Unmanned aerial vehicle cluster collaborative target encircling elastic control method for coping with network attack

Abstract

The invention discloses an unmanned aerial vehicle cluster collaborative target hunting elastic control method for coping with network attacks, and relates to the field of unmanned aerial vehicles, and the method comprises the steps: constructing a flight model of an unmanned aerial vehicle and a control law of the unmanned aerial vehicle; constructing a potential field function based on the body radius of the unmanned aerial vehicle at the current moment and the communication distance of the unmanned aerial vehicle at the current moment; adopting a distributed Kalman consensus filtering algorithm to obtain motion information of the target at the next moment; based on the motion information of each unmanned aerial vehicle in the unmanned aerial vehicle cluster at the current moment, the upper and lower bounds of the network attack at the current moment and the motion information of the target at the next moment, obtaining the motion information of each unmanned aerial vehicle at the next moment by taking a potential field function as a constraint and adopting a flight model of the unmanned aerial vehicle and a control law of the unmanned aerial vehicle; and controlling the unmanned aerial vehicles to surround the target according to the motion information of each unmanned aerial vehicle at the next moment until the surrounding of the target is completed. According to the invention, the accuracy, flexibility and anti-interference performance of unmanned aerial vehicle cluster cooperative target encircling elastic control can be improved.

Description

Technical Field

[0001] The present application relates to the field of drones, and in particular to a flexible control method for collaborative target capture by a drone cluster in response to network attacks. Background Art

[0002] In recent years, distributed control of multiple drones has shown promising applications in various fields, such as formation control, containment control, and capture control. Among these applications, capture control is particularly important. Its core goal is to guide multiple drones to form a stable target-centric configuration through carefully designed controllers. However, drones, as typical cyber-physical systems, are highly dependent on communication networks for data transmission and processing. Without effective defense strategies, drones are highly vulnerable to malicious cyberattacks, which can severely damage the performance of drone swarms and hinder the intended capture mission.

[0003] In some cases, cyberattack signals are often modeled as perturbations to actuators or sensors, but less attention has been paid to their potential negative impact on communication networks. Compared to attacks targeting single control units, cyberattacks are often more destructive. Furthermore, most approaches to cyberattacks rely on observer strategies to observe and compensate for attack signals, but these methods are computationally complex. Summary of the Invention

[0004] The purpose of this application is to provide a flexible control method for cooperative target capture of drone swarms in response to network attacks, which can improve the accuracy, flexibility and anti-interference of the flexible control of cooperative target capture of drone swarms.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] This application provides a flexible control method for cooperative target capture by drone swarms in response to network attacks, including:

[0007] Constructing a flight model and a control law of a UAV; constructing the control law of the UAV by introducing a dead zone function;

[0008] Obtain the motion information of each drone in the drone cluster at the current moment, the motion information of the target at the current moment, the body radius of the drone at the current moment, the communication distance of the drone at the current moment, and the upper and lower bounds of the network attack at the current moment;

[0009] Based on the current UAV body radius and the current UAV communication distance, a potential field function is constructed;

[0010] Based on the motion information of the current target and the motion information of each drone, a distributed Kalman consensus filter algorithm is used to predict the motion information of the target at the next moment corresponding to each drone;

[0011] Based on the motion information of each drone in the drone cluster at the current moment, the upper and lower bounds of the network attack at the current moment, and the motion information of the target at the next moment, with the potential field function as a constraint, the drone's flight model and the drone's control law are used to obtain the motion information of each drone at the next moment. According to the motion information of each drone at the next moment, the drones are controlled to encircle the target until the encirclement of the target is completed.

[0012] According to the specific embodiments provided in this application, this application has the following technical effects:

[0013] The present application provides a flexible control method for cooperative target capture of a swarm of drones in response to network attacks. The motion information of the target obtained at the current moment and the motion information of each drone are filtered by a distributed Kalman consensus filtering algorithm, which can suppress the noise in the motion information of the target, so that the predicted motion information of the target at the next moment is accurate, providing a basis for subsequent precise calculations. By taking the potential field function as a constraint and adopting the flight model and control law of the drone, the motion information of the drone at the next moment can be used to capture the target in real time according to the motion information of the target at the current moment, without defining a precise formation, thereby improving the flexibility of the flexible control of cooperative target capture of the swarm of drones. By constructing a control law of the drone that introduces a dead zone function for calculation, the control law of the drone uses the dead zone function to dynamically suppress network attacks, which can ensure the motion information of the drone at the next moment and improve the anti-interference ability of the flexible control of cooperative target capture of the swarm of drones. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0015] Figure 1 This is a schematic diagram of the roundup process for this application.

[0016] Figure 2 A flowchart of a flexible control method for collaborative target capture by a drone swarm in response to network attacks provided in one embodiment of the present application.

[0017] Figure 3A detailed flowchart of a flexible control method for cooperative target capture by a drone swarm in response to network attacks provided in one embodiment of the present application.

