A high-speed unmanned aerial vehicle cluster cooperative search method based on correlation positioning error pheromone

Abstract

The application discloses a high-speed unmanned aerial vehicle cluster cooperative search method based on associated positioning error pheromone. I, constructing single-machine AOA passive direction finding positioning technology; II, designing high-speed unmanned aerial vehicle inversion position control law; III, grouping method of bird flock neighborhood interaction; IV, associated positioning error pheromone cooperative search method; V, realizing high-speed unmanned aerial vehicle cluster cooperative search by applying the associated error pheromone method, and outputting a final motion trajectory graph and a pheromone graph. The application provides a method for realizing cooperative search positioning of high-speed unmanned aerial vehicle clusters in a scene lacking prior information, expands application occasions of the high-speed unmanned aerial vehicle clusters, and enhances combat field adaptability; for a three-freedom model based on overload, an inversion position control rate is designed, and accurate position control of a nonlinear aircraft model is realized; the method associates pheromone updating strategies with positioning errors, and introduces various improvement strategies to accelerate method convergence.

Description

Technical Field

[0001] This invention is a high-speed unmanned aerial vehicle (UAV) swarm cooperative search method based on associated positioning error pheromones, belonging to the field of unmanned aerial vehicle technology. Background Technology

[0002] High-speed unmanned aerial vehicles (UAVs) are special tactical drones that use booster rocket engines as their main power unit. They possess advantages such as rapid response, precision strike, and difficulty in interception, making them an important component of modern combat systems. Physically, compared to conventional UAVs, high-speed UAVs have simplified aerodynamic surfaces, making them more suitable for long-range target engagement. However, they also exhibit high speeds but poorer aerodynamic performance. Mathematically, the state variables within a high-speed UAV system are highly coupled, and the number of state variables exceeds the control variables, making it a typical underactuated system that is difficult to control directly using classical control methods. All of these factors significantly increase the difficulty of controlling a single high-speed UAV.

[0003] Inversion control, as an important modern control method, obtains the control laws at each level of the system by recursively constructing the Lyapunov function of the closed-loop system. It is applicable to both linear and nonlinear systems, and therefore has a natural advantage in constructing the single-unit position control law of high-speed unmanned aerial vehicles.

[0004] The mainstream application of high-speed unmanned aerial vehicles (UAVs) lies in long-range guided strikes. Target spatial positions are determined through reconnaissance methods such as the UAV's own sensors or external radar positioning, and then the high-speed UAV is guided to the target to achieve effective destruction. However, in actual combat scenarios, due to enemy interference and limitations in the accuracy of friendly sensors, a single UAV cannot directly obtain target information sufficient for guidance. Furthermore, obtaining prior spatial information through other reconnaissance methods may alert the target, leading to mission failure. Therefore, employing multiple high-speed UAVs to continuously correct target position errors through inter-unit information exchange, and achieving accurate approach to targets without prior information, makes the construction of a high-speed UAV swarm collaborative search system essential.

[0005] A high-speed unmanned aerial vehicle (UAV) swarm system is a system in which a certain number of high-speed UAVs are organized to complete combat missions within an open system architecture. It is based on swarm intelligence, utilizes network communication information, relies on inter-UAV interaction and collaboration, and builds upon individual combat capabilities to construct a highly survivable and autonomous combat system. Through close cooperation, the high-speed UAV swarm expands individual functions, possesses better intelligence and autonomy, and forms the foundation for collaborative search capabilities.

[0006] Currently, mainstream collaborative search methods include centralized and distributed approaches. Centralized methods apply optimization techniques, transforming the search problem into a constrained optimization problem and using optimization methods to find an approximate optimal solution. While simple and convenient, these methods place high demands on the performance and real-time capabilities of the optimization methods and cannot perform searches in the presence of sensor errors, making them offline decision-making methods. In contrast, distributed methods involve parallel online collaborative decision-making, employing techniques such as model predictive control, optimal situational decision-making, and intelligent optimization to continuously correct the search path, resulting in higher accuracy. Among these, intelligent optimization techniques are highly versatile and have been successfully applied in the field of UAV swarm search.

[0007] Ant colonies in nature possess the ability to quickly find food in unknown environments, always moving along the shortest distance between food and their nest. The key to this lies in the pheromones left behind by the ants during their foraging journeys. By comparing differences in pheromone concentration at various points, ant colonies can quickly adjust their movement patterns, pruning low-value routes and maximizing collective energy gains. This unique mechanism of ant colonies has given rise to pheromone search methods, capable of solving shortest path selection problems with complex spatial constraints, and therefore possessing the potential for application in cooperative search.

[0008] To address the aforementioned issues, this invention presents a cooperative search method for high-speed unmanned aerial vehicle (UAV) swarms based on correlated positioning error pheromones. It employs passive direction finding and positioning technology using angle-of-arrival (AOA) to acquire target information. Then, building upon the classical pheromone search method, it introduces cooperative error positioning and time information to construct a correlated error pheromone cooperative search method, generating individual UAV motion strategies. A single-UAV inversion position control law is designed to enable the high-speed UAV swarm to perceive, search for, and approach targets without prior information, expanding the functionality of high-speed UAV swarms and enhancing their adaptability to combat environments. Summary of the Invention

[0009] 1. Purpose of the invention:

[0010] The purpose of this invention is to provide a cooperative search method for high-speed unmanned aerial vehicle (UAV) swarms based on correlated positioning error pheromones. This method enables the cooperative localization and approximation of unknown objects by a high-speed UAV swarm, expanding the functionality of high-speed UAVs and improving battlefield combat capabilities.

