Formation control method for unmanned aerial vehicle cluster

Abstract

The invention discloses a formation control method for an unmanned aerial vehicle cluster, and relates to the technical field of intelligent formation of unmanned aerial vehicles, and the method comprises the following steps: S1, obtaining an initial space state of the unmanned aerial vehicle cluster; s2, performing adjacent constraint construction by using the initial state vector; s3, performing geometric mapping by using the local adjacency relation graph; s4, performing global formation generation by using the relative geometric position matrix; and S5, performing dynamic formation control calculation by using the relative geometric position matrix and the target formation reference model. By setting multi-dimensional measurement calculation based on an initial state vector and combining a combined constraint set of a spatial distance, a speed difference and an angular speed difference, a weighted local adjacency relation graph is formed, and according to the scheme, spatial proximity and dynamic consistency can be considered while the adjacency relation is constructed; the generated graph structure not only reflects the geometric closeness degree between the unmanned aerial vehicles, but also quantifies the compatibility of the unmanned aerial vehicles in motion and attitude changes.

Description

Technical Field

[0001] This invention relates to the field of intelligent formation technology for unmanned aerial vehicles (UAVs), specifically to a formation control method for a UAV swarm. Background Technology

[0002] Unmanned aerial vehicle (UAV) swarm formation control is a crucial area of ​​current research in intelligent unmanned systems, with wide applications in disaster relief, regional patrols, and collaborative reconnaissance. Its core lies in enabling UAVs to maintain a pre-defined formation and flexibly adjust it in dynamic environments through mutual perception and collaborative computation among multiple UAVs.

[0003] In existing technologies, common methods often use a single spatial distance threshold or a fixed number of neighbors as the basis for determining the connection when constructing adjacency relationships. These methods ignore the consistency of the dynamic characteristics of UAVs, such as speed and angular velocity, which makes the generated adjacency relationship graph unable to accurately reflect the real cluster coupling state, thus affecting subsequent geometric mapping and target formation generation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for formation control of unmanned aerial vehicle (UAV) swarms, thereby resolving the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a method for formation control of a drone swarm, comprising the following steps:

[0007] S1. Obtain the initial spatial state of the drone cluster and obtain the initial state vector;

[0008] S2. Use the initial state vector to construct the proximity constraint and obtain the local adjacency graph;

[0009] S3. Use the local adjacency graph to perform geometric mapping to obtain the relative geometric position matrix;

[0010] S4. Use the relative geometric position matrix to generate the global formation and obtain the target formation reference model;

[0011] S5. Dynamic formation control calculations are performed using the relative geometric position matrix and the target formation reference model to obtain the control command set for the UAV cluster.

[0012] To further optimize this technical solution, step S1 first involves multi-source information acquisition, then state estimation and fusion, and finally initial state vector construction. The resulting output shows that the UAV cluster size is N, and GNSS / IMU fusion and time synchronization are completed under a unified timestamp t0. The individual state vector of UAV i is expressed by the following formula:

[0013]

[0014] In the formula,

[0015] s i (t0) represents the single-machine state vector of the i-th UAV, which is a 12-dimensional vector;

[0016] This represents the position estimate after GNSS / IMU fusion;

[0017] This represents the depth estimate after fusion;

[0018] This represents the angular velocity of the IMU after calibration and alignment, expressed in ENU;

[0019] Indicates IMU linear acceleration;

[0020] Therefore, the initial state vector is:

[0021]

[0022] In the formula,

[0023] S0 represents the initial state vector of the entire formation at time t0, which is a 12N-dimensional vector;

[0024] N represents the number of drones.

[0025] To further optimize this technical solution, step S2 first performs pairwise quantization measurement calculation, then performs joint proximity constraint determination, then constructs the adjacency matrix and edge set, then performs edge weight mapping and weighted graph formation, and finally outputs the local adjacency relationship graph.

[0026] In step S2, when performing pairwise quantization calculations, for any two distinct machines i,j (i≠j), the following values ​​are calculated:

[0027] Spatial distance: d ij =‖p i (t0)-p j (t0)‖2 represents the geometric proximity of the two machines in the ENU coordinate system;

[0028] Speed ​​difference: u ij =||v i (t0)-v j (t0)||2 represents the measure of the convergence of motion between the two machines;

[0029] Angular velocity difference: q ij =||ω i(t0)-ω j (t0)||2 represents the consistency of the attitude change rate of the two aircraft.

