Multi-unmanned aerial vehicle intelligent target avoidance search method based on prior probability information guidance

Abstract

The invention belongs to the technical field of multi-unmanned aerial vehicle collaborative planning, and particularly relates to a multi-unmanned aerial vehicle intelligent avoidance target searching method based on prior probability information guidance, which comprises the following steps: step 1, establishing an unmanned aerial vehicle platform model, a sensor model and a target motion model; 2, an environment information graph model is established, wherein the environment information graph model comprises a target search probability graph, an environment uncertain information graph and a digital pheromone graph; and step 3, performing path planning by combining particle swarm optimization (PSO) and a Dubins curve under a distributed model predictive control (DMPC) framework, tracking the target by the unmanned aerial vehicle according to avoidance dynamics of the target, and stopping moving once the target is captured by the unmanned aerial vehicle until all the targets are captured. According to the method, the motion prediction model is established, the probability map is updated in real time, efficient search of multiple targets is achieved under the condition that the number of the targets, the initial positions and the motion states are known, and the problems that a collaborative search algorithm is low in efficiency and resources of multiple unmanned aerial vehicles are wasted are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-UAV collaborative planning, and in particular to a multi-UAV intelligent avoidance target search method guided by prior probability information. Background Art

[0002] With the rapid development of unmanned aerial vehicle (UAV) technology, it has been widely used in search and rescue, agricultural monitoring and urban planning. Target search is the first step in many tasks. Only after the target is located can the subsequent tracking and strike tasks be carried out. Focusing on multi-UAV target search in a dynamic environment, compared with the limited coverage and response speed of a single UAV, multi-UAV collaborative search has greatly improved the search efficiency and success rate through task allocation, path planning and communication coordination, especially in confrontational tasks such as military reconnaissance and border patrol. In these tasks, the target usually has a certain degree of intelligence and will adopt strategies to avoid reconnaissance, such as changing the moving path, hiding signals and interfering with reconnaissance signals, making it difficult for traditional collaborative search methods to effectively track the target. Therefore, how to respond quickly and adjust the search strategy in real time in a large area is a very challenging task.

[0003] Collaborative search for mobile targets in wide-area environments is a challenging task. The uncertainty of target positions directly makes the search path planning task extremely complicated. Most existing methods focus on static or general dynamic targets, and rarely involve intelligent evasive dynamic targets. In such environments, traditional target search methods are difficult to adapt to the real-time evasive behavior of targets, resulting in reduced search efficiency. Therefore, a multi-UAV intelligent evasive target search method based on prior probability information is proposed to solve the above problems. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the shortcomings of the prior art, the present invention provides a multi-UAV intelligent avoidance target search method guided by prior probability information, which solves the problems raised in the above background technology.

[0006] (II) Technical solution

[0007] In order to achieve the above-mentioned purpose, the present invention specifically adopts the following technical solutions:

[0008] A multi-UAV intelligent avoidance target search method based on prior probability information guidance includes the following steps:

[0009] Step 1: Establish UAV platform model, sensor model, and target motion model;

[0010] Step 2: Establish an environmental information graph model, including target search probability graph, environmental uncertainty information graph, and digital pheromone graph;

[0011] Step 3: Path planning is performed under the framework of distributed model predictive control (DMPC) combined with particle swarm optimization (PSO) and Dubins curve. The UAV tracks the target according to its avoidance dynamics. Once the target is captured by the UAV, it stops moving until all targets are captured.

[0012] Furthermore, the specific content of establishing the UAV platform model in step 1 is: a small fixed-wing UAV is used to cruise at a constant speed and a fixed altitude, and when searching for a path for the UAV, its flight performance and related constraints need to be fully considered. The UAV flies forward at a fixed speed and is limited by the maximum yaw angle.

[0013] Furthermore, the specific content of establishing the sensor model in step 1 is as follows: the drone uses sensors to detect targets during the search process, and continuously updates the target existence probability based on the sensor detection information. However, due to factors such as sensor measurement uncertainty and environmental occlusion, the target may be missed or misjudged. Assuming that the drone sensor detection probability p d ∈[0,1], indicating the probability that the drone detects the target when there is a target; the false alarm probability of the sensor p f ∈[0,1], which indicates the probability that the UAV detects the target even though the target does not exist. When the UAV is performing the search mission, it considers the sensor detection probability and false alarm probability, and uses the Bayesian criterion to dynamically update the probability of target existence in the mission area according to the detection information of its own sensors.