[0011] 2. Technical Solution:

[0012] This invention addresses the cooperative search problem in high-speed unmanned aerial vehicle (UAV) swarms. It employs AOA passive direction finding and localization technology to obtain the possible spatial positions and localization errors of the detected objects, and uses a bird flock-like neighborhood interaction technique to group the UAV swarm. Then, an improved pheromone cooperative search method based on correlated localization errors is proposed, and intra-group cross-directional finding and localization errors are used to accelerate pheromone convergence, generating individual UAV motion strategies. Finally, an inversion position control law is designed to control the motion of individual UAVs, the specific steps of which are as follows:

[0013] Step 1: Constructing a single-machine AOA passive direction finding and positioning technology

[0014] First, we design a detection method using sensors mounted on each aircraft. Considering that the spatial orientation of the target to be searched is unknown and may be extremely far away, a passive direction-finding positioning method is used to obtain the position probability:

[0015]

[0016] Where n is the number of aircraft participating in cross-direction finding; exp() represents an exponential function; σ T Let T be the variance of the Gaussian distribution of the target object T; i Let be the measurement result of the i-th high-speed aircraft in the group for the target to be searched; p be the spatial probability of the target to be searched; F(x,y) be the probability density of the associated positions x and y, which can be represented as an error ellipse as follows:

[0017]

[0018] Where (a,b,c,d,e,g) are the undetermined coefficients of each term of the polynomial, which can be obtained by solving F(x,y) using the method of undetermined coefficients; (x T ,y T Let r1 and r2 be the major and minor axes of the error ellipse, respectively. Let be the orientation of the major axis of the error ellipse, and w be an intermediate term in the calculation. After expanding the direction finding result F(x,y), the parameters of the error ellipse can be obtained as follows:

[0019] Δ=b 2 -4ac

[0020] x T =(2cd-be) / Δ

[0021] y T =(2ae-bd) / Δ

[0022]

[0023]

[0024]

[0025]

[0026] Define the major axis r1 as the positioning error, which is used to determine the positioning accuracy of this round.

[0027] Step 2: Design the inversion position control law for high-speed unmanned aerial vehicles

[0028] S21, Unmanned Aerial Vehicle Model Based on Overload Control

[0029] Generally, the actuators of unmanned aerial vehicles (UAVs) are mostly rocket engines, making direct constant-speed control difficult. Therefore, classic inner-loop controllers often choose to directly control each direction. Considering the kinematic characteristics of UAVs in three-dimensional space and neglecting the response time of the inner-loop controller, the following three-degree-of-freedom model can be constructed:

[0030]

[0031] V min ≤V≤V max

[0032] Where, vector P = [xyh] T This indicates the position of the unmanned aerial vehicle's center of mass, with a specific altitude h that is positive vertically upwards; V is the airspeed of the aircraft, limited to the speed range [V]. min V max Between ]; [ψ θ] represent the ballistic deflection angle (default is 0 rad when the aircraft is facing east, right deflection is positive) and the ballistic inclination angle of the unmanned aerial vehicle (default is 0 rad when the aircraft is horizontal to the ground, upward tilt is positive); X1=[xyh V ψ θ] T The initial state constituting the model; u = [n x n y n h ] T The vector represents the overload coefficient of an unmanned aerial vehicle, defined as the ratio of the resultant force of all external forces except gravity to its weight; g is the acceleration due to gravity, with a value of 9.8 m / s². 2 .

[0033] According to formula (4), this system is a typical nonlinear system, with its state variables coupled to each other, making it difficult to directly apply conventional control methods. To simplify the control process, channel decoupling is considered to achieve a one-to-one correspondence between input components and position components in each direction.

[0034] S22, Design of Three-Channel Decoupling Inversion Position Control Law

[0035] To design an inversion position controller, it is necessary to decouple the three-state model of the high-speed aircraft based on overload to obtain decoupled models in three directions, and then design inversion position control laws for the models of the three channels respectively. Without considering the constraints on airspeed V and ballistic angle [ψθ], the original model (4) can be rearranged, and a new state vector can be designed. The new state vector X contains the position quantity P = (x, y, h) and its first derivative. Combining formula (4), the mapping relationship is obtained:

[0036]

[0037] It can be seen that both this mapping and its inverse mapping are smooth, from which the second-order differential equation of the position can be obtained, taking the position component in the x-direction as an example:

[0038]

[0039] in, The partial derivative operation formula (6) explains the specific form of the mapping (5) in the x-direction, from which the differential equation of the state variable X can be obtained, as follows:

[0040]

[0041] Furthermore, equation (7) can be decoupled as follows:

[0042]

[0043]

[0044]

[0045] Where x1 = x, y1 = y, h1 = h; Vectors b1, b2, and b3 are represented as follows:

[0046]

[0047] B is the linearization matrix. It can be seen that at this point, all three systems are decoupled into a three-input, single-output parametric strict feedback form, with outputs respectively (y... x ,y y ,y h This satisfies the prerequisites for constructing the inversion controller.