[0030] To further optimize this technical solution, in step S2, when performing joint proximity constraint determination, a threshold R is set. d >0、R v >0、R ω >0, and define the joint constraint set:

[0031] C = {C1, C2, C3};

[0032] Where C1:d ij ≤R d ;C2:u ij ≤R v ;C3:q ij ≤R ω ;

[0033] In the formula,

[0034] R d Spatial distance threshold;

[0035] R v Speed ​​difference threshold;

[0036] R ω Angular velocity difference threshold.

[0037] To further optimize this technical solution, in step S2, when performing joint proximity constraint determination, a determination indicator is set:

[0038]

[0039] χ ij =1 indicates that the two machines satisfy the constraints of spatial proximity and dynamic consistency, allowing the establishment of an edge between nodes i and j; χ ij =0 indicates no edge connection.

[0040] To further optimize this technical solution, in step S2, when constructing the adjacency matrix and edge set, the adjacency matrix is ​​A, and its expression is:

[0041] A ij =χ ij A ii =0;

[0042] A∈{0,1} N×N Used to represent the edge relationships within a drone swarm;

[0043] A ij The element in the i-th row and j-th column of the matrix is ​​used to indicate whether there is a connection between drones i and j;

[0044] A ii =0: The diagonal elements are always zero, indicating that the drone does not establish an edge connection with itself.

[0045] To further optimize this technical solution, in step S2, when constructing the adjacency matrix and edge set, the edge set is: E local ={(i,j)∣A ij =1};

[0046] E local The edge set of the local adjacency graph, containing all pairs of unmanned aerial vehicles that satisfy the joint constraints;

[0047] (i,j): Represents an undirected edge between drones i and j;

[0048] A ij =1: As a filtering condition, it means that i and j are determined to have a connection in the adjacency matrix;

[0049] E local Summarize all node pairs that satisfy the joint constraints to form the structural skeleton of the graph.

[0050] To further optimize this technical solution, in step S2, when performing edge weight mapping and weighted graph formation, a monotonically decreasing normalized mapping is set. This mapping reflects the proximity strength, and its expression is:

[0051]

[0052] In the three expressions, φ d ,φ v ,φ ω These expressions reflect their relative proximity in the dimensions of spatial distance, velocity difference, and angular velocity difference, respectively. All three expressions are monotonically decreasing, indicating that the closer the difference value is to zero, the closer the proximity intensity is to 1.

[0053] To further optimize this technical solution, in step S2, when performing edge weight mapping and weighted graph formation, a non-negative coefficient α is manually set based on task requirements. d ,α v ,α ω α d +α v +α ω =1, used for linear weighted fusion among the three neighboring indices, the resulting edge weight expression is:

[0054] W ij =χ ij (α d φ d (d ij )+α v φ v(u ij )+α ω φ ω (q ij ));

[0055] When χ ij When W = 1, ij ∈[0,1] reflects the comprehensive proximity strength between the two machines; when χ ij When W = 0, ij =0 indicates no edge connection, and the edge weight is used as a priority reference when splicing geometric relationships and robust matching in step S3.

[0056] To further optimize this technical solution, in step S2, when outputting the local adjacency graph, the node set for all UAV nodes in the cluster is set as V = {1,2,…,N}. Combining the adjacency matrix A and the edge weight matrix W, the expression for the local adjacency graph is:

[0057]

[0058] In a second aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program instructions, when executed by the processor, implement the steps of a formation control method for a drone swarm as described in the first aspect of the present invention.

[0059] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of a formation control method for a drone swarm as described in the first aspect of the present invention.