[0018] Figure 4 A regional schematic diagram of the potential field function provided in one embodiment of the present application.

[0019] Figure 5 A schematic diagram of an attack mode provided in one embodiment of the present application.

[0020] Figure 6 Schematic diagram of the target state estimation results, where (a) is the position estimation, (b) is the velocity estimation, and (c) is the control input estimation.

[0021] Figure 7 Schematic diagram of random perturbation with an upper bound of 5 in scenario 1.

[0022] Figure 8 Schematic diagram of the spatial distribution of collaborative roundups at different times in scenario 1, where (a) is the spatial distribution at t = 0s, (b) is the spatial distribution at t = 15s, (c) is the spatial distribution at t = 30s, and (d) is the spatial distribution at t = 50s.

[0023] Figure 9 Schematic diagram of the capture error in scenario 1.

[0024] Figure 10 Schematic diagram of the relative distance between drones in scenario 1.

[0025] Figure 11 Schematic diagram of the flight speed of the UAV and the target in scene 1.

[0026] Figure 12 Schematic diagram of the heading angle of the drone in scene 1.

[0027] Figure 13 Schematic diagram of the flight track angle of the UAV in scene 1.

[0028] Figure 14 Schematic diagram of the control input of the drone in scene 1.

[0029] Figure 15 Schematic diagram of control input without inhibition strategy in scenario 1.

[0030] Figure 16 Schematic diagram of random perturbation with an upper bound of 10 in scenario 2.

[0031] Figure 17Schematic diagram of the spatial distribution of collaborative roundups at different times, where (a) is the spatial distribution at t=0s, (b) is the spatial distribution at t=15s, (c) is the spatial distribution at t=30s, and (d) is the spatial distribution at t=50s.

[0032] Figure 18 Schematic diagram of the capture error in scenario 2.

[0033] Figure 19 Schematic diagram of the relative distance between drones in scene 2.

[0034] Figure 20 Schematic diagram of flight speed in scene 2.

[0035] Figure 21 Schematic diagram of the heading angle of the drone in scene 2.

[0036] Figure 22 Schematic diagram of the flight track angle of the UAV in scene 2.

[0037] Figure 23 Schematic diagram of the control input of the drone in scene 2.

[0038] Figure 24 Schematic diagram of control input without inhibition strategy in scenario 2.

[0039] Figure 25 Schematic diagram of the spatial distribution of collaborative roundup at different times in scenario 3, where (a) is the spatial distribution at t = 15s, (b) is the spatial distribution at t = 29s, (c) is the spatial distribution at t = 31s, and (d) is the spatial distribution at t = 50s.

[0040] Figure 26 Schematic diagram of the roundup error in scenario 3.

[0041] Figure 27 Schematic diagram of flight speed in scene 3.

[0042] Figure 28 Schematic diagram of the heading angle of the drone in scene 3.

[0043] Figure 29 Schematic diagram of the flight track angle of the UAV in scene 3.

[0044] Figure 30 Schematic diagram of the control input of the drone in scene 3. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0046] In roundup or escort missions, a swarm of drones is often required to encircle a target within a convex hull formed by its own drones. Traditional roundup control methods predefine the desired drone formation based on a time-varying formation function. Based on this, a control law with a formation function compensation term is designed to drive the swarm to encircle the target. However, given the complexity of real-world flight environments, designing formation parameters that meet specific mission requirements is difficult.

[0047] Typical network attacks fall into two categories: False Data Injection (FDI) and Denial of Service (DoS). FDI generally occurs when an attacker infiltrates a drone's communication network and sends false data to the receiving unit, tampering with the actual state information. This misleads other drones into using the erroneous data to resolve state information, disrupting the swarm system's coordination and potentially leading to mission failure or even hijacking. DoS occurs when an attacker gains access to a system and then blocks the information flow of the communication network by sending a large amount of invalid information, preventing the system from exchanging state information. Compared to DoS, FDI is more covert. For DoS, related research often assumes a certain dormant period for the attack and then utilizes event-triggered mechanisms to transform the problem into a cooperative control problem under intermittent communication conditions. A common solution for FDI is to construct an observer to estimate the attack signal and then implement resilient control through compensation strategies.

[0048] When it comes to the capture problem, two control strategies are commonly used, depending on the cooperative or non-cooperative nature of the target. One is to treat the target as a friendly agent and solve it through tracking-inclusion control. The other is to estimate the motion state of the non-cooperative target based on an observer, and then use a controller to drive the swarm to establish a stable configuration centered on the target. However, due to the complexity of the task environment, it is difficult to define universal formation parameters. Therefore, it is necessary to design a new capture control strategy that is free of geometric configuration constraints.