[0014] Furthermore, the specific content of establishing the target motion model in step 1 is: simplifying the structure and motion posture of the target into particles; for dynamic targets with intelligent avoidance reconnaissance characteristics, considering that the actual searched target has a certain anti-reconnaissance capability, the target can sense the presence of nearby drones. When the drone enters the perception range, the target will make an evasive action. When the target senses the drone, it will adjust its motion direction and move away from the direction of the drone. The target will update its position along the avoidance direction. The target motion equation is designed as:

[0015] Where: p t is the position of the target at time t, p t+1 is the target position at the next moment, v t is the speed of the target, Δt is the time step, is the unit vector of the avoidance direction, which is calculated by the relative position between the target and the drone: r = p t -puav =(x t -x uav ,y t -y uav );

[0016] In order to get away from the drone, the target will move in the opposite direction of its position, avoiding the direction is the unit vector in the opposite direction of the relative position vector:

[0017] Considering that there are multiple drones, the target performs comprehensive avoidance based on the positions of all drones, and performs a weighted average of the position effects of all drones: Among them, w i is the avoidance strength, defined as The closer the drone is, the greater its influence on the avoidance direction.

[0018] Furthermore, the specific content of establishing the target search probability map in step 2 is as follows: before the multi-UAV system searches the mission area, each UAV initializes its own search information map. In the process of searching and tracking dynamic targets, the following three situations should be considered: I: when the approximate initial position of the target is known; II: when the position, speed and direction of the target are all known; III: when the position, speed and direction of the target are all known; Use a probability-based dynamic model to update the target position, and use the probability density function p xy (t) is used to describe the probability that the target is at position (x, y) at time t, and considering that the target may move to an adjacent grid cell in each time step, the dynamic update is described by the probability change Δp(m,l,k,i) between grids: Δp(m,l,k,i)=p(m+Δ,l,k)-p(m,l,k); where: Δ corresponds to adjacent grid points in different directions, Δp(m,l,k,i) represents the spatial gradient of the probability of the existence of target l in a certain direction i of grid m, which is calculated by drone k; in order to further correct the uncertainty in the target movement process, the scattering intensity coefficient d(l,i) is introduced, and finally the following update formula is obtained:

[0019]

[0020] Furthermore, in step 2, an environmental uncertainty information map is established. As the degree of collaborative search of the region by multiple UAVs deepens, the degree of understanding of the mission area by UAVs gradually increases, and the uncertainty decreases. When the grid (x, y) is detected, χ xy (t+1)=η·χ xy (t); when the grid (x, y) is not detected, χ xy (t)+0.5(1-χ xy(t)); where η is the attenuation factor of information uncertainty in the environment, η∈[0,1], χ xy (t=0)=1.

[0021] Furthermore, the specific content of establishing the digital pheromone map in step 2 is: using the digital pheromone map to assist decision-making control, quantifying the changes in the drone detection area by constructing the pheromone map, guiding the drone to reduce repeated searches of the same area, and revisiting the area where the target has been detected; establishing the attraction pheromone and repulsion pheromone of the digital pheromone map.

[0022] Furthermore, in step 3, under the DMPC framework, the system transforms the large-scale search planning problem into a series of short-term planning problems, establishes a multi-UAV collaborative search rolling planning model, and assumes that x(k+q|k) and u(k+q|k) are the predictions of the system state and control input at time k+q, respectively. For simplicity, let the N-step predicted state at time k be X(k), and the control input be U(k). Then the search efficiency of the system after N-step prediction is recorded as: Where P is the length of the rolling optimization prediction window. If P = 3, then at time k, the optimization model for solving the optimal task decision of the system can be described as: The overall search effectiveness function J of the multi-UAV system after using the control input in the state is composed of the regional target discovery benefit J in the target search process T 、Expected detection benefits of multiple UAVsJ F 、Multi-UAV collaboration benefitsJ C , UAV execution cost C1, UAV collision avoidance C2, and threat cost C3.