[0048] Based on the decoupled equations (8), (9), and (10), a single-machine inversion position control law is constructed. Taking the x-channel as an example, omitting the derivation process, the following formula can be obtained:

[0049]

[0050] Where, x d The desired position in the x-direction; z 11 For the output quantity y x With x d Error; α1 is the harmonic parameter; z 12 Let x2 be the error between state variables x1 and α1. When:

[0051]

[0052] Then the Lyapunov function V x2 derivative c 11 c 12 Let c1 = min{c 11 ,c 12},but That is, V x2 Stability, and thus the Lyapunov function V x1 Stable. At this point, z 11 , z 12 All of them will decay exponentially to 0.

[0053] Similarly, inversion control laws are constructed for the y-channel and h-channel respectively. First, the inversion position control law for the y-channel is constructed as follows:

[0054]

[0055] Among them, y d The desired position in the y-direction; z 21 For the output quantity y y With y d Error; α2 is the harmonic parameter; z 22 Let be the error between the state variables y2 and α2. When:

[0056]

[0057] Where c 21 c 22 Let c2 be a constant term, and let c2 = min{c 21 ,c 22}, then the Lyapunov function V y2 derivative That is, V at this time y2 Stable, and consequently, stable in Lyapunov. 21 , z 22 The exponential decays to 0.

[0058] Furthermore, the inversion position control law for the h-channel is constructed as follows:

[0059]

[0060] Among them, h d The desired position in the h direction; z 31 For the output quantity y h with h d Error; α3 is the harmonic parameter; z 32 Let h2 be the error between state variables h2 and α3. When:

[0061]

[0062] Where c 31 c 32 Let c3 be a constant term, and let c3 = min{c 31 ,c 32}, then the Lyapunov function V h2 derivative At this time V h2 Stable, and consequently, stable in Lyapunov. 31 , z 32 The exponential decays to 0.

[0063] S23, Solving the control variables of a three-channel system

[0064] The conditions (12), (14), and (16) for the three-channel inversion control law obtained in step S22 to hold all involve three input quantities [n x ,n y ,n h This problem cannot be solved independently. Therefore, it is necessary to establish a homogeneous system of equations by simultaneously applying the three preconditions. Finally, the control variables that satisfy the conditions are obtained by solving the homogeneous system of equations.

[0065] First, by simultaneously applying the three conditions, we can obtain the homogeneous system of equations:

[0066]

[0067] Wherein, the coefficient diagonal matrix C1 = diag[c 11 c 21 c 31 ], C2 = diag[c 12 c 22 c 32 The actual position quantity P = [xy h] T Expected position quantity P d =[x d y d h d ] T Intermediate quantity Z1 = [z 11 z 21 z 31 ] T Z2 = [z 12 z22 z 32 ] T Gravitational force G = [0 0 -g] T After calculating the state variable X at time t, the control variable u at time t+1 can be obtained by solving the homogeneous system of equations (17). c ,Right now:

[0068]

[0069] Step 3: Grouping method based on bird flock neighborhood interaction

[0070] During flocks of European starlings, each individual bird interacts only with a fixed number of nearby individuals at a time. The starlings compare and prioritize the distances to other birds, then select the closest individuals to interact with. For example... Figure 2 As shown, for individuals 3 and 4, their respective three nearest neighbors are individuals 2, 1, 5 and 1, 3, 5, respectively. Since they do not have each other in their fixed neighbors, they can be divided into two groups. Based on this mechanism, a grouping method for unmanned aerial vehicles (UAVs) that mimics bird flock neighborhood interactions can be designed.

[0071] In the initial search phase, the number of aircraft and targets is read to determine the number of groups and the number of members within each group. If the groups cannot evenly distribute the unmanned swarm, a remainder method is used to randomly assign the remaining individuals to different groups. Then, an unmanned aircraft in space is randomly selected, its distance to other individuals is calculated, and the distances are sorted. The aircraft with the nearest expected number of individuals is selected and removed from the original swarm, forming a new group. The same process is then repeated for the next group until all aircraft are grouped.

[0072] Step 4: Pheromone Cooperative Search Method Based on Correlation Localization Error

[0073] S41, Generating pheromone structures

[0074] When a large colony of ants is in an unknown environment, they collectively explore the unknown space and search for food, using pheromones as a communication medium. The concentration of pheromones influences the entire colony's value judgment of a local area, attracting the whole group to move towards higher-value areas. Inspired by the ant colony's pheromone-based foraging behavior, a high-speed UAV search collaboration method based on error-correlated pheromones is designed. After dividing the search map into two-dimensional grids, the high-speed UAV swarm can be regarded as a group of ants on the gridded map. Each UAV has the ability to generate pheromone trajectories, as follows:

[0075]

[0076] Where h kp(t) represents the pheromone concentration generated by the k-th spacecraft at position p at time t; p is a two-dimensional grid coordinate point, obtained by dividing the map into M×N coordinate points using a minimum grid with side length Q. The x and y coordinates of these coordinate points can be represented by x, y, and y respectively. m ,y n This indicates that at time t, all the pheromone information of the k-th spacecraft constitutes its pheromone trajectory structure H. k (t).