[0049] Based on this, this application proposes a flexible control method for cooperative target capture of drone swarms in response to network attacks. By designing a controller based on self-organizing rules, it is possible to achieve the capture of maneuverable targets under network attacks and avoid collisions. When some drones fail, the remaining drones can quickly form a new configuration and continue to complete the capture mission. The specific capture process is as follows: Figure 1 shown.

[0050] The convex hull formed by the clusters is defined as:

[0051] Define the target position p T The distance from the convex hull is:

[0052] Obviously, when the target is surrounded by a swarm of drones, that is, p T ∈con(p), H(p)=0.

[0053] Based on the cluster self-organization rules, scalable UAV cluster collaborative target capture control under dynamic topology is realized, and the state meets the following conditions:

[0054] (1) The target can be surrounded by a swarm of drones:

[0055] (2) Cluster speeds tend to be consistent: q i (t) is the speed of the i-th UAV at time t, q j (t) is the speed of the i-th UAV at time t, q T (t) is the speed of the target at time t.

[0056] (3) Distance constraints between machines: in, is the minimum safety distance, ||p ij || is the relative distance between the i-th UAV and the j-th UAV, and μ is the communication distance of the UAVs.

[0057] This application solves the problem of coordinated target capture and control of drone clusters when communications are subject to network attacks.

[0058] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0059] In an exemplary embodiment, Figure 2 and Figure 3As shown, a flexible control method for collaborative target capture of a drone cluster in response to network attacks is provided. The method is executed by a computer device, and can be specifically executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to a server as an example for illustration, including the following steps S1 to S5.

[0060] Step S1: constructing a flight model of a UAV and a control law of the UAV; the control law of the UAV is constructed by introducing a dead zone function.

[0061] Specifically, use pictures Represents the communication interaction between drones. and They are node sets and communication edges respectively. It's a picture The adjacency matrix, a ij is the communication weight. If (i,j)∈ε, a ij >0; otherwise, a ij =0. is the Laplace matrix, where (l ij The neighbor set of the drone is defined as the element in the i-th row and j-th column of the Laplace matrix: M i (t)={j∈{1,2,…,n},||p ij (t)||≤μ,j≠i}.

[0062] As a typical cyber-physical system, drones primarily rely on inter-machine communication to transmit state information. However, during swarming missions, they may be vulnerable to cyber attacks from non-cooperative targets. These attackers manipulate the drone system's controllers and communication links to inject false data, tampering with real state information. This misleads control and decision-making, potentially disrupting the swarm's stable configuration and potentially rescuing the target.

[0063] To this end, consider the speed channel of the drone being attacked by a network, and model the attack in the case of false data injection as: Q ij (t) = q j (t)+ζ ij (t). Where, Q ij (t) is the speed information of the jth UAV received by the i-th UAV after being attacked by the network at time t, q j (t) is the real speed information of the j-th UAV, ζ ij (t) is a bounded false signal, and the bounded false signal ζ ij (t) is bounded and satisfies c≥0. Figure 5As shown in the figure, considering the bidirectional communication characteristics of the network and the complexity of the attack, it is assumed that in the same communication channel, the attacks on the two nodes are different, that is, ζ ij (t)≠ζ ji (t). And for different upper bounds d of attack signals ij , take d=max{d ij ,i,j=1,2,…,n(i≠j)}.

[0064] Furthermore, the flight model of the UAV is expressed as:

[0065]

[0066] in, is the axial position of the i-th UAV at time t, p ix (t) is the axial position of the X axis of the i-th UAV at time t, p iy (t) is the Y-axis position of the i-th UAV at time t, p iz (t) is the axial position of the Z axis of the i-th UAV at time t; is the axial velocity of the X axis of the i-th UAV at time t; V i (t) is the speed of the i-th UAV at time t; ψ i (t) is the heading angle of the i-th UAV at time t; γ i (t) is the flight path angle of the i-th UAV at time t; is the axial velocity of the Y axis of the i-th UAV at time t; is the axial velocity of the Z axis of the i-th drone at time t; is the derivative of the velocity of the i-th UAV at time t; g is the acceleration due to gravity; T i (t) is the thrust of the i-th UAV at time t; D i (t) is the resistance of the i-th UAV at time t; m i is the mass of the i-th UAV; is the derivative of the heading angle of the i-th UAV at time t; L i (t) is the lift of the i-th UAV at time t; φ i (t) is the tilt angle of the i-th UAV at time t; is the derivative of the flight path angle of the i-th UAV at time t; n i (t) is the overload of the i-th UAV at time t.