[0023] Furthermore, the specific contents of the trajectory planning combining the particle swarm optimization algorithm with the Dubins path in step 3 are as follows: in the DMPC iterative solution process, the system uses the PSO algorithm to optimize the control signal to obtain the control input sequence U(k) of the subsystem; in the control time domain, each UAV calculates the initial heading according to the predicted target position and direction, and adjusts the control signal according to the decision result to calculate the new position and heading while detecting the target and updating the environmental information; in addition, the system optimizes the motion trajectory of the UAV by generating a Dubins curve path, optimizes the heading angle adjustment of the UAV by using the particle swarm, and then generates a Dubins path that meets the minimum turning radius constraint; the whole process is executed cyclically in each control cycle to ensure that the UAV can adapt to environmental changes in real time and complete the target search;

[0024] Consider N v The initial position of each drone is The initial heading angle is where j = 1, 2, ..., N vThe goal is to plan the control sequence of each UAV in the rolling time domain N to maximize the overall mission benefit while satisfying the UAV’s minimum turning radius R min Isokinetic constraints;

[0025] PSO is used to optimize the change in the heading angle of the UAV and determine its trajectory at each prediction time step. The fitness function J of PSO is i Combining multiple factors, such as target discovery, path smoothness, and collision avoidance; PSO iteratively updates the position and velocity of particles to find the optimal solution;

[0026] In each iteration, the fitness function J i Evaluate the performance of particle i and update the individual optimal solution p of the particle i,best and the global optimal solution g best ; When the algorithm converges, use the global optimal particle g best Decode and obtain the control signal of the drone.

[0027] (III) Beneficial effects

[0028] Compared with the prior art, the present invention provides a multi-UAV intelligent avoidance target search method based on prior probability information guidance, which has the following beneficial effects:

[0029] Aiming at the realistic scenario where the trajectory of the target to be searched is random and changeable and has the characteristic of avoiding reconnaissance, the present invention makes full use of environmental detection information, establishes a motion prediction model, updates the probability map in real time, and realizes efficient search for multiple targets when the number, initial position and motion state of the targets are known, thus solving the problems of low efficiency of collaborative search algorithm and waste of multiple UAV resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of the method of the present invention;

[0031] Figure 2 is a system block diagram of the method of the present invention;

[0032] Figure 3 This is the path effect diagram of the traditional search method;

[0033] Figure 4 This is a path effect diagram of the search method of the present invention;

[0034] Figure 5 A comparison chart of search efficiency between the traditional method and the method proposed by the present invention;

[0035] Figure 6 This is a comparison chart of coverage between the traditional method and the method proposed in this invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] Example

[0038] like Figure 1-6 As shown, an embodiment of the present invention proposes a multi-UAV intelligent avoidance target search method based on prior probability information guidance, comprising the following steps:

[0039] Step 1: Establish UAV platform model, sensor model, and target motion model;

[0040] The specific content of establishing the UAV platform model is: a small fixed-wing UAV is used to cruise at a constant speed and fixed altitude, and when searching for the UAV, its flight performance and related constraints need to be fully considered. The UAV flies forward at a fixed speed and is limited by the maximum yaw angle. The state vector of the i-th UAV at time k is:

[0041] s i (k) = [P uav (k),ψ i (k)]

[0042] Where: P uav (k)=(x i (k),y i (k)) is the UAV position coordinate, ψ i (k) is the UAV yaw angle, then the discrete system motion model of the i-th UAV is expressed as follows:

[0043]

[0044] |u i |≤ω max

[0045] The UAV at the current moment can only obtain the target grid position and heading information at the next moment. In the actual search process, the UAV is continuously maneuvering, so it is necessary to plan the flight track from the current grid to the target grid for the UAV online based on the target grid position determined by each decision result.

[0046] Where: v is the UAV flight speed; Δt is the time step; u i (k) = Δψ i (t) is the decision variable of UAV, i.e., the change of yaw angle.

[0047] The specific content of establishing the sensor model is as follows: the drone uses sensors to detect targets during the search process, and continuously updates the target existence probability based on the sensor detection information. However, due to factors such as sensor measurement uncertainty and environmental occlusion, targets may be missed or misjudged. Assume that the drone sensor detection probability p d ∈[0,1], indicating the probability that the drone detects the target when there is a target; the false alarm probability of the sensor p f ∈[0,1], which indicates the probability that the drone detects the target even though the target does not exist. When the drone is performing the search mission, it dynamically updates the probability p of the target existing in the mission area based on the detection information of its own sensors. xy (t). Considering the sensor detection probability and false alarm probability, the Bayesian criterion is used to update the target existence probability, which is described as follows:

[0048]

[0049] The above formula represents the relationship between the probability of the target existing in the grid at the next moment and the detection probability of the sensor at the previous moment, where: d =0.8,p f =0.2, b(n)=0 means the target is not found; b(n)=1 means the target is found.