[0060] Compared with the prior art, the present invention provides a method for formation control of UAV swarms, which has the following beneficial effects:

[0061] The formation control method for this UAV swarm, by setting up multi-dimensional calculations based on the initial state vector and combining a joint constraint set of spatial distance, velocity difference, and angular velocity difference, forms a weighted local adjacency graph. This scheme can simultaneously consider spatial proximity and dynamic consistency when constructing adjacency relationships, so that the generated graph structure not only reflects the geometric proximity between UAVs, but also quantifies their compatibility in motion and attitude changes. The resulting local adjacency graph can provide robust and accurate structural support for subsequent geometric mapping and formation generation, fundamentally solving the problem of insufficient dimensions in adjacency relationship construction in existing technologies. Attached Figure Description

[0062] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart illustrating a method for formation control of a drone swarm proposed in this invention.

[0064] Figure 2 This is a schematic diagram of the proximity constraint construction process for a formation control method for UAV swarms proposed in this invention.

[0065] Figure 3 This is a schematic diagram of the geometric mapping process for a formation control method for unmanned aerial vehicle (UAV) swarms proposed in this invention.

[0066] Figure 4 This is a schematic diagram of the dynamic formation control calculation process of the formation control method for a drone swarm proposed in this invention. Detailed Implementation

[0067] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0068] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0069] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0070] Example 1:

[0071] Reference Figures 1-4 This is the first embodiment of the present invention, which provides a method for formation control of a drone swarm, including the following steps:

[0072] S1. Obtain the initial spatial state of the drone cluster and obtain the initial state vector;

[0073] This step provides complete and accurate status data for subsequent steps;

[0074] Step S1 first involves collecting multi-source information, including:

[0075] The three-dimensional position coordinates of each UAV are obtained through mature GNSS positioning technology;

[0076] The attitude angular velocity and acceleration are collected using the inertial measurement unit (IMU) built into the drone.

[0077] The above data is synchronized with timestamps to ensure that the data is comparable at the same time.

[0078] Step S1 then performs state estimation and fusion, using the mature Kalman filtering method to fuse GNSS and IMU data to eliminate noise and drift caused by a single sensor;

[0079] The fusion results provide smoothed velocity information and more accurate position estimation for the UAV.

[0080] Step S1 ends with the construction of the initial state vector, which arranges the position and velocity of each UAV in order of number to form a complete set of initial states;

[0081] This set is organized into an initial state vector, which serves as a unified description of the spatial state of the entire UAV swarm.

[0082] In the output of step S1, the drone swarm size is N, and GNSS / IMU fusion and time synchronization are completed under a unified timestamp t0. The individual state vector of drone i is expressed by the following formula:

[0083]

[0084] In the formula,

[0085] s i (t0) represents the single-machine state vector of the i-th UAV, which is a 12-dimensional vector;

[0086] This represents the position estimate after GNSS / IMU fusion;

[0087] This represents the depth estimate after fusion;

[0088] This represents the angular velocity of the IMU after calibration and alignment, expressed in ENU;

[0089] This indicates the linear acceleration of the IMU, which has been compensated for gravity and converted to ENU.

[0090] Therefore, the initial state vector is:

[0091]

[0092] In the formula,

[0093] S0 represents the initial state vector of the entire formation at time t0, which is a 12N-dimensional vector;

[0094] N represents the number of drones.

[0095] S2. Use the initial state vector to construct the proximity constraint and obtain the local adjacency graph;

[0096] Step S2 takes the initial state vector S0 output from step S1 as input and extracts the p values ​​of each UAV from S0. i (t0), v i (t0), ω i (t0);

[0097] Calculate the difference between the spatial distance and dynamic consistency between each pair of elements;

[0098] Make edge determination based on the joint constraint set C = {C1, C2, C3};

[0099] Generate a local adjacency graph (including adjacency matrix and edge weights) as the sole input for the next step of geometric mapping.

[0100] Step S2 first performs pairwise quantization metric calculation, then performs joint proximity constraint determination, then constructs the adjacency matrix and edge set, then performs edge weight mapping and weighted graph formation, and finally outputs the local adjacency graph.

[0101] In step S2, when performing pairwise quantization calculations, for any two distinct machines i,j (i≠j), the following values ​​are calculated:

[0102] Spatial distance: d ij =||p i (t0)-p j (t0)||2 represents the geometric proximity of the two machines in the ENU coordinate system; the smaller the value, the closer they are.

[0103] Speed ​​difference: u ij =||v i (t0)-v j (t0)||2 represents the measure of the similarity of motion between the two machines; the smaller the value, the closer the motion states are.