[0077] S42. Pheromite trajectory update strategy to avoid searching known paths

[0078] The key to implementing the pheromone search method is updating the pheromone trajectory. To avoid the phenomenon of aircraft repeatedly searching known paths under the guidance of the classic pheromone search method, an improved pheromone trajectory update strategy is designed, the formula of which is as follows:

[0079]

[0080]

[0081] In formula (20), γ is the attenuation parameter, controlled between (0,1); r is the search radius of the aircraft; S is the number of aircraft members currently capable of interaction; (x i (t),y i (t) is the position of the i-th aircraft in set S at time t, and (x(t),y(t)) is the grid point at position p at time t. It is the pheromone attenuation value at position p in the pheromone structure of the i-th aircraft. When the distance of position p relative to the aircraft in set S is less than the search radius r, an attenuation effect is applied to the pheromone of the k-th aircraft in set S at that position.

[0082] S43. Pheromones Update Strategy Based on Time-Adaptive Environmental Cognition

[0083] High-speed unmanned aerial vehicles (UAVs) lack prior information about the environment when they first begin a search. As the collaborative search progresses, their understanding of the environment gradually increases. To reflect this effect, we correlate this environmental understanding with a time scale and construct the following update strategy:

[0084]

[0085] Where κ is a cognitive ability coefficient between (0,1), which can be set manually; ω is a modulating factor used to regulate the rate of increase in pheromone concentration. The above formula shows that the cluster's cognitive ability towards the environment will gradually increase over time, but this increase in cognitive ability is limited: in the early stages of the search, the cluster's cognitive ability towards the overall environment is weak, but it grows rapidly; after a longer time scale, the cluster's cognitive ability towards the overall environment is higher, but the growth rate gradually approaches saturation.

[0086] S44. Local Elite Leadership Strategy Based on Correlation with Positioning Errors

[0087] To improve the efficiency of local search, a local elite leadership strategy is designed to correlate localization errors. When an individual within the local ensemble approaches the target at a certain position, the pheromone concentration along its trajectory increases, prompting other individuals in the ensemble to move closer to this trajectory. In other words, the pheromone trajectory of the elite aircraft influences other individuals within the local ensemble. This influence is related to the passive direction-finding localization error of AOA, and is described as follows:

[0088] H k (t+1)=H k (t)+ρΔH s (t),k∈S

[0089]

[0090]

[0091] Where ρ∈[0,1] is the leadership ability coefficient; the pheromone trajectory of the k-th aircraft in set S at time t+1 is H. k (t+1); It is H k The pheromone concentration at any grid point on (t+1) cannot exceed the set concentration range [h]. min ,h max ]. ΔH s (t) represents the update amount of the pheromone structure of the neighbor set S at time t.

[0092] E is the positioning error determined by the major axis r1 of the error ellipse obtained from the passive orientation positioning of AOA within set S. max This is the maximum positioning error set; at time t, when the AOA positioning error is less than E max At that time, the distance between each individual in the neighbor set S and the center of the positioning error ellipse is compared, and the individual with the closest position is regarded as the elite individual opt. The pheromone structure of the elite opt is H. opt (t) exerts a leadership effect on individuals within the entire set. However, this leadership effect is not unlimited; as the positioning error E gradually decreases and approaches the set minimum positioning error E, the influence diminishes.min The leadership ability of elite individuals (opt) gradually weakens. When E is less than E... min Afterwards, it was determined that the positioning requirements had been met, and the search could be shifted to other targets.

[0093] Step 5: Apply the pheromone method based on correlation error to achieve cooperative search of high-speed unmanned aerial vehicle swarms, and output the motion trajectory map and the final pheromone map.

[0094] S51, unmanned aerial vehicle swarm grouping and local passive direction finding and positioning, to obtain local positioning error.

[0095] At t=t s (stt s ∈(0,T max At any given moment, the grouping of each UAV corresponding to the target to be searched is determined based on the grouping method of mimicking bird flock neighborhood interaction. Each unmanned aerial vehicle (UAV) is grouped to perform passive AOA (Aspect-Oriented Alignment) positioning, obtaining the positioning error ellipse, positioning error E, and the spatial probability p of the target to be searched. If p can be used to establish the initial pheromone field of the UAV at t=0, then the obtained positioning error E and the minimum positioning accuracy E are used... min In comparison. If E is less than E min If the minimum precision has been reached, then the number of targets N is considered to be [value missing]. T Decrease. If N T If the quantity is reduced to 0, proceed to step S54 and the program ends. If it is not 0, repeat step S51 to perform regrouping and direction finding.

[0096] S52. Use the error-correlated pheromone search method to determine the individual movement strategies within each group.

[0097] For each UAV group S, the pheromone search method using associated positioning error proposed in step four is implemented. For each individual within group S, the pheromone trajectory (19) is updated sequentially using three update strategies (20), (21), and (22) to update the pheromone concentration. For each individual, the point with the maximum pheromone concentration within the current pheromone trajectory is taken as t = t s +ΔT expected motion position [x] d (t s ),y d (t s )], which are input to their respective inversion controllers.