[0050] The specific content of establishing the target motion model is: simplifying the structure and motion posture of the target into a mass point, ignoring the motion of the UAV in the Z-axis direction, and recording the state vector of the j-th target at time k as:

[0051] s j (k) = [p t (k),φ j (k)]

[0052] Where: p t (k)=(x target (k),y target (k)) is the target position coordinate, φ(t) is the target yaw angle. Then the discrete system motion model of the jth target is expressed as follows:

[0053]

[0054]

[0055] p t is the position of the target at time t, p t+1 is the target position at the next moment. t is the speed of the target, Δt is the time step, is the unit vector of the target's current direction of motion.

[0056] For dynamic targets with intelligent avoidance detection characteristics, considering that the actual searched target has a certain anti-detection capability, the target can sense the presence of nearby drones. When the drone enters the sensing range, the target will make evasive actions. When the target senses the drone, it will adjust its direction of movement away from the direction of the drone. The target will update its position along the avoidance direction, and the new motion equation becomes:

[0057]

[0058] in: is the unit vector of the avoidance direction. The avoidance direction is calculated by the relative position between the target and the drone:

[0059] r=p t -p uav =(x t -x uav ,y t -y uav )

[0060] In order to get away from the drone, the target will move in the opposite direction of its position, avoiding the direction is the unit vector in the opposite direction of the relative position vector:

[0061]

[0062] Considering that there are multiple drones, the target performs comprehensive avoidance based on the positions of all drones. The weighted average of the position influences of all drones is:

[0063]

[0064] w i is the avoidance strength, defined as The closer the drone is, the greater its influence on the avoidance direction.

[0065] Step 2: Establish an environmental information graph model, including environmental uncertainty information graph, digital pheromone graph, and target search probability graph;

[0066] The specific content of establishing the target search probability map is as follows: before the multi-UAV system searches the mission area, each UAV initializes its own search information map. During the search and tracking of dynamic targets, the probability-based dynamic model is used to update the target position. The following probability density function p xy (t) describes the probability that the target is located at position (x, y) at time t. By making full use of the detected environmental information to update the probability map, the target distribution can be better predicted, the search time can be reduced, and the target search problem in a time-varying environment can be better solved.

[0067] When the approximate initial position of the target is known, each target exists independently in the search environment. Since the target's direction and speed of movement are unknown, its position variables (x t ,y t ) are also independent of each other, so they obey the two-dimensional Gaussian distribution respectively: In the entire search environment, the probability distribution information is N t The cumulative sum of independent target probability distribution values ​​is used to initialize the target search probability map by using a multi-peak normal distribution:

[0068] r(i,j) 2 =(x i -x t ) 2 +(y j -y t ) 2

[0069]

[0070] Among them, r(i,j) represents the distance of the target from the position (x,y).

[0071] When the position and velocity of the target are known, the probability density function takes into account the expected position of the target after it moves, assuming that the target moves in a straight line:

[0072]

[0073] Among them, v t is the target's moving speed, t0 is the target's delayed moving time, t0 = 10s.

[0074] When the position, velocity, and direction of the target are known, the probability density function is further refined to take into account not only the change in the target's expected position, but also the direction error:

[0075]

[0076] Among them, θ(i,j) refers to the relative angle between any point on the map and the peak center of the Gaussian distribution, θ t (i, j) refers to the initial direction of the dynamic target, δ θ is the standard deviation of the angle distribution.

[0077] In each time step, the target may move to an adjacent grid cell. This dynamic update is described by the probability change Δp(m,l,k,i) between grids:

[0078] Δp(m,l,k,i)=p(m+Δ,l,k)-p(m,l,k)

[0079] Where: Δ corresponds to adjacent grid points in different directions. Δp(m,l,k,i) represents the spatial gradient of the probability of the existence of target l in a certain direction i of grid m, calculated by drone k. This update method takes into account the possible direction and speed of the target within the time step, ensuring dynamic correction of the target position at different time steps.

[0080] In order to further correct the uncertainty in the target movement process, the scattering intensity coefficient d(l,i) is introduced, and finally the following update formula is obtained:

[0081]

[0082] Through this formula, the probability of the target is corrected in different directions to be more consistent with the actual movement, ensuring accurate tracking of the target during the search process.

[0083] The specific content of establishing the environmental uncertainty information map is: In order not to miss possible targets in the mission area, the UAV's search path should cover the unknown area as much as possible, that is, the area with large uncertainty in the search probability map. As the degree of collaborative search of multiple UAVs in the area deepens, the UAV's understanding of the mission area gradually increases and the uncertainty decreases.

[0084]

[0085] Where η is the attenuation factor of information uncertainty in the environment, η∈[0,1],χ xy (t=0)=1.