[0104] Angular velocity difference: q ij =||ω i (t0)-ω j (t0)||2 represents the consistency of the attitude change rate of the two aircraft. The smaller the value, the more consistent the rotation characteristics of the aircraft.

[0105] In step S2, when performing the joint proximity constraint determination, a threshold R is set. d >0、R v >0、R ω >0, and define the joint constraint set:

[0106] C = {C1, C2, C3};

[0107] Where C1:d ij ≤R d ;C2:u ij ≤R v ;C3:q ij ≤R ω ;

[0108] In the formula,

[0109] R d Spatial distance threshold: Used to limit the maximum permissible relative distance between drones. When the geometric distance d between two drones... ij No more than R d At that time, it was assumed that they were spatially proximate and possessed the physical conditions to establish a connection.

[0110] R v Speed ​​difference threshold: This limits the maximum permissible speed difference between drones. When the speed difference between the two drones is u... ij No more than R v At that time, it was assumed that they tended to be consistent in their motion state and would not quickly break away from their adjacency relationship due to excessive relative speed;

[0111] R ω Angular velocity difference threshold: This limits the maximum permissible angular velocity difference between drones. When the angular velocity difference q between two drones... ij No more than R ω When this occurs, it indicates that their posture changes are relatively consistent and will not disrupt local adjacency relationships due to rapid deflection of the body posture.

[0112] Set the judgment indicator:

[0113]

[0114] χ ij =1 indicates that the two machines satisfy the constraints of spatial proximity and dynamic consistency, allowing the establishment of an edge between nodes i and j; χ ij =0 indicates no edge connection.

[0115] In step S2, when constructing the adjacency matrix and edge set, the adjacency matrix is ​​A, and its expression is:

[0116] A ij =χij A ii =0;

[0117] A∈{0,1} N×N Used to represent the edge relationships within a drone swarm;

[0118] A ij The element in the i-th row and j-th column of the matrix is ​​used to indicate whether there is a connection between drones i and j;

[0119] A ii =0: The diagonal elements are always zero, indicating that the drone does not establish an edge connection with itself;

[0120] The edge set is: E local ={(i,j)∣A ij =1};

[0121] E local The edge set of the local adjacency graph, containing all pairs of unmanned aerial vehicles that satisfy the joint constraints;

[0122] (i,j): Represents an undirected edge between drones i and j;

[0123] A ij =1: As a filtering condition, it means that i and j are determined to have a connection in the adjacency matrix;

[0124] E local Summarize all node pairs that satisfy the joint constraints to form the structural skeleton of the graph.

[0125] In step S2, during the edge weight mapping and weighted graph formation, a monotonically decreasing normalized mapping is set to reflect the proximity strength. Its expression is:

[0126]

[0127] In the three expressions, φ d ,φ v ,φ ω These reflect their relative proximity in the dimensions of spatial distance, velocity difference, and angular velocity difference, respectively. All three expressions are monotonically decreasing, indicating that the closer the difference value is to zero, the closer the proximity intensity is to 1.

[0128] Based on task requirements, a non-negative coefficient α is manually set. d ,α v ,α ω α d +α v +α ω =1, used for linear weighted fusion among the three neighboring indices, the resulting edge weight expression is:

[0129] W ij =χ ij (α d φ d (d ij )+α v φ v (u ij )+α ω φ ω (q ij ));

[0130] When χ ij When W = 1, ij ∈[0,1] reflects the comprehensive proximity strength between the two machines; when χ ij When W = 0, ij =0 indicates no edge connection, and the edge weight is used as a priority reference when splicing geometric relationships and robust matching in step S3.

[0131] In step S2, when outputting the local adjacency graph, for all drone nodes in the cluster, the node set is set as V = {1, 2, ..., N}. Combining the adjacency matrix A and the edge weight matrix W, the expression for the local adjacency graph is:

[0132]

[0133] This expression adds weight information to the existing adjacency structure, so that the output adjacency graph maintains the clarity of the topology and has quantitative strength features for subsequent formation control.