[0098] S53. Input the individual motion strategy within each group into the inversion controller to control the individual's motion in the two-dimensional plane. Since the aircraft only searches in the two-dimensional plane, set the desired altitude h for each aircraft. d The height remains constant at time t=0, thus obtaining the desired position vector P. d (ts +ΔT), and calculate P d (t s +ΔT) first and second derivatives. Then read the unmanned aerial vehicle state X1(t) s ) is used to generate the linearized matrix B(t) s ), intermediate quantity matrix Z1(t s ) and Z2(t s Finally, t is generated according to the control quantity formula (18). s +ΔT time control quantity u c (t s +ΔT), used to control model (4), and generate model motion state X1(t) s +ΔT). After all individual states have been updated, record the motion state X1(t) of each individual. s +ΔT) and pheromone map H(t) s +ΔT) to the corresponding motion state record structure array X r =[X1(0),X1(Δ)] T ),X1(2ΔT),....,X1(t s After the update is complete, the pheromone maps of each group are overlaid, and the result is saved to the overall pheromone record structure H=[H(0),H(ΔT)]. T ),H(2Δ T ),...,H(t s +Δ T )).

[0099] S54. Determine if the method termination condition has been reached.

[0100] After all individuals have completed their updates, determine the number N of targets to be searched. T Is it 0? If N T If it is not 0, then determine t. s +Δ T Has the maximum simulation duration T been reached? max If the condition is not met, the program proceeds normally. At this point, t = t + Δ. T This round of simulation ends; proceed to step S51. Repeat the above process continuously until the maximum simulation duration T is reached. max Or the number of targets N T The value is 0. At this point, the simulation ends, and the process jumps to step S55.

[0101] S55, Output motion trajectory map and final pheromone map

[0102] Output the motion trajectory map, generating an M×N grid map with a minimum grid side length of Q based on the initial settings; based on the initial target state X... TDraw the target to be searched and represent it with purple dots. Based on this, calculate the target based on the motion record structure array X of each body. r The three-dimensional position component P generates the corresponding motion trajectory, which is drawn using a green line. Finally, based on X... r The final position component generates the final position of each aircraft. For the final pheromone map, after generating the 3D map, the final pheromone map is drawn based on the pheromone concentration of the overall pheromone recording structure H at each grid, and is represented by blue grid lines.

[0103] Overall Flowchart Reference Figure 3 .

[0104] This invention proposes a cooperative search method for high-speed unmanned aerial vehicle (UAV) swarms based on correlated positioning error pheromones. Its advantages and benefits are as follows: 1. It provides a method for high-speed UAV swarms to achieve cooperative search and positioning in scenarios lacking prior information, expanding the application scenarios of high-speed UAV swarms and enhancing their adaptability to combat environments. 2. For a three-freedom model based on overload, an inversion position control law is designed, achieving accurate position control of the nonlinear aircraft model. 3. An improved pheromone search method correlated with positioning errors is proposed, linking the pheromone update strategy with the positioning error, and introducing multiple improved strategies to accelerate method convergence. Attached Figure Description

[0105] Figure 1 High-speed unmanned aerial vehicle inversion control block diagram

[0106] Figure 2 Bird flock neighborhood interaction concept diagram

[0107] Figure 3 Flowchart of high-speed unmanned aerial vehicle swarm search based on correlation error pheromones

[0108] Figure 4 motion trajectory diagram

[0109] Figure 5 Final pheromone map

[0110] The labels and symbols in the diagram are explained as follows:

[0111] Y d —Desired position vector;

[0112] [x d ,y d ,h d — Components of the desired position vector

[0113] Y — System output position vector

[0114] [y x ,y y ,yh —Components of the system output position vector

[0115] [z 11 ,z 21 ,z 31 — Three-channel error

[0116] [α1, α2, α3] — Three-channel harmonic parameters

[0117] [z 12 ,z 22 ,z 32 — Three-channel second-order error

[0118] Differential process

[0119] X1 — System state vector

[0120] [n xc ,n yc ,n hc —Control Components

[0121] [L1,L2,L3]—The four individuals closest to individual #4, with larger indices indicating greater distances.

[0122] [S1,S2,S3]—The four individuals closest to individual #3, with larger indices indicating greater distances.

[0123] i,j — Counters

[0124] t — Simulation duration

[0125] Δ T Minimum simulation step size

[0126] T max ——Maximum simulation duration

[0127] T n —Number of cluster groups

[0128] S sum —Structure of each group member

[0129] N T N m —Number of targets and number of aircraft

[0130] opt—A designation for elite individuals in this group Detailed Implementation

[0131] The effectiveness of the proposed method is illustrated below through a specific example of cooperative search in high-speed UAV swarms based on error-correlated pheromones. The experimental computer used an Intel(R) Core(TM) i7-7700 processor with a CPU frequency of 3.60 GHz and 32 GB of memory. The simulation software was MATLAB 2020. A cooperative search method for high-speed UAV swarms based on error-correlated pheromones is as follows:

[0132] Step 1: Constructing a single-machine AOA passive direction finding and positioning technology

[0133] To realistically simulate measurement errors, random Gaussian noise with zero mean and variance of 0.1 is added to the possible probability p of the target space to be searched. This noise simulates reasonable errors caused by sensor accuracy and external interference during the positioning process.

[0134] Step 2: High-speed unmanned aerial vehicle inversion position control

[0135] S21, Unmanned Aerial Vehicle Model Based on Overload Control

[0136] Set the total number of unmanned aerial vehicle swarms to N. m The number of drones is 64; the search radius of the drone is r = 10 km; the uniform initial velocity of the drone is V = 0.7 March = 238 m / s; the maximum velocity is V max =1.5V=357m / s; minimum velocity V min =0.8V≈190m / s. Set the trajectory inclination angle θ to be [value], and the trajectory deviation angle ψ to be [value]. Random numbers are generated. The map is set to 100km × 100km, with a minimum grid side length of 1km. The map is then divided into a grid of M = N = 101. The initial positions of the unmanned aerial vehicles (UAVs) are randomly distributed between x ∈ [0, 10]km, y ∈ [0, 40]km, and h ∈ [15, 25]m. The number of targets to be searched is set to N. T The initial positions are randomly distributed between x∈[40,48]km and y∈[10,30]km. For the unmanned aerial vehicle, the positions of all targets are unknown.