[0086] The specific contents of establishing a digital pheromone map are as follows: a digital pheromone map is used to assist decision-making and control. By constructing a pheromone map, the changes in the drone detection area are quantified. The drone is guided to reduce repeated searches in the same area and revisit areas where the target has been detected to achieve effective collaborative search. The update method of the attraction pheromone and repulsion pheromone of the digital pheromone map is as follows:

[0087]

[0088] E a represents the evaporation coefficient of the attraction pheromone, G a represents the propagation coefficient of the attracting pheromone to the surrounding grid cells, D a Indicates the autonomous release content of attracting pheromones. r represents the evaporation coefficient of repelling pheromone, G r represents the propagation coefficient of repulsive pheromone, D r represents the autonomous release content of repellent pheromone. Define the pheromone switch matrix K s for:

[0089]

[0090] Among them, Δt is the time interval between two detection visits of the grid unit (x, y), δ T is the revisit time threshold, if Δt≤δ T , k x,y =0, indicating that a UAV has detected this unit in a short time and does not need to be revisited; otherwise, the grid unit needs to be revisited. and They represent the amount of attractive pheromone and repulsive pheromone transmitted from the neighboring grid during the search period t-1 to t, respectively. The propagation influence range of pheromone is defined as the 8 neighboring cells N around the cell. s =8.

[0091]

[0092] Step 3: Path planning is performed by combining particle swarm optimization (PSO) and Dubins curve under the framework of distributed model predictive control (DMPC);

[0093] In the search process, a distributed target search method based on distributed model predictive control (DMPC) is adopted. In order to improve the efficiency of solving the multi-UAV collaborative search planning problem, under the DMPC framework, the system transforms the large-scale search planning problem into a series of short-term planning problems, and establishes a multi-UAV collaborative search rolling planning model. Let x(k+q|k) and u(k+q|k) be the predictions of the system state and control input at time k+q, respectively. For simplicity, let the N-step predicted state at time k be X(k), and the control input be U(k). Then the search efficiency of the system after N-step prediction is recorded as:

[0094]

[0095] Where P is the length of the rolling optimization prediction window, and P = 3. Then at time k, the optimization model for solving the optimal task decision of the system can be described as

[0096]

[0097] According to the actual scene requirements and the probabilistic map model, the objective function of UAV search is designed. By solving the objective function, the UAV can make reasonable decisions.

[0098] Regional target discovery benefit in target search process J T : The target discovery function refers to the probability of discovering a target during the flight of the UAV from the current time t to the next time t+1. This indicator is described by the cumulative target discovery probability of the UAV cluster, which is defined as follows:

[0099]

[0100] Expected detection benefits of multiple UAVsJ F :Since the uncertainty of the environmental grid decreases with the increase of the number of visits by the UAV, in order to guide the UAV to search the areas with high uncertainty that have not been visited for a long time and reduce the uncertainty in the entire search environment as soon as possible, it is assumed that the range of the area that may be detected in the planning time domain is S k , expected detection benefit J F Defined as:

[0101]

[0102] In this way, the UAV path planning system can prioritize areas where the target is likely to be detected but where there is greater uncertainty, thereby maximizing the overall search efficiency and effectiveness.

[0103] Benefits of multi-UAV collaboration C : In the process of collaborative area search, multiple UAVs should try to ensure that multiple UAVs can perform tasks safely, reduce repeated search and detection of the same area by multiple UAVs, and guide UAVs to revisit the areas that have been detected. Therefore, the collaborative benefit can be defined as the environmental search state S on the UAV search track xy Function of (t):

[0104]

[0105] Execution cost C1: In order to reduce the number of UAV turns during the collaborative search process, effectively reduce energy consumption, and guide the UAV to complete the collaborative search task with the lowest execution cost, the change in the UAV heading angle is used here to represent the task execution cost C1 within the detection area at time t:

[0106]

[0107] Where θ(t) represents the heading of the UAV at time t, N represents the control window of rolling optimization, and N=3.

[0108] UAV collision avoidance C2: In order to avoid collisions between UAVs during the collaborative search process, a virtual repulsive force is introduced to construct an anti-collision mechanism between UAVs. The repulsive force generated by other UAVs on UAV i at time t is:

[0109]

[0110] Where: F ij represents the repulsive force exerted by drone j on drone i, k f represents the repulsion coefficient, ω f Repulsion factor, Dij Indicates the distance between drones, E ij The unit vector representing the direction of the force exerted by drone i on drone j, φ ij Indicates the heading of the drone i and E ij β represents the angle between the position of the drone at time t+1 and the repulsive force F i (t) is the angle between them.