[0134] Existing common practices are mostly based on connecting edges with a single distance threshold or connecting edges with a fixed number of nearest neighbors, ignoring the impact of the consistency of velocity and attitude change rates on formation stability. Step S2 makes a judgment based on the joint constraint set C = {C1, C2, C3}, and forms a quantitative strength by normalizing edge weight mapping and convex combination coefficients. The output is a weighted local adjacency graph, which provides strength information and robustness support for subsequent geometric mapping.

[0135] S3. Use the local adjacency graph to perform geometric mapping to obtain the relative geometric position matrix;

[0136] Step S3 uses the local adjacency graph output in step S2. The proximity relationships corresponding to each edge in the graph are transformed into geometric constraints. These constraints are then iteratively assembled in a unified coordinate system to construct a consistent geometric framework within the group. Finally, a relative geometric position matrix that can characterize the relative positional relationships among all UAVs is output.

[0137] Step S3 first performs the transformation from edge weights to geometric constraints, then performs relative geometric splicing, and finally performs global uniform mapping;

[0138] In step S3, when transforming edge weights into geometric constraints, for any connected edge (i,j)∈E local Based on the edge weight W obtained in step S2 ij ∈[0,1], set geometric constraints:

[0139] g ij =W ij ·(p i (t0)-p j (t0));

[0140] In the formula, g ij It represents the weighted relative position of nodes i and j in the ENU coordinate system;

[0141] Boundary W ij The larger the value, the closer the relationship between the two machines, and the stronger the binding force of the geometric constraint in subsequent mappings.

[0142] Step S3 involves concatenating the geometric constraints of all edges into a matrix form during relative geometric splicing:

[0143] G={g ij |(i,j)∈E local};

[0144] Where G represents a set matrix of geometric constraints, containing the weighted relative position vectors of all connected edges, describing the splicing relationship from local to global.

[0145] In step S3, during the global uniformity mapping, the geometric constraint set G is projected onto a unified global relative framework using the least squares approach:

[0146]

[0147] in, This represents a relative geometric position matrix, with each row corresponding to the relative position coordinates of a drone.

[0148] This process will apply all local constraints g ij By merging, we obtain a relatively geometric solution that is consistent across the entire group.

[0149] In mature formation geometry construction methods, geometric derivation based on a fixed topology or a fixed reference machine is common (e.g., constructing relative coordinates based on the navigator). Step S3, based on weighted geometric stitching of the adjacency graph, adjusts the geometric constraint strength by edge weights, achieving a comprehensive mapping of local information. This is significantly different from existing methods.

[0150] S4. Use the relative geometric position matrix to generate the global formation and obtain the target formation reference model;

[0151] Step S4 is based on the relative geometric position matrix P obtained in step S3. rel , will P rel The relative positional relationships in the model are mapped to the preset global formation template. Through geometric operations such as translation, rotation and scaling, the relative position matrix is ​​aligned to the global reference coordinate system to obtain a consistent spatial layout of the group as a whole, i.e., the target formation reference model.

[0152] Step S4 first performs matching of the relative matrix to the global template. Using mature point set matching methods (such as rigid body registration algorithms based on minimum mean square error), P is... rel Match the preset target formation template;

[0153] The target formation template is directly set based on existing requirements and mature geometric design methods. The determination process is as follows: select the overall formation form according to the task scenario, such as rhombus, arrow, circle or matrix.

[0154] In the global reference coordinate system, using geometric point layout or regular grid methods, individual UAVs are evenly or orderly distributed in space to form a standardized spatial layout;

[0155] By calibrating with a fixed coordinate system, the layout becomes a globally consistent geometric template.

[0156] Step S4 then performs geometric transformation alignment. Based on the matched relative position matrix and the target template, mature rigid body transformation methods are used, including translation, rotation, and scale normalization, to align P. rel Mapping to the global reference coordinate system yields globally consistent formation coordinates;

[0157] Step S4 finally generates the target formation reference model. Based on the globally consistent formation coordinates, the overall formation layout is output in the global reference coordinate system to obtain the target formation reference model, which serves as the geometric reference input for subsequent control laws.