[0067] Furthermore, in order to realize the cluster's self-organizing configuration to capture the target, the control input of the system is defined as follows: the navigation item of the i-th UAV at time t The potential field term of the i-th UAV at time t The speed of the i-th UAV at time t is consistent with the The control law of the UAV is expressed as:

[0068]

[0069] Among them, u i (t) is the auxiliary control input of the i-th UAV at time t; u T (t) is the control input of the target at time t; c1>0, c2>0 are the control gains; is the position error between the i-th UAV and the target at time t; is the speed error between the i-th UAV and the target at time t; is the navigation item of the i-th UAV at time t; a ij (t) is the communication weight between the i-th UAV and the j-th UAV at time t; is along position p i The gradient of c The potential field function used by the UAV for distance adjustment; is the position error between the i-th UAV and the j-th UAV at time t; is the potential field term of the i-th UAV at time t; is the dead zone function, M i is the number of neighboring drones of the i-th drone, ζ is the false signal of network attack; a il (t) is the communication weight between the i-th UAV and the l-th UAV at time t; is the error between the speed of the lth UAV after being attacked and the target speed received by the i-th UAV at time t; a jl (t) is the communication weight between the jth UAV and the lth UAV at time t; is the error between the speed of the j-th UAV and the target speed at time t; is the error between the speed of the lth UAV after being attacked and the target speed received by the jth UAV at time t; is the velocity consistency term of the i-th UAV at time t.

[0070] Furthermore, the target state value can be estimated asymptotically. The position error between the i-th UAV and the target at time t is expressed as:

[0071]

[0072] The expression of the speed error between the i-th UAV and the target at time t is:

[0073]

[0074] Among them, p i (t) is the position of the i-th UAV at time t; pT (t) is the position of the target at time t; q i (t) is the speed of the i-th UAV at time t, Among them, q ix (t) is the speed of the X axis of the i-th UAV at time t, q iy (t) is the velocity of the Y axis of the i-th UAV at time t, q iz (t) is the speed of the Z axis of the i-th UAV at time t; q T (t) is the speed of the target at time t.

[0075] Furthermore, the expression of the communication weight between the i-th UAV and the j-th UAV is:

[0076]

[0077] Among them, ||p ij || is the relative distance between the i-th UAV and the j-th UAV; μ is the communication distance of the UAV; χ is the attenuation factor.

[0078] Furthermore, the expression of the dead zone function is:

[0079]

[0080] Where W is the adjacent speed error variable of the dead zone function; W (f) is the fth component; d is the upper bound of network attack.

[0081] This application is based on the dead zone function and can handle the external interference of neighboring drone speed caused by network attacks. By modifying the control law, it can prevent ζ ij (t) can avoid the drastic fluctuation of the control signal caused by the system, and ensure that the stability of the system is not affected. In addition, the design of the dead zone function can adaptively adjust the number of neighboring drones |M i |The impact of (threshold value with |M i | changes), even if the communication channels in the cluster are subjected to large-scale network attacks, the robustness of control can be ensured under different network scales.

[0082] For the sum of false signals received by the i-th drone, the following inequality must be satisfied.

[0083] -|M i |d<ζ i (t)<|M i |d.

[0084] Where, is the sum of all false signals received by the i-th UAV from its neighbors.

[0085] Further, consider W(f) =q (f) +ζ (f) , then The positive definiteness of is analyzed as follows:

[0086] Case 1: When W (f) >|M i |d, at this time Because W (f) Exceeding the positive boundary, the adjusted value W (f) -|M i |d still with q (f) Same direction (because the attack was cut off), so

[0087] Case 2: When W (f) <-|M i |d, at this time Because W (f) Below the negative boundary, the adjusted value W (f) +|M i |d still with q (f) Same direction (interference is compensated), so

[0088] Case 3: When |W (f) |≤|M i |d, Obviously

[0089] In summary, This property provides key support for subsequent stability analysis (non-increasing property of Lyapunov function).

[0090] Step S2: Obtain the current motion information of each drone in the drone cluster, the current target's motion information, the current drone radius, the current drone communication range, and the current upper and lower bounds of the network attack. Motion information includes position, velocity, and acceleration.

[0091] Step S3: Construct a potential field function based on the current UAV body radius and the current UAV communication distance.

[0092] Furthermore, the safety control of the drone cluster as the basis for achieving collaborative tasks is an issue that must be considered. To this end, this application designs a potential field function that can simultaneously achieve collision avoidance and network connectivity maintenance to meet the distance constraint (C3). Figure 4 As shown, the expression of the potential field function is:

[0093]

[0094] Among them,c (·) is the potential field function; is the action function;κ is the input variable of the potential field function; r is the collision avoidance buffer distance; μ is the communication distance of the UAV; is the minimum safe distance, r1 is the radius of the drone.

[0095] when When ||p ij (t)||∈( r ,μ), the UAVs are affected by gravity and maintain communication network connectivity. Based on this, the expected distance constraint is achieved through the integrated design of different potential fields.