[0111] Threat cost C3: refers to the risk of assessing the possibility that the drone may pass through or stay in the threat area. Each no-fly zone is defined as a hemisphere, and the collision avoidance benefit is calculated by checking whether the shortest distance from the current position of each trajectory to the last track point and the center of the hemisphere is less than the radius;

[0112]

[0113] Where: X(t+1|t) represents the trajectory of the drone from time t to time t+1. f represents the set of no-fly zone centers, O r Represents a collection of no-fly zone radii.

[0114] In summary, the overall search performance of the multi-UAV system in state x(k) with control input u(k) is defined as:

[0115] J k (x(k),u(k))=ω1J T +ω2J F +ω3J C -ω4C1-ω5C2-ω6C3

[0116] in: is the weight coefficient, and different coefficients indicate the importance attached to the benefits. It should be reasonably set according to the needs of the actual application scenario.

[0117] The specific content of trajectory planning combined with particle swarm optimization algorithm and Dubins path is as follows: In the DMPC iterative solution process, an important issue is to solve the local optimization of each UAV subsystem to obtain the control input sequence U(k) of the subsystem. Considering the unique advantages of particle swarm optimization (PSO) algorithm in solving nonlinear optimization problems, the system uses PSO algorithm to optimize the control signal. In the control time domain, each UAV calculates the initial heading according to the predicted target position and direction, and adjusts the control signal according to the decision results to calculate the new position and heading while detecting the target and updating the environmental information. In addition, the system optimizes the motion trajectory of the UAV by generating a Dubins curve path. The core idea of ​​this method is to optimize the heading angle adjustment of the UAV through particle swarm optimization, and then generate a Dubins path that meets the minimum turning radius constraint. The whole process is executed cyclically in each control cycle to ensure that the UAV can adapt to environmental changes in real time and complete the target search.

[0118] Consider N v The initial position of each drone is The initial heading angle is where j = 1, 2, ..., N v The goal is to plan the control sequence of each UAV in the rolling time domain N to maximize the overall mission benefit while satisfying the UAV’s minimum turning radius R min Equidynamic constraints.

[0119] In this method, PSO is used to optimize the change in the heading angle of the drone to determine its trajectory at each prediction time step. The fitness function J of PSO combines multiple factors such as target discovery, path smoothness, collision avoidance, etc. PSO searches for the optimal solution by iteratively updating the position and velocity of the particles:

[0120] v i (t+1)=w(t)·v i (t)+c1r1(p i,best (t)-x i (t))+c2r2(g best -x i (t))

[0121] x i (t+1)=x i (t)+v i (t+1)

[0122] Among them, x i (t) is the solution of the i-th particle at the t-th iteration, v i (t) is the velocity of the particle, p i,best is the best historical solution of the particle, g bestis the global optimal solution. c1 and c2 are learning factors, usually set to 2. r1 and r2 are random numbers between [0-1]. w(t) is the inertia weight, which is dynamically adjusted with the number of iterations:

[0123]

[0124] Among them, w max =0.9 is the initial inertia weight, which is used to speed up the search in the initial stage. min =0.4 is the minimum inertia weight, which is used in the later stage to make the algorithm converge to the optimal solution more accurately. max =50 is the maximum number of iterations.

[0125] In each iteration, the fitness function J i Evaluate the performance of particle i and update the individual optimal solution p of the particle i,best and the global optimal solution g best When the algorithm converges, the global optimal particle g is used best Decode the control signal of the drone:

[0126]

[0127] Among them, V j =V max For fixed speed.

[0128] The proposed algorithm is compared with the standard heuristic search method based on information graphs. The essential difference between the two algorithms is that the standard method does not establish a target motion model. Therefore, we designed a scenario with random and variable target trajectories, especially with the characteristics of evasion detection, to verify the effectiveness of the target motion model. Figure 3 This is the path effect diagram of the traditional search method. Figure 4 The figure is a path effect diagram of the search method of the present invention. The initial position of the drone is represented by blue solid points T1-T4, and the target motion trajectory is represented by red. In all four algorithms, the initial information of the target is the same, see Table 1.

[0129] Table 1 Initial state of the target.

