and error processing to obtain the collaborative positioning result between the UAV formations.

2. The method according to claim 1, characterized in that The array signal between the master UAV and the slave UAV is identified, and a bidirectional LOS vector is obtained according to the identification result and the navigation and positioning information of the master UAV, including the following steps: Configure the main UAV with a global satellite navigation system and inertial equipment; Obtaining navigation and positioning information of the master drone through the global satellite navigation system and the inertial device; Identify the array signal broadcasted from the drone and obtain the positioning information of the drone; A bidirectional LOS vector between the master drone and the slave drone is calculated according to the master drone navigation positioning information and the slave drone positioning information.

3. The method according to claim 1, characterized in that The step of combining the navigation and positioning information of the master drone and the bidirectional LOS vector to obtain the position and posture of the slave drone comprises the following steps: Obtaining a first attitude rotation matrix between the main UAV carrier coordinate systems according to the main UAV navigation and positioning information; According to the first attitude rotation matrix and the bidirectional LOS vector, a second attitude rotation matrix between the master UAV carrier coordinate system and the slave UAV carrier coordinate system and a third attitude rotation matrix between the slave UAV carrier coordinate systems are calculated; Obtaining a relative attitude angle between the master UAV and the slave UAV according to the first attitude rotation matrix, the second attitude rotation matrix and the third attitude rotation matrix; According to the relative attitude angle, the position and posture of the slave drone is obtained.

4. The method according to claim 1, characterized in that: The combined graph neural network model and the deep reinforcement learning model predict the optimal formation layout of the drones according to the posture of the slave drones and the navigation and positioning information of the master drone, including the following steps: Initialize the drone formation according to the position and posture of the slave drone and the navigation and positioning information of the host drone; wherein the drone formation includes an initial position, a relative angle and a speed; Determine the relative position and relative distance of the UAV according to the initial position; Predicting the positioning error of the UAV formation layout through a graph neural network model according to the relative distance, the relative angle and the speed of each of the UAVs; At each time step, construct a state vector according to the relative position, the relative angle, the angle and the positioning error; Inputting the state vector into the policy network of the deep reinforcement learning model, performing the action of adjusting the relative position and the relative angle, and updating the drone formation and the positioning error; The strategy network is rewarded or punished according to the reward function and the positioning error, and the strategy network is updated until the positioning error is less than a preset limit value, thereby obtaining the optimal formation layout of the UAVs.

5. The method according to claim 4, characterized in that The method of predicting the positioning error of the UAV formation by using a graph neural network model according to the relative distance, the relative angle and the speed of each UAV comprises the following steps: Using the relative distance, relative angle and speed of each drone as node attributes to generate a graph structure; Perform convolution processing on the graph structure through multi-layer convolution operations to update the feature vector of each node; The feature vector is fully connected to obtain the positioning error of the UAV formation.

6. The method according to claim 1, characterized in that The step of performing coordinate transformation and error processing on the optimal formation layout to obtain the collaborative positioning result between the UAV formations includes the following steps: Obtaining position and posture measurement information of slave UAVs of the optimal formation layout; Performing coordinate system conversion on the position and attitude measurement information of the slave UAV to obtain absolute navigation information; According to the slave inertial navigation error model and the array measurement error model, the absolute navigation information is optimally estimated to obtain the inertial navigation error and the measurement error; According to the inertial navigation error and the measurement error, the absolute navigation information is error compensated to obtain a collaborative positioning result between UAV formations.

7. The method according to claim 1, characterized in that The method further comprises the following steps: Constructing and training the graph neural network model; Build and train the deep reinforcement learning model.

8. A cooperative positioning system between drone formations, characterized in that: include: The first module is used to identify the array signal between the master drone and the slave drone, and obtain a bidirectional LOS vector according to the identification result and the navigation and positioning information of the master drone; The second module is used to combine the navigation and positioning information of the master drone and the bidirectional LOS vector to obtain the position and posture of the slave drone; The third module is used to combine the graph neural network model and the deep reinforcement learning model to predict the optimal formation layout of the drones according to the posture of the slave drones and the navigation and positioning information of the master drone; The fourth module is used to perform coordinate transformation and error processing on the optimal formation layout to obtain the collaborative positioning result between the UAV formations.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 7.

10. A computer storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to implement the method according to any one of claims 1 to 7 when executed by the processor.

Images