An improved method for UAV formation control based on flocking

Abstract

The present invention discloses an improved algorithm for UAV formation control based on flocking, which belongs to the technical field of UAV formation. The algorithm comprises three parts: V-Flocking algorithm, multi-task clustering algorithm and VHF+Flocking obstacle avoidance algorithm. The V-Flocking algorithm introduces the characteristics of biological mechanisms in the Vicsek model, such as limited field of view and random line of sight, and further improves the convergence efficiency by improving the topological structure. The multi-task clustering algorithm enables the Flocking algorithm to realize the transition from aggregation to splitting into multiple subsets, so that the UAV formation can perform multiple subtasks. The VHF+Flocking obstacle avoidance algorithm transforms the obstacle avoidance problem into a task path planning problem, so that the drawback of the original obstacle avoidance algorithm oscillating in a narrow space is solved.

Description

Technical Field

[0001] The invention relates to the technical field of unmanned aerial vehicle formations, and in particular to an improved method for unmanned aerial vehicle formation control based on flocking. Background Art

[0002] Multi-rotor drones have the advantages of low price, small size, fast response, and simple operation. They are widely used in aerial photography, inspection, environmental monitoring, data collection, material transportation, etc. The functions of a single drone are limited, and it is difficult to complete multiple complex tasks and repetitive operations at the same time. Benefiting from drone flight control technology and flight control systems, drone formation flight can be easily achieved at this stage. After the technology matures, the main application directions of drone formations are formation light performances, formation operations, and cluster operations.

[0003] The core advantages of drone formations are: low cost per unit, large number of groups, high group intelligence, and multi-task collaborative operations. As a large number of drones of different types and performances form formations and are deployed to perform various tasks, reasonable and efficient autonomous collaborative control methods are inevitably needed to enhance their task execution capabilities.

[0004] To sum up, high intelligence and autonomy are the future development trends in the field of drone formations. The research on drone formations has been at the forefront. The progress and achievements of this research are widely used in emergency, security, entertainment and other industries, and have high economic and social value.

[0005] The UAV formation control algorithm mainly studies three aspects: the formation formation generation, the formation multi-task execution, and the formation formation obstacle avoidance. The existing UAV formation control algorithm is mainly centralized control. Compared with centralized control, each UAV participating in the formation in distributed control interacts with adjacent UAVs with its own moving target position, real-time speed, attitude and other information. In practical applications, the distributed algorithm has a relatively good control effect, the control algorithm is simple, and the calculation amount is also small. The existing distributed algorithm does not consider the aerodynamics, kinematics and structure of the UAV itself, but only regards it as a particle; it does not consider the composition of a single UAV and the parameters of the actual sensor, which is too idealized and far from practical application. Summary of the invention

[0006] The technical problem to be solved by the present invention is to address the shortcomings of the overly idealistic UAV formation algorithm in the background technology, and to provide an improved method for UAV formation control based on flocking, which not only uses a distributed control algorithm to improve the advantages of UAV formations, but also fully considers the practical problems in the application of UAV formation scenarios.

[0007] The present invention adopts the following technical solutions to solve the above technical problems:

[0008] An improved method for UAV formation control based on flocking, which includes three parts: biological mechanism clustering algorithm, multi-task clustering algorithm, and vector histogram clustering obstacle avoidance algorithm;

[0009] Among them, the V-Flocking algorithm is used to introduce the characteristics of the biological mechanism in the Vicsek model, consider the limited field of view and random sight lines, and further improve the convergence efficiency by improving the topological structure;

[0010] Multi-task clustering algorithm is used to make the Flocking algorithm realize the process of aggregation and splitting into multiple subsets, so that the drone formation can perform multiple subtasks;

[0011] VHF+Flocking obstacle avoidance algorithm is used to transform the obstacle avoidance problem into a task path planning problem, effectively solving the problem of the original obstacle avoidance algorithm oscillating in a narrow space;

[0012] The V-Flocking algorithm considers the directionality of UAV communication, adds a communication interval to the original Flocking algorithm, and evenly divides the communication interval into several domain intervals, and only communicates with the nearest neighbors in each domain interval;

[0013] Each communication interval is superimposed with a normal distribution probability, which not only conforms to the characteristics of biological vision, but also takes into account the radiation range of the data transmission module antenna;