[0130]

[0131] In the initial stage, when the target moves slowly, the target is close to the initial position of the drone, or the detection range of the drone is sufficient to cover the target's moving distance, motion prediction has little effect on the search efficiency. Both algorithms searched for the first two targets within 25 iterations. However, when the target moves quickly or has evasive behavior, the model without motion prediction will have difficulty adapting to dynamic changes because it only relies on the current position, resulting in frequent repositioning of the drone, increased energy consumption and maneuvering, and reduced capture success rate. In contrast, the system with motion prediction can adjust in advance to maintain high tracking accuracy. When searching for T3, compared Figure 3 and Figure 4 It can be seen that the algorithm that considers motion prediction can allow the drone to predict the location where the target will arrive. Especially in a confrontational environment, it can achieve early response to the rapid movement and turning of the target. Quickly "encircling" the fast-moving or turning target effectively avoids the search delay caused by the dynamic change of the target and reduces the situation where almost the entire area needs to be searched to capture the target. In order to further analyze and compare the search capabilities of different algorithms for moving targets, we repeated simulation experiments 30 times for different algorithms. Figure 5 A comparison chart of search efficiency between the traditional method and the method proposed by the present invention; Figure 6 The following is a comparison chart of coverage between the traditional method and the method proposed in this invention. Under a given search environment, the algorithm without motion prediction completes the search for all targets in 77 iterations, with a coverage rate of 83.97%. The algorithm with motion prediction completes the search for all targets in 75 iterations, with a coverage rate of 69.61%. Compared with the algorithm without motion prediction, the search efficiency is improved by 2.60% and the coverage rate is reduced by 17.09%. This shows that intelligent avoidance target motion prediction provides an efficient and accurate strategy for drones to search for targets in dynamic environments, which can cover less area and complete the capture of all targets with higher efficiency.

[0132] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-UAV intelligent avoidance target search method based on prior probability information guidance, characterized by: The steps include: Step 1: Establish UAV platform model, sensor model, and target motion model; Step 2: Establish an environmental information graph model, including target search probability graph, environmental uncertainty information graph, and digital pheromone graph; Step 3: Path planning is performed under the framework of distributed model predictive control (DMPC) combined with particle swarm optimization (PSO) and Dubins curve. The UAV tracks the target according to its avoidance dynamics. Once the target is captured by the UAV, it stops moving until all targets are captured.

2. According to claim 1, a multi-UAV intelligent avoidance target search method based on prior probability information guidance is characterized by: The specific content of establishing the UAV platform model in step 1 is: a small fixed-wing UAV is used to cruise at a constant speed and fixed altitude, and when searching for the UAV, its flight performance and related constraints need to be fully considered. The UAV flies forward at a fixed speed and is limited by the maximum yaw angle.

3. According to claim 1, a multi-UAV intelligent avoidance target search method based on prior probability information guidance is characterized by: The specific content of establishing the sensor model in step 1 is as follows: the drone uses sensors to detect targets during the search process, and continuously updates the target existence probability based on the sensor detection information. However, due to factors such as sensor measurement uncertainty and environmental occlusion, the target may be missed or misjudged. Assuming that the drone sensor detection probability p d ∈[0,1], indicating the probability that the drone detects the target when there is a target; the false alarm probability of the sensor p f ∈[0,1], which indicates the probability that the UAV detects the target even though the target does not exist. When the UAV is performing the search mission, it considers the sensor detection probability and false alarm probability, and uses the Bayesian criterion to dynamically update the probability of target existence in the mission area according to the detection information of its own sensors.

4. According to claim 1, a multi-UAV intelligent avoidance target search method based on prior probability information guidance is characterized by: The specific content of establishing the target motion model in step 1 is: simplifying the structure and motion posture of the target into particles; for dynamic targets with intelligent avoidance and reconnaissance characteristics, considering that the actual searched target has a certain anti-reconnaissance capability, the target can sense the presence of nearby drones. When the drone enters the perception range, the target will make an evasive action. When the target senses the drone, it will adjust its motion direction and move away from the direction of the drone. The target will update its position along the avoidance direction. The target motion equation is designed as: Where: p t is the position of the target at time t, p t+1 is the target position at the next moment, v t is the speed of the target, Δt is the time step, is the unit vector of the avoidance direction, which is calculated by the relative position between the target and the drone: r = p t -p uav =(x t -x uav ,y t -y uav ); In order to get away from the drone, the target will move in the opposite direction of its position, avoiding the direction is the unit vector in the opposite direction of the relative position vector: Considering that there are multiple drones, the target performs comprehensive avoidance based on the positions of all drones, and performs a weighted average of the position effects of all drones: Among them, w i is the avoidance strength, defined as The closer the drone is, the greater its influence on the avoidance direction.