[0096] Step S4: Based on the motion information of the target at the current moment and the motion information of each UAV, a distributed Kalman consensus filter algorithm is used to predict the motion information of the target at the next moment corresponding to each UAV;

[0097] Furthermore, the calculation formula for the motion information of the target at the next moment corresponding to each UAV is:

[0098]

[0099] in, is the motion information of the target at time t+1 corresponding to the i-th UAV; is the estimated value of the motion information of the target at time t corresponding to the i-th UAV; is the rate of change of the estimated value of the motion information of the target at time t corresponding to the i-th UAV; Δt is the time interval between time t and time t+1; A is the system matrix; K i is the gain matrix of the i-th UAV; w i (t) is the observation value of the target by the i-th UAV at time t; H i is the observation matrix; σ is a positive constant; P i is the covariance of the prior estimation error of the i-th UAV; M i is the number of neighboring drones of the i-th drone; is the estimated value of the motion state of the target at time t corresponding to the j-th UAV.

[0100] Specifically, in order to address the impact of uncertain data in a noisy environment on the cluster's acquisition of the accurate motion state of the target, this application adopts a distributed Kalman consensus filtering algorithm to achieve noise suppression and state estimation of the maneuvering target, and applies the estimation results to the subsequent collaborative capture control algorithm.

[0101] Assume the state equation of the target is:

[0102] The observation equation of UAV i is: w i (t) = H i ξ(t)+υ i (t).

[0103] Where, is the target state to be estimated at time t; A is the system matrix, B is the noise input matrix; w i (t) is the observation value of the target state by the i-th UAV at time t; H i is the observation matrix; ω(t),υ i (t) are Gaussian white noise, ω(t) represents the process noise at time t and υ i (t) represents the measurement noise of the i-th UAV at time t, and the covariance satisfies:

[0104]

[0105] Where E(·) is the covariance of the noise, R i is the covariance matrix, δ kl is the Kronecker delta function. If k = l, δ kl =1; if k≠l, δ kl =0.

[0106] Step S5: Based on the motion information of each drone in the drone cluster at the current moment, the upper and lower bounds of the network attack at the current moment, and the motion information of the target at the next moment, with the potential field function as a constraint, the drone's flight model and the drone's control law are used to obtain the motion information of each drone at the next moment. According to the motion information of each drone at the next moment, the drones are controlled to encircle the target until the encirclement of the target is completed.

[0107] Furthermore, based on the motion information of each UAV at the current moment, the upper and lower bounds of the network attack, and the motion information of the target at the next moment, the control law of the UAV is adopted to obtain the auxiliary control input of each UAV at the next moment.

[0108] According to the auxiliary control input of each UAV at the next moment, the actual control input of each UAV at the next moment is calculated.

[0109] Based on the actual control input of each UAV at the next moment, the flight model of the UAV is used to obtain the motion information of each UAV at the next moment.

[0110] Furthermore, the calculation formula for the actual control input of each UAV at the next moment is:

[0111]

[0112] in, is the actual control input of the i-th UAV at time t; n i (t) is the overload of the i-th UAV at time t; u iz (t) is the auxiliary control input of the Z axis of the i-th UAV at time t; g is the acceleration of gravity; γ i (t) is the flight path angle of the i-th UAV at time t; u ix (t) is the auxiliary control input of the X axis of the i-th UAV at time t; ψ i (t) is the heading angle of the i-th UAV at time t; u iy (t) is the auxiliary control input of the Y axis of the i-th UAV at time t; φ i (t) is the tilt angle of the i-th UAV at time t; T i (t) is the thrust of the i-th UAV at time t; m i is the mass of the i-th UAV; D i (t) is the drag of the i-th UAV at time t.

[0113] In an exemplary embodiment, a stability analysis is performed on steps S1 to S5 of the present application, and the specific analysis is as follows.

[0114] Considering that the system consists of n drones and a dynamic target, the drone model adopts the flight model of the drone in step 1. The FDI received by the drone is expressed as Q ij (t) represents and satisfies the inequality -|M i |d<ζ i (t)<|M i |d. Under the control law, the UAV swarm can achieve coordinated capture of maneuvering targets and meet distance constraints.

[0115] 1. Prove that the target can be captured by drones.

[0116] Define the round-up position and velocity errors:

[0117]

[0118] The time derivative of the capture velocity error is obtained:

[0119]

[0120] Further derivation yields:

[0121]

[0122] Considering the undirected nature of cluster communication, it is easy to know that:

[0123]

[0124] but can be transformed into:

[0125] Right now:

[0126] From c1>0, c2>0, it is clear that It is proved that drone swarms can realize the capture of mobile targets.

[0127] 2. Prove that the cluster has consistent speed.

[0128] Define the Lyapunov candidate function as follows:

[0129]

[0130] Taking the derivative with respect to time, we get:

[0131]

[0132] Substituting the control law into the equation, we can obtain:

[0133]

[0134] definition

[0135] Convert the above formula into:

[0136]

[0137] According to the previous analysis of the properties of the dead zone function, That is, V(t)≤V(0). For the set:

[0138] According to Lasalle's invariance principle, solutions starting from Ω tend to converge to a set:

[0139] Then we can get, That is, q1(t)=q2(t)=…=q T (t), indicating that the speeds of all drones can converge to the target speed under cyber attack.