[0137] S22, Design of Three-Channel Decoupling Inversion Position Control Law

[0138] Set the simulation step size ΔT = 0.1s, and the maximum simulation duration T. max =300s; every time interval ΔT, read the state X1(t) from the spacecraft. s ) and desired position P d (t s ), calculate B(t) s +ΔT), Z1(t) s +ΔT) and Z2(t s+ΔT), input to the three-channel inversion controller.

[0139] S23, Solving the control variables of a three-channel system

[0140] Set the inversion controller parameters C1 = diag[1 1 1] and C2 = diag[1 1 1]. Solve formula (18) based on the three-channel inversion controller results to obtain the output u at the next moment. c (t s +Δ T ).

[0141] Step 3: Grouping method based on bird flock neighborhood interaction

[0142] Set the number of targets to be searched, T. n =5. According to T n A grouping method mimicking bird flocking neighborhood interactions was used to determine the initial number of members in each group to be 12-13. The unmanned aerial vehicles were grouped into S groups. sum ={S1,S2,...,S5}.

[0143] Step 4: Pheromone Cooperative Search Method Based on Correlation Localization Error

[0144] S41, Generating pheromone structures

[0145] In the initial stage, an initial pheromone structure of M×N is generated for each unmanned aerial vehicle, and its initial probability density map uses the possible probability p of the target space to be searched obtained in step one.

[0146] S42. Pheromite trajectory update strategy to avoid searching known paths

[0147] To avoid the phenomenon of aircraft repeatedly searching for known paths under the guidance of the classic pheromone search method, an improved pheromone trajectory update strategy is used. The attenuation parameter γ = 0.1 is set; the pheromone concentration of the aircraft is updated using formula (20).

[0148] S43. Pheromones Update Strategy Based on Time-Adaptive Environmental Cognition

[0149] High-speed unmanned aerial vehicles (UAVs) lack prior information about the environment when they first begin searching. As the collaborative search process progresses, their understanding of the entire environment gradually increases. To reflect this effect, the understanding of the environment is correlated with the time scale. The cognitive ability coefficient κ = 0.2 and the adjustment factor ω = 1 are set, and the pheromone concentration of the UAV is updated using formula (21).

[0150] S44. Local Elite Leadership Strategy Based on Correlation with Positioning Errors

[0151] To improve the efficiency of local search, a local elite leadership strategy is designed to correlate localization errors. When an individual within the local ensemble approaches the target at a certain position, the pheromone concentration along its trajectory increases, prompting other individuals in the ensemble to move closer to this trajectory. In other words, the pheromone trajectory of the elite aircraft influences other individuals within the local ensemble. This influence is related to the passive direction finding and localization error of AOA, and a maximum localization error E is set. max =5000m, minimum positioning error E min =500m, minimum pheromone concentration h min = -0.1, maximum pheromone concentration h max =1. Leadership ability coefficient ρ = 0.1. Apply formula (22) to update the pheromone trajectory of each aircraft.

[0152] Step 5: Apply the pheromone method based on correlation error to achieve collaborative pheromone search for high-speed unmanned aerial vehicles (UAVs), and output the final motion trajectory map and pheromone map.

[0153] S51, unmanned aerial vehicle swarm grouping and local passive direction finding and positioning, to obtain local positioning error.

[0154] At t=t s (stt s ∈(0,T max At any given moment, the grouping of each UAV corresponding to the target to be searched is determined based on the grouping method of mimicking bird flock neighborhood interaction. Each unmanned aerial vehicle (UAV) is grouped to perform passive AOA (Aspect-Oriented Alignment) positioning, obtaining the positioning error ellipse, positioning error E, and the spatial probability p of the target to be searched. If p can be used to establish the initial pheromone field of the UAV at t=0, then the obtained positioning error E and the minimum positioning accuracy E are used... min In comparison. If E is less than E min If the minimum precision has been reached, then the number of targets N is considered to be [value missing]. T Decrease. If N T If the quantity is reduced to 0, proceed to step S54 and the program ends. If it is not 0, repeat step S51 to perform regrouping and direction finding.

[0155] S52. Use the error-correlated pheromone search method to determine the individual movement strategies within each group.

[0156] For each UAV group S, the pheromone search method using associated positioning error proposed in step four is implemented. For each individual within group S, the pheromone trajectory (19) is updated sequentially using three update strategies (20), (21), and (22) to update the pheromone concentration. For each individual, the point with the maximum pheromone concentration within the current pheromone trajectory is taken as t = t s +ΔT expected motion position [x] d (ts ),y d (t s )], which are input to their respective inversion controllers.