5. According to claim 1, a multi-UAV intelligent avoidance target search method based on prior probability information guidance is characterized by: The specific contents of establishing the target search probability map in step 2 are as follows: before the multi-UAV system searches the mission area, each UAV initializes its own search information map. During the search and tracking of dynamic targets, the following three situations should be considered: I: when the approximate initial position of the target is known; II: When the position and speed of the target are known; III: When the position, speed and direction of the target are known; Use the probability-based dynamic model to update the target position, and use the probability density function p xy (t) is used to describe the probability that the target is at position (x, y) at time t, and considering that the target may move to an adjacent grid cell in each time step, the dynamic update is described by the probability change Δp(m,l,k,i) between grids: Δp(m,l,k,i)=p(m+Δ,l,k)-p(m,l,k); where: Δ corresponds to adjacent grid points in different directions, Δp(m,l,k,i) represents the spatial gradient of the probability of the existence of target l in a certain direction i of grid m, which is calculated by drone k; in order to further correct the uncertainty in the target movement process, the scattering intensity coefficient d(l,i) is introduced, and finally the following update formula is obtained:

6. The method for intelligently avoiding a target by multiple UAVs based on prior probability information guidance according to claim 5 is characterized in that: In step 2, an environmental uncertainty information map is established. As the degree of collaborative search of the area by multiple UAVs deepens, the degree of understanding of the mission area by UAVs gradually increases, and the uncertainty decreases. When the grid (x, y) is detected, χ xy (t+1)=η·χ xy (t); when the grid (x, y) is not detected, χ xy (t)+0.5(1-χ xy (t)); where η is the attenuation factor of information uncertainty in the environment, η∈[0,1], χ xy (t=0)=1.

7. The method for intelligently avoiding a target by multiple UAVs based on prior probability information guidance according to claim 1 is characterized in that: The specific contents of establishing the digital pheromone map in step 2 are as follows: using the digital pheromone map to assist decision-making and control, by constructing the pheromone map, quantifying the changes in the drone detection area, guiding the drone to reduce repeated searches of the same area, and revisiting the area where the target has been detected; Building digital pheromone maps of attraction and repulsion pheromones.

8. The method for intelligently avoiding a target by multiple UAVs based on prior probability information guidance according to claim 1 is characterized in that: In step 3, under the DMPC framework, the system transforms the large-scale search planning problem into a series of short-term planning problems, establishes a multi-UAV collaborative search rolling planning model, and assumes x(k+q|k) and u(k+q|k) are the predictions of the system state and control input at time k+q, respectively. For simplicity, let the N-step predicted state at time k be X(k), and the control input be U(k). Then the search efficiency of the system after N-step prediction is recorded as: Where P is the length of the rolling optimization prediction window. If P = 3, then at time k, the optimization model for solving the optimal task decision of the system can be described as: The overall search effectiveness function J of the multi-UAV system after using the control input in the state is composed of the regional target discovery benefit J in the target search process T 、Expected detection benefits of multiple UAVs J F 、Multi-UAV collaboration benefitsJ C , UAV execution cost C1, UAV collision avoidance C2, and threat cost C3.

9. The method for intelligently avoiding a target by multiple UAVs based on prior probability information guidance according to claim 8 is characterized in that: The specific contents of the trajectory planning combining the particle swarm optimization algorithm with the Dubins path in step 3 are as follows: in the DMPC iterative solution process, the system uses the PSO algorithm to optimize the control signal to obtain the control input sequence U(k) of the subsystem; in the control time domain, each UAV calculates the initial heading according to the predicted target position and direction, and adjusts the control signal according to the decision result to calculate the new position and heading while detecting the target and updating the environmental information; in addition, the system optimizes the motion trajectory of the UAV by generating a Dubins curve path, optimizes the heading angle adjustment of the UAV by using the particle swarm, and then generates a Dubins path that meets the minimum turning radius constraint; the whole process is executed cyclically in each control cycle to ensure that the UAV can adapt to environmental changes in real time and complete the target search; Consider N v The initial position of each drone is The initial heading angle is where j = 1, 2, ..., N v The goal is to plan the control sequence of each UAV in the rolling time domain N to maximize the overall mission benefit while satisfying the UAV’s minimum turning radius R min Isokinetic constraints; PSO is used to optimize the change in the heading angle of the UAV and determine its trajectory at each prediction time step. The fitness function J of PSO is i Combining multiple factors, such as target discovery, path smoothness, and collision avoidance; PSO iteratively updates the position and velocity of particles to find the optimal solution; In each iteration, the fitness function J i Evaluate the performance of particle i and update the individual optimal solution p of the particle i,best and the global optimal solution g best ; When the algorithm converges, use the global optimal particle g best Decode and obtain the control signal of the drone.

Images