[0158] The target formation reference model expression is:

[0159] P ref =1 N t T +s(ΠP rel )R T ;

[0160] In the formula,

[0161] The relative geometric position matrix output by step S3 (each row corresponds to the relative coordinates of a UAV);

[0162] ∏∈{0,1} N×N: Matching (permutation) matrix, derived from the point set matching results, used to match P rel The row order corresponds one-to-one with the roles / positions in the preset target formation template (each row has exactly one 1);

[0163] R∈SO(3): Rotation matrix, derived from rigid body transformation alignment;

[0164] s>0: Scale factor, derived from scale normalization;

[0165] Translation vectors, derived from global translation alignment;

[0166] A column vector of all ones, used to copy t to each drone;

[0167] The target formation reference model, i.e. the global formation coordinate matrix finally output in step S4, is subsequently used as the geometric reference for generating control commands in step S5.

[0168] S5. Dynamic formation control calculations are performed using the relative geometric position matrix and the target formation reference model to obtain the control command set for the UAV cluster.

[0169] Step S5 obtains the current actual state of the formation and aligns it with the global reference in S4. The desired velocity command matrix is ​​synthesized through the structure of "global tracking term + relative consistency term + damping term", and then split into control command sets for each UAV.

[0170] Step S5 first acquires the actual state at control time t. k By utilizing mature GNSS+IMU fusion (such as Kalman filtering), the spatial position of each UAV is estimated in real time in the global coordinate system, and the velocity state of each UAV is estimated in real time from sensor data, thus obtaining the following:

[0171] Control time t k The actual location matrix of the drone swarm:

[0172] Where the i-th row of the matrix represents the i-th UAV at time t k The actual three-dimensional position vector:

[0173] p i (t k )=[x i (t k ),y i (t k ),z i (t k )];

[0174] In the formula,

[0175] x i (t k ) represents the i-th UAV t in the global coordinate system. k The x-axis coordinate at time;

[0176] y i (t k ) represents the i-th UAV t in the global coordinate system. k The y-coordinate at time;

[0177] z i (t k ) represents the i-th UAV t in the global coordinate system. k The z-axis coordinate at time;

[0178] Control time t k The actual velocity matrix obtained below:

[0179] Where the i-th row of the matrix corresponds to the three-dimensional velocity vector of the i-th UAV:

[0180] v i (t k ) = [v i,x (t k ),v i,y (t k ),v i,z (t k )];

[0181] In the formula,

[0182] v i,x (t k ) represents the i-th drone t k The velocity component along the x-axis at any given moment;

[0183] v i,y (t k ) represents the i-th drone t k The velocity component along the y-axis at any given moment;

[0184] v i,z (t k ) represents the i-th drone t k The velocity component along the z-axis at any given moment;

[0185] Step S5 then performs global tracking error calculation based on the target formation reference model P. ref By comparing the actual position of each UAV with its corresponding position in the target formation reference model point by point, the global position error matrix is ​​obtained:

[0186] E g(t k ) = P ref -P(t k );

[0187] E g Measuring the deviation of each drone from its desired position in the target formation is the dominant factor driving formation convergence.

[0188] Step S5 then performs relative consistency error calculation. Based on the edge weight matrix W obtained in step S2, the degree matrix is ​​further constructed, and its expression is:

[0189] D = diag(∑ j W ij );

[0190] This matrix is ​​a diagonal matrix, where the i-th diagonal element is:

[0191] D ii =∑ j W ij ;

[0192] This represents the sum of connection weights between the i-th drone and all other drones;

[0193] The edge weight matrix W and the degree matrix D together construct the graph Laplacian matrix L, which is expressed as follows:

[0194] L = DW;

[0195] Thus, the relative consistency error matrix is ​​obtained:

[0196]

[0197] When E r When = 0, it means that the relative layout between nodes is consistent with the reference.

[0198] Step S5 then performs the desired speed command synthesis, given a non-negative gain K. p ,K r ,K d A desired velocity command matrix is ​​established based on three non-negative gains, which correspond to the local position error, relative geometric error, and velocity state adjustment intensity, respectively. The specific determination process is as follows:

[0199] K p The gain is determined by the system's convergence requirements for position error. By analyzing the deviation between the UAV and the target formation and combining the dynamic capabilities of the flight control system in position adjustment, a suitable gain is set so that the UAV can approach the target position within a reasonable time scale.