[0014] In order to prevent the drones from colliding with each other, the minimum range of the angle range θ that the drones need to communicate is (-90°, 90°). In a two-dimensional scene, with the drone individual as the center and R as the radius of the circular area, the individual's field of view will be changed from the global field of view to the fan-shaped limited field of view defined by the angle range θ and R, and the left and right boundaries of the field of view have an angle of θ / 2 with the individual's movement direction; among them, other drone individuals within the field of view of drone individual i are the real neighbors of the drone;

[0015] Divide the angle interval θ into m sub-intervals in equal proportion. According to the characteristics of the rod antenna, the closer to the side, the smaller the antenna gain. At the same time, considering that the need for the drone to communicate with the drones on the sides is much lower than that with the drone in front, the Gaussian distribution is used to represent the communication probability of each sub-interval. The specific calculation is as follows:

[0016]

[0017] Among them, f(θ1) is the communication probability of the sub-interval in the θ1 angle direction, and θ1 is the angle between the central axis of the angle interval and the forward direction;

[0018] The drone can obtain the nearest neighbor j in several surrounding communication intervals through location information and communicate with it according to the communication probability of each interval; the introduction of communication probability is to further improve the convergence efficiency, which is equivalent to setting the weight of the neighbor; the neighbor that an individual needs to consider is a combination of direction and distance, so a discriminant function is introduced. When the distance is small and the direction angle is small, it should be considered most;

[0019]

[0020] Among them, f is the introduced discriminant function. The larger the value of this function is, the more the influence of the drone’s neighbors should be considered. j is the location of neighboring drone j, q i is the position of the drone individual i;

[0021] The discriminant function has a threshold f0. When the value is greater than the threshold, neighbor j is considered as a valid neighbor to be considered by the algorithm. Let the neighbor set interval after the improved topological structure be M i , j∈M i ;

[0022] After the introduction of the angle range of drone communication, the position update rule and direction update rule are also followed during movement; taking individual i as an example, its position update rule is the same as above, and the direction update rule becomes:

[0023]

[0024] Among them, θ1(t+1) is the movement direction that drone individual i should take at time t+1, and ω represents the number of valid neighbors;

[0025] Δθ is noise, which is an adjustable function. Angle represents the average of all individuals in the field of view of i at time t+1, including their own movement direction. is the unit motion direction complex vector of individual i, satisfying:

[0026]

[0027] Combine the above formula:

[0028]

[0029] in, To introduce the effect of biological mechanism on the acceleration of UAV, p i is the velocity vector of UAV individual i;

[0030] The acceleration term after considering the communication probability and changing the topology model is:

[0031]

[0032] in, is the acceleration vector of the drone under the Flocking algorithm, ▽V(q) is the gradient of the cluster potential function V(q), is the Laplace term of the position vector, and p is the velocity vector;

[0033] So the improved formula is as follows:

[0034]

[0035] In the formula, c1 and c2 are impact factors, c3 is also an impact factor, φ a is the action function defined, q j represents the position of UAV j, q i represents the position of UAV i, n ij represents the unit vector from drone i to drone j, is the adjustable coefficient;

[0036] Set the weight through this impact factor;

[0037] The multi-task clustering algorithm refers to adding multiple group targets when the drone formation needs to perform multiple tasks. The strategy adopted by the clustering method is the principle of proximity and selection according to weight, which specifically includes the following:

[0038] Assume the total number of drones is num, num1 drones go to q1 at speed p1, num2 drones go to q2 at speed p2, then nunm drones go to qn at speed pn;

[0039] num=num1+num2+……numn (10)

[0040] Define the potential function It indicates that the individual is attracted by the target, which is inversely proportional to the square of the distance between the group target and the drone, and directly proportional to nunm. The potential functions of each group target to the drone are compared, and the group target with the largest potential function is taken as the potential destination, which is divided into n groups. The number of drones required for each mission target is roughly achieved through this coefficient;

[0041]

[0042] in, is the position vector of the k-th group target rk, k is a parameter, representing 1, 2, 3, 4...,

[0043] The second round of screening is carried out, and the drone with the largest potential function in each group is selected as the head of the sub-cluster. By increasing the counter, each time interval, according to the size of the gravitational potential function, each target captures a drone. When the counter is equal to the number of drones required by the target, the remaining drones are stopped from being attracted. The remaining drones in the group become free groups again, and the potential function is calculated according to the remaining group targets until all drones are divided by their respective group targets. The cluster head broadcasts all individual numbers of the sub-cluster and moves towards the single group target again according to the formula;

[