[0140] 3. Prove that there is no collision between drones.

[0141] Let's assume that at time t1, UAV k collides with UAV l, that is Then we have:

[0142]

[0143] According to the collection It can be seen that:

[0144] a kl (t)Ψ c (||p k (t)-p l (t)||)≤V c (t)<V(t)≤V(0).

[0145] However, when Sometimes, there is c (||p k (t)-p l (t)||)→∞. Obviously, this is This assumption is contradictory. Therefore, the assumption is not true. This means that there will be no collision between drones during the roundup process.

[0146] Furthermore, according to the potential function design, when ||p ij (t)||∈(r1+d s ,μ), the gravity forces the distance between the UAVs to decrease and maintain connectivity.

[0147] In summary, the control strategy designed in this paper can realize the coordinated target capture control of drone clusters under FDI attack and avoid collision.

[0148] Note 3: It should be noted that when ζ ij When (t) = 0, the control law can be degenerated into:

[0149]

[0150] Similar to the above proof, it is easy to see that the control law can achieve the coordinated capture of maneuvering targets under FDI-free cyber attacks. This means that the control law designed in this paper is an extension of this type of swarm control problem. The deadband function only suppresses false signals and does not affect the stability of the closed-loop system.

[0151] In an exemplary embodiment, steps S1 to S5 of the present application are simulated and verified, and the specific simulation is as follows.

[0152] The effectiveness of the proposed algorithm was verified through simulations. For scenarios 1 and 2, the effectiveness of the UAV collaborative capture control strategy was verified under spoofing signal attacks with different upper bounds. For scenario 3, the robustness of the control algorithm was verified under the condition of partial node failure, specifically whether the system could reconfigure the swarm configuration and complete the capture mission.

[0153] The performance parameters of the drone can be customized. Drag and lift are calculated using the following equations.

[0154]

[0155] Where ρ0 = 1.225 kg / m 3 、S w =1.37m 2 and C D0 =0.02 are respectively the atmospheric density, wing area and zero lift drag coefficient. n =1 and k d =0.1 represents the load factor and resistance coefficient respectively. m i = 20kg, χ = 0.5, r1 = 10m and μ = 100m. The actual control variables satisfy the constraints: -1.5 ≤ n i ≤2.0、10N≤T i ≤125N and -80°≤φ i ≤80°.

[0156] The control input of the target is: Covariance matrix and R i =0.5I3, the initial estimated covariance matrix P0 = 30I9. The control parameters are selected as: c1 = 0.7, c2 = 2.5. The initial speed of the drone is randomly selected within [55, 70] m / s. The state estimation results of the target are as follows Figure 6 As shown in the figure, in the presence of noise, each UAV achieved good results in the axial position, velocity and acceleration of the target.

[0157] Scenario 1: Target roundup under FDI attack (d=5).

[0158] First, the speed deception signal is set as Figure 7 The upper bound shown is a random perturbation of 5. The simulation results are shown in Figures 8-15 As shown. Among them, Figure 8 The spatial distribution of the cooperative roundup system at different times is described. Figure 9 It can be seen that under the conditions of measurement noise and FDI attack, the capture error can converge to a reasonable range, that is, the target is successfully captured by the drone cluster. Figure 10 It shows that the relative distances between the UAVs all satisfy the distance constraint (C3), indicating that the design of the potential function is effective. Figure 11-13 The dynamic evolution of the swarm's flight speed, heading angle, and flight path angle over time is described. Under the action of the cooperative control strategy, the various state variables can converge to a consistent state, prompting the swarm to form a stable flight configuration. Figure 14 and Figure 15The actual control inputs with and without the disturbance suppression strategy are shown. Clearly, the transformed control inputs are relatively smooth under the influence of multiple disturbances, which is consistent with engineering practice. In contrast, the control inputs without the suppression strategy remain saturated for extended periods, failing to meet the requirements of a practical capture task. Analysis shows that the proposed strategy is well suited to addressing the collaborative target capture control problem in the presence of FDI attacks and sensor measurement noise.

[0159] Scenario 2: Target roundup under FDI attack (d=10).

[0160] Adjust the deception signal to Figure 16 The random disturbance signal with an upper bound of 10 is used to verify the control effect of the system. The simulation results are shown in Figure 17-24 As shown. Figure 17-18 It can be seen that the drone cluster can still capture the target. Figures 19-22 It can be seen that the evolution process of each state quantity is relatively smooth. Figure 23-24 , which further verifies the effectiveness of our control strategy.

[0161] Scenario 3: Target roundup with node changes under FDI attack.