[0200] K rThe geometric consistency requirement for maintaining the formation is determined by analyzing the connectivity of the local adjacency graph and the importance of relative position constraints, so that the UAV can converge to the target formation without destroying the stable structure of its geometric formation.

[0201] K d The speed feedback gain is determined by the system's requirements for speed suppression and smooth control. By examining the dynamic characteristics of the UAV and the response limitations of the actuators, an appropriate speed feedback gain is set to suppress excessively rapid speed changes and ensure the smoothness and safety of formation maneuvers.

[0202] The expression for the desired speed command matrix is:

[0203]

[0204] In the formula,

[0205] The first term K p E g : To push various agencies toward their P ref The target point in the curve converges;

[0206] Second term - K r E r : Suppress relative geometric deviations and maintain relative consistency under adjacent intensities embodied by W;

[0207] Third item - K d V: Velocity damping, which suppresses oscillations and overshoot, and ensures smooth convergence.

[0208] Step S5 finally encapsulates and outputs the control instruction set, converting V... cmd (t k The command is broken down into three-dimensional desired velocity instructions for each UAV, expressed as follows:

[0209]

[0210] In the formula,

[0211] U(t k ): At time t k The obtained drone swarm control command set;

[0212] u i (t k ): The i-th drone at time t k The control command vector;

[0213] Indicates the time of the i-th drone at time t k The desired velocity vector is calculated by the control law and serves as the direct input to the flight control system.

[0214] This indicates each control command vector u i (t k ) is a three-dimensional vector containing three components: a velocity component along the x-axis, a velocity component along the y-axis, and a velocity component along the z-axis;

[0215] i = 1, ..., N: indicates that there are N drones in the cluster, and each drone is at time t k Each has a corresponding control command;

[0216] The matrix form of this expression is:

[0217]

[0218] In the formula,

[0219] The i-th row corresponds to the three-dimensional velocity command of the i-th UAV;

[0220] Each column represents the velocity components in the x, y, and z directions.

[0221] U(t k This refers to the control command set of the drone swarm, which can be directly sent to the speed closed loop of each drone (or the flight controller can convert the desired speed into the desired attitude / throttle and other low-level quantities).

[0222] Traditional methods commonly fall into two categories:

[0223] Single tracking based solely on a global reference point (V cmd ∝P ref -P), ignoring the strength of adjacent structures;

[0224] Based solely on fixed topology unification (V cmd ∝-LP), lacking explicit tracking of specific global formation targets.

[0225] Step S5 introduces both a global tracking term and a relatively consistent term induced by the weighted graph structure of S2 into the same control law, and supplements it with velocity damping, thereby achieving an adjustable quantitative balance between formation maintenance and global convergence.

[0226] Example 2:

[0227] This embodiment also provides a computer device applicable to a formation control method for a drone swarm, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the formation control method for a drone swarm as proposed in the above embodiment.

[0228] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a formation control method for a drone swarm as proposed in the above embodiments.

[0229] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0230] If a function is implemented as 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 this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0231] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0232] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0233] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0234] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for formation control of a drone swarm, characterized in that, Includes the following steps: S1. Obtain the initial spatial state of the drone cluster and obtain the initial state vector; S2. Use the initial state vector to construct the proximity constraint and obtain the local adjacency graph; S3. Use the local adjacency graph to perform geometric mapping to obtain the relative geometric position matrix; S4. Use the relative geometric position matrix to generate the global formation and obtain the target formation reference model; S5. Dynamic formation control calculations are performed using the relative geometric position matrix and the target formation reference model to obtain the control command set for the UAV cluster.

2. The method for formation control of a drone swarm according to claim 1, characterized in that, Step S1 first involves multi-source information acquisition, then state estimation and fusion, and finally initial state vector construction. The output shows that the UAV cluster size is N, and GNSS / IMU fusion and time synchronization are completed under a unified timestamp t0. The individual state vector of UAV i is expressed by the following formula: In the formula, s i (t0) represents the single-machine state vector of the i-th UAV, which is a 12-dimensional vector; This represents the position estimate after GNSS / IMU fusion; This represents the depth estimate after fusion; This represents the angular velocity of the IMU after calibration and alignment, expressed in ENU; Indicates IMU linear acceleration; Therefore, the initial state vector is: In the formula, S0 represents the initial state vector of the entire formation at time t0, which is a 12N-dimensional vector; N represents the number of drones.