[0157] S53. Input the individual motion strategy within each group into the inversion controller to control the individual's motion in the two-dimensional plane. Since the aircraft only searches in the two-dimensional plane, set the desired altitude h for each aircraft. d The height remains constant at time t=0, thus obtaining the desired position vector P. d (t s +ΔT), and calculate P d (t s +ΔT) first and second derivatives. Then read the unmanned aerial vehicle state X1(t) s ) is used to generate the linearized matrix B(t) s ), intermediate quantity matrix Z1(t s ) and Z2(t s Finally, t is generated according to the control quantity formula (18). s +ΔT time control quantity u c (t s +ΔT), used to control model (4), and generate model motion state X1(t) s +ΔT). After all individual states have been updated, record the motion state X1(t) of each individual. s +ΔT) and pheromone map H(t) s +ΔT) to the corresponding motion state record structure array X r =[X1(0),X1(Δ)] T ),X1(2ΔT),....,X1(t s After the update is complete, the pheromone maps of each group are overlaid, and the result is saved to the overall pheromone record structure H=[H(0),H(ΔT)]. T ),H(2Δ T ),...,H(t s +Δ T )).

[0158] S54. Determine if the method termination condition has been reached.

[0159] After all individuals have completed their updates, determine the number N of targets to be searched. T Is it 0? If N T If it is not 0, then determine t. s +Δ T Has the maximum simulation duration T been reached? max If the condition is not met, the program proceeds normally. At this point, t = t + Δ. TThis round of simulation ends; proceed to step S51. Repeat the above process continuously until the maximum simulation duration T is reached. max Or the number of targets N T The value is 0. At this point, the simulation ends, and the process jumps to step S55.

[0160] S55, Output motion trajectory map and final pheromone map

[0161] Output the motion trajectory map, generating an M×N grid map with a minimum grid side length of Q based on the initial settings; based on the initial target state X... T Draw the target to be searched and represent it with purple dots. Based on this, calculate the target based on the motion record structure array X of each body. r The three-dimensional position component P generates the corresponding motion trajectory, which is drawn using a green line. Finally, based on X... r The final position component generates the final position of each aircraft. For the final pheromone map, after generating the 3D map, the final pheromone map is drawn based on the pheromone concentration of the overall pheromone recording structure H at each grid, and is represented by blue grid lines.

Claims

1. A high-speed unmanned aerial vehicle (UAV) swarm cooperative search method based on associated positioning error pheromones, characterized in that: The method includes the following steps: Step 1: Construct a single-machine AOA passive direction finding and positioning system Step 2: Designing Inverse Position Control for High-Speed ​​Unmanned Aerial Vehicles S21, Unmanned Aerial Vehicle Model Based on Overload Control; S22. Design of Three-Channel Decoupled Inversion Position Control Law: In order to design the inversion position controller, it is necessary to decouple the three-state model of the high-speed aircraft based on the overload to obtain the decoupled model in three directions, and then design the inversion position control law for the model of the three channels respectively. S23. Solving for the control input of the three-channel system: The conditions for the three-channel inversion control law obtained in step S22 to hold all involve three inputs. It cannot be solved alone; therefore, it is necessary to establish three simultaneous preconditions to construct a homogeneous system of equations; finally, the control quantity that satisfies the conditions is obtained by solving the homogeneous system of equations. Step 3: Grouping based on bird flock neighborhood interactions Step 4: Coronavirus Cooperative Search Based on Correlation Positioning Error S41. Generate pheromone structure; S42. Pheromonic trajectory update strategy to avoid searching known paths: In order to avoid the phenomenon of aircraft repeatedly searching known paths under the guidance of the classic pheromone search method, an improved pheromone trajectory update strategy is designed. S43. Pheromones update strategy based on time-adaptive environmental cognition: When a high-speed unmanned aerial vehicle first begins its search, it lacks prior information about the environment. As the collaborative search process progresses, its cognition of the entire environment gradually increases. S44. Local Elite Leadership Strategy for Correlation Positioning Error: When an individual in a local set approaches a certain position of the target, the pheromone concentration on its trajectory will increase, prompting other individuals in the set to move closer to this trajectory. That is, the pheromone trajectory of the elite aircraft influences other individuals in the local set. Step 5: Apply the pheromone method of correlation error to realize the collaborative search of high-speed unmanned aerial vehicle swarm, and output the motion trajectory map and the final probability map.

2. The high-speed unmanned aerial vehicle swarm cooperative search method based on associated positioning error pheromones according to claim 1, characterized in that: The step S22 involves the design of a three-channel decoupled inversion position control law, wherein... The formula for obtaining the channel is as follows: ； in, for Expected position quantity in the direction; For output quantity and error; For harmonic parameters; State variables and Interval error; when the following conditions are met: ； Then the Lyapunov function derivative ; , Let be a constant term; ,but ,Right now Stability, and therefore the Lyapunov function Stable; at this time , All will decay exponentially to 0. represents a vector; u represents the overload coefficient vector of the unmanned aerial vehicle.