[0162] To demonstrate the distributed self-organizing control strategy to give the cluster the autonomous adjustment function. Based on scenario 2, assuming that UAV 4 fails and loses connection within 30 seconds, the reconstruction capability of the cluster system is verified. The simulation results are as follows Figures 25-30 As shown. Figure 25-26 As shown in Figure 3, 30 seconds ago, the four drones achieved the capture of the target in a tetrahedron configuration. Figure 25 As shown in (c), the target is outside the flight convex hull, and the capture error (23) fluctuates. Then, under the control law, the target is captured by the triangular cluster, and the capture error converges quickly. This shows that the cluster has good reconstruction capabilities. Figures 27-30 The flight speed, heading angle, flight trajectory angle and actual control input variables shown are all able to converge quickly and reach consistency after a fault occurs.

[0163] The beneficial effects of the flexible control method for cooperative target capture of drone swarms in response to network attacks proposed in this application are mainly manifested in:

[0164] (1) In order to realize the target capture without geometric constraints, the potential field function is used as a constraint, the UAV flight model and the UAV control law are adopted, so that the UAV's motion information at the next moment can be used in real time to capture the target according to the target's motion information at the current moment. There is no need to define a precise formation, and the cluster can be directly driven to capture the target instead of forming the desired formation first. Therefore, the flexibility of the flexible control of the UAV cluster's collaborative target capture is improved.

[0165] (2) Network attacks are modeled as attacks on the speed channel in the communication network, and a distributed elastic controller with a collision avoidance mechanism is designed. This controller dynamically suppresses the impact of attack signals through a dead-zone strategy with an adaptive threshold. The threshold is dynamically adjusted according to the number of adjacent drones and the upper bound of the attack signal, ensuring control robustness for different network sizes and improving the anti-interference ability of the elastic control of the drone cluster's coordinated target capture.

[0166] (3) Through the distributed Kalman consensus filtering algorithm, the distributed estimation of the target state in a noisy environment is realized, which can suppress the noise in the target's motion information and make the predicted motion information of the target at the next moment accurate, providing a basis for subsequent precise calculations.

[0167] This application addresses the challenges of target state estimation, false data injection (FDI) attack suppression, and collaborative capture control with collision avoidance in noisy environments. By modeling FDI attacks as false signal injections into the velocity channel, a resilient controller with adaptive thresholds is designed to effectively suppress the interference of attacks of varying intensities on the system. Simulation results show that the drone swarm can still achieve stable capture under attacks of varying intensities and node failure scenarios, and the system exhibits good robustness and adaptability in complex mission environments. Future research will further explore joint defense strategies for multimodal hybrid attacks and conduct actual flight verification.

[0168] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0169] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A flexible control method for cooperative target capture by drone swarms in response to network attacks, characterized by: The method for elastically controlling the coordinated target capture of a drone swarm in response to a network attack includes: Constructing a flight model and a control law of a UAV; constructing the control law of the UAV by introducing a dead zone function; Obtain the motion information of each drone in the drone cluster at the current moment, the motion information of the target at the current moment, the body radius of the drone at the current moment, the communication distance of the drone at the current moment, and the upper and lower bounds of the network attack at the current moment; Based on the current UAV body radius and the current UAV communication distance, a potential field function is constructed; Based on the motion information of the current target and the motion information of each drone, a distributed Kalman consensus filter algorithm is used to predict the motion information of the target at the next moment corresponding to each drone; Based on the motion information of each drone in the drone cluster at the current moment, the upper and lower bounds of the network attack at the current moment, and the motion information of the target at the next moment, with the potential field function as a constraint, the drone's flight model and the drone's control law are used to obtain the motion information of each drone at the next moment. According to the motion information of each drone at the next moment, the drones are controlled to encircle the target until the encirclement of the target is completed.

2. The method for controlling the coordinated target capture of a drone swarm in response to a network attack according to claim 1 is characterized in that: The flight model of the UAV is expressed as: in, is the axial velocity of the X axis of the i-th UAV at time t; V i (t) is the speed of the i-th UAV at time t; ψ i (t) is the heading angle of the i-th UAV at time t; γ i (t) is the flight path angle of the i-th UAV at time t; is the axial velocity of the Y axis of the i-th UAV at time t; is the axial velocity of the Z axis of the i-th drone at time t; is the derivative of the velocity of the i-th UAV at time t; g is the acceleration due to gravity; T i (t) is the thrust of the i-th UAV at time t; D i (t) is the resistance of the i-th UAV at time t; m i is the mass of the i-th UAV; is the derivative of the heading angle of the i-th UAV at time t; L i (t) is the lift of the i-th UAV at time t; φ i (t) is the tilt angle of the i-th UAV at time t; is the derivative of the flight path angle of the i-th UAV at time t; n i (t) is the overload of the i-th UAV at time t.