3. The formation control method for a drone swarm according to claim 1, characterized in that, Step S2 first performs pairwise quantization measurement calculation, then performs joint proximity constraint determination, then constructs the adjacency matrix and edge set, then performs edge weight mapping and weighted graph formation, and finally outputs the local adjacency graph. In step S2, when performing pairwise quantization calculations, for any two distinct machines i,j (i≠j), the following values ​​are calculated: Spatial distance: d ij =||p i (t0)-p j (t0)||2 represents the geometric proximity of the two machines in the ENU coordinate system; Speed ​​difference: u ij =||v i (t0)-v j (t0)||2 represents the measure of the convergence of motion between the two machines; Angular velocity difference: q ij =||ω i (t0)-ω j (t0)||2 represents the consistency of the attitude change rate of the two aircraft.

4. The formation control method for a drone swarm according to claim 3, characterized in that, In step S2, when performing the joint proximity constraint determination, a threshold R is set. d >0、R v >0、R ω >0, and define the joint constraint set: C = {C1, C2, C3}; Where C1:d ij ≤R d ;C2:u ij ≤R v ;C3:q ij ≤R ω ; In the formula, R d Spatial distance threshold; R v Speed ​​difference threshold; R ω Angular velocity difference threshold.

5. The method for formation control of a drone swarm according to claim 3, characterized in that, In step S2, when performing the joint proximity constraint determination, a determination indicator is set: χ ij =1 indicates that the two machines satisfy the constraints of spatial proximity and dynamic consistency, allowing the establishment of an edge between nodes i and j; χ ij =0 indicates no edge connection.

6. The formation control method for a drone swarm according to claim 3, characterized in that, In step S2, when constructing the adjacency matrix and edge set, the adjacency matrix is ​​A, and its expression is: A ij =x ij ,A ii =0; A∈{0,1} N×N Used to represent the edge relationships within a drone swarm; A ij The element in the i-th row and j-th column of the matrix is ​​used to indicate whether there is a connection between drones i and j; A ii =0: The diagonal elements are always zero, indicating that the drone does not establish an edge connection with itself.

7. The method for formation control of a drone swarm according to claim 3, characterized in that, In step S2, when constructing the adjacency matrix and edge set, the edge set is: E local ={(i,j)∣A ij =1}; E local The edge set of the local adjacency graph, containing all pairs of unmanned aerial vehicles that satisfy the joint constraints; (i,j): Represents an undirected edge between drones i and j; A ij =1: As a filtering condition, it means that i and j are determined to have a connection in the adjacency matrix; E local Summarize all node pairs that satisfy the joint constraints to form the structural skeleton of the graph.

8. The formation control method for a drone swarm according to claim 3, characterized in that, In step S2, when performing edge weight mapping and weighted graph formation, a monotonically decreasing normalized mapping is set. This mapping reflects the proximity strength, and its expression is: In the three expressions, φ d ,φ v ,φ ω These expressions reflect their relative proximity in the dimensions of spatial distance, velocity difference, and angular velocity difference, respectively. All three expressions are monotonically decreasing, indicating that the closer the difference value is to zero, the closer the proximity intensity is to 1.

9. The method for formation control of a drone swarm according to claim 3, characterized in that, In step S2, when performing edge weight mapping and weighted graph formation, a non-negative coefficient α is manually set based on task requirements. d ,α v ,α ω α d +α v +α ω =1, used for linear weighted fusion among the three neighboring indices, the resulting edge weight expression is: W ij =x ij (a d f d (d ij )+a v f v (u ij )+a ω f ω (q ij )); When χ ij When W = 1, ij ∈[0,1] reflects the comprehensive proximity strength between the two machines; when χ ij When W = 0, ij =0 indicates no edge connection, and the edge weight is used as a priority reference when splicing geometric relationships and robust matching in step S3.

10. The method for formation control of a drone swarm according to claim 3, characterized in that, In step S2, when outputting the local adjacency graph, for all drone nodes in the cluster, the node set is set as V = {1,2,…,N}. Combining the adjacency matrix A and the edge weight matrix W, the expression for the local adjacency graph is:

Images