3. The high-speed unmanned aerial vehicle swarm cooperative search method based on associated positioning error pheromones according to claim 1, characterized in that: The specific process of solving the three-channel system control quantity in step S23 is as follows: First, by simultaneously applying the three conditions, we obtain the homogeneous system of equations: ； Among them, the coefficient diagonal matrix , Actual position quantity Expected position quantity Intermediate quantity , Heavy force ; obtained in calculation state quantity at time 1 Then, by solving the homogeneous system of equations, we obtain... Control of time ,Right now: 。 4. The high-speed unmanned aerial vehicle swarm cooperative search method based on associated positioning error pheromones according to claim 1, characterized in that: The improved pheromone trajectory update strategy formula in step S42 is as follows: ； ； In the formula, represent Time of the first The aircraft in Location-generated pheromone concentration; yes The pheromone concentration at any grid point. For set Inner A flying vehicle Pheromones at any given moment; It's the attenuation parameter, controlled within... between; It is the search radius of the aircraft; This represents the number of aircraft members currently interacting with the system. yes Time Collection Inner The location of the aircraft yes time Grid points of location; It is the first In the pheromone structure of the aircraft Location pheromone decay value; when Relative set of positions The distance of the internal aircraft is less than the search radius. At that time, for the set The first The pheromones from the aircraft at that location exert a decaying effect.

5. The high-speed unmanned aerial vehicle swarm cooperative search method based on associated positioning error pheromones according to claim 1, characterized in that: The pheromone update strategy based on time-adaptive environmental cognition in step S43 is as follows: ； in, It is a man-made setting. Cognitive ability coefficient between; It is a regulatory factor used to regulate the rate at which pheromone concentration increases. represent Time of the first The aircraft in Location-generated pheromone concentration; yes The pheromone concentration at any grid point. For set Inner A flying vehicle Pheromone trajectory at any given moment.

6. The high-speed unmanned aerial vehicle swarm cooperative search method based on associated positioning error pheromones according to claim 1, characterized in that: The local elite leadership strategy for associating positioning errors in step S44 is as follows: ； in, It is a leadership ability coefficient; set Inner A flying vehicle The pheromone trajectory at time 1 is ; yes The pheromone concentration at any grid point must not exceed the set concentration range. ; express Time Neighbor Set The amount of pheromone structure update; It is based on the set Major axis of the error ellipse obtained by passive direction finding and positioning using internal AOA The determined positioning error; This is the maximum positioning error set; in At any given time, when the AOA positioning error is less than At that time, the neighbors gathered. The distance between each individual and the center of the positioning error ellipse is compared, and the individual with the closest position is considered the elite individual. Elite The pheromone structure is It exerts a leadership effect on individuals within the entire set; however, this leadership effect is not unlimited, and it diminishes with positioning errors. Gradually reduce and approach the set minimum positioning error Elite individuals His leadership ability gradually weakened; when Less than Afterwards, believing that the positioning requirements had been met, it turned to searching for other targets.

7. The high-speed unmanned aerial vehicle swarm cooperative search method based on associated positioning error pheromones according to claim 1, characterized in that: The specific process of step five is as follows: S51, unmanned aerial vehicle swarm grouping and local passive direction finding and positioning, to obtain local positioning error. exist ( At any given moment, the grouping of each UAV corresponding to the target to be searched is determined based on the grouping method of mimicking bird flocking neighborhood interaction. Each unmanned aerial vehicle (UAV) was grouped to perform AOA passive direction finding and positioning, and the positioning error ellipse and positioning error were obtained. The probability of the target space to be searched If in hour, Used to establish the initial pheromone field of the aircraft; then the obtained positioning error is used With minimum positioning accuracy In comparison; if Less than If the minimum precision has been reached, then the number of targets is considered to have been reduced. Reduce; if If the quantity is reduced to 0, proceed to step S54 and the program ends; if it is not 0, repeat step S51 to regroup and perform direction finding and positioning. S52. Use the error-correlated pheromone search method to determine the individual movement strategies within each group. Group each unmanned aerial vehicle Step four proposes a pheromone search method based on associated positioning errors for grouping. The pheromone trajectory structure of each individual is updated sequentially using an improved pheromone trajectory update strategy, a pheromone update strategy based on time-adaptive environmental cognition, and a local elite leadership strategy based on associated localization errors to update the pheromone concentration. For each individual, the point with the maximum pheromone concentration within the current pheromone trajectory is taken as... Expected movement position at all times The inputs are sent to their respective inversion controllers; S53. Input the individual motion strategies within each group into the inversion controller to control the individual's motion in the two-dimensional plane. Set the desired height for a single machine. Unchanging, for The time-altitude is used to obtain the desired position vector. and calculate First and second derivatives; then read the unmanned aerial vehicle status. Used to generate linearized matrices Intermediate quantity matrix and Finally, the control quantity formula is used to generate the result. Time control quantity Used to control the model and generate the model's motion state. After all individual states have been updated, record the motion state of each individual. Pheromones To the corresponding motion state record structure array After the update is complete, the pheromone maps of each group are overlaid, and the results are saved to the overall pheromone record structure. ; S54. Determine if the method termination condition has been reached. After all individuals have completed their updates, determine the number of targets to be searched. Is it 0? If it is not 0, then judge. Has the maximum simulation duration been reached? If the condition is not met, the program proceeds normally. This round of simulation ends, proceed to step S51; repeat the above process continuously until the maximum simulation time is reached. or number of targets The value is 0; at this point, the simulation ends, and the process jumps to step S55. S55, Output motion trajectory map and final pheromone map Output the motion trajectory graph, generating a minimum grid side length based on the initial settings. of Grid map; based on initial target state Draw the target to be searched; based on the structure array of motion records for each body. Mid-3D position components Generate the corresponding motion trajectory; finally, based on The final position component generates the final position of each aircraft; for the final pheromone map, after generating the 3D map, the overall pheromone record structure is used. The final pheromone map is plotted based on the pheromone concentration at each grid point.

Images