3. The method for controlling the coordinated target capture of a drone swarm in response to a network attack according to claim 1 is characterized in that: The control law of the UAV is expressed as: Among them, u i (t) is the auxiliary control input of the i-th UAV at time t; u T (t) is the control input of the target at time t; c1>0, c2>0 are the control gains; is the position error between the i-th UAV and the target at time t; is the speed error between the i-th UAV and the target at time t; is the navigation item of the i-th UAV at time t; a ij (t) is the communication weight between the i-th UAV and the j-th UAV at time t; is along position p i The gradient of c The potential field function used by the UAV for distance adjustment; is the position error between the i-th UAV and the j-th UAV at time t; is the potential field term of the i-th UAV at time t; is the dead zone function, M i is the number of neighboring drones of the i-th drone, ζ is the false signal of network attack; a il (t) is the communication weight between the i-th UAV and the l-th UAV at time t; is the error between the speed of the lth UAV after being attacked and the target speed received by the i-th UAV at time t; a jl (t) is the communication weight between the jth UAV and the lth UAV at time t; is the error between the speed of the j-th UAV and the target speed at time t; is the error between the speed of the lth UAV after being attacked and the target speed received by the jth UAV at time t; is the velocity consistency term of the i-th UAV at time t.

4. The method for controlling the coordinated target capture of a drone swarm in response to a network attack according to claim 3 is characterized in that: The expression of the position error between the i-th UAV and the target at time t is: The expression of the speed error between the i-th UAV and the target at time t is: Among them, p i (t) is the position of the i-th UAV at time t; p T (t) is the position of the target at time t; q i (t) is the speed of the i-th UAV at time t; q T (t) is the speed of the target at time t.

5. The method for flexible control of drone swarm coordinated target capture in response to network attacks according to claim 3 is characterized in that: The expression of the communication weight between the i-th UAV and the j-th UAV is: Among them, ||p ij || is the relative distance between the i-th UAV and the j-th UAV; μ is the communication distance of the UAV; χ is the attenuation factor.

6. The method for flexible control of drone swarm coordinated target capture in response to network attacks according to claim 3 is characterized in that: The expression of the dead zone function is: Where W is the adjacent speed error variable of the dead zone function; W (f) is the fth component; d is the upper bound of network attack.

7. The method for controlling the coordinated target capture of a drone swarm in response to a network attack according to claim 1 is characterized in that: The expression of the potential field function is: Among them, c (·) is the potential field function; is the action function;κ is the input variable of the potential field function; r is the collision avoidance buffer distance; μ is the communication distance of the UAV; is the minimum safe distance, r1 is the radius of the drone.

8. The method for flexible control of drone swarm coordinated target capture in response to network attacks according to claim 1 is characterized in that: The calculation formula for the motion information of the target at the next moment corresponding to each UAV is: in, is the motion information of the target at time t+1 corresponding to the i-th UAV; is the estimated value of the motion information of the target at time t corresponding to the i-th UAV; is the rate of change of the estimated value of the motion information of the target at time t corresponding to the i-th UAV; Δt is the time interval between time t and time t+1; A is the system matrix; K i is the gain matrix of the i-th UAV; w i (t) is the observation value of the target by the i-th UAV at time t; H i is the observation matrix; σ is a positive constant; P i is the covariance of the prior estimation error of the i-th UAV; M i is the number of neighboring drones of the i-th drone; is the estimated value of the motion state of the target at time t corresponding to the j-th UAV.

9. The method for controlling the flexible capture of a drone swarm in response to a network attack according to claim 1 is characterized in that: Based on the motion information of each drone at the current moment, the upper and lower bounds of the network attack, and the motion information of the target at the next moment, with the potential field function as a constraint, the drone's flight model and control law are used to obtain the motion information of each drone at the next moment. Specifically, the following information is obtained: Based on the motion information of each UAV at the current moment, the upper and lower bounds of the network attack, and the motion information of the target at the next moment, the control law of the UAV is used to obtain the auxiliary control input of each UAV at the next moment; Calculate the actual control input of each UAV at the next moment based on the auxiliary control input of each UAV at the next moment; Based on the actual control input of each UAV at the next moment, the flight model of the UAV is used to obtain the motion information of each UAV at the next moment.

10. The flexible control method for cooperative target capture by drone swarms in response to network attacks according to claim 9 is characterized in that: The calculation formula for the actual control input of each drone at the next moment is: Among them, n i (t) is the overload of the i-th UAV at time t; u iz (t) is the auxiliary control input of the Z axis of the i-th UAV at time t; g is the acceleration of gravity; γ i (t) is the flight path angle of the i-th UAV at time t; u ix (t) is the auxiliary control input of the X axis of the i-th UAV at time t; ψ i (t) is the heading angle of the i-th UAV at time t; u iy (t) is the auxiliary control input of the Y axis of the i-th UAV at time t; φ i (t) is the tilt angle of the i-th UAV at time t; T i (t) is the thrust of the i-th UAV at time t; m i is the mass of the i-th UAV; D i (t) is the drag of the i-th UAV at time t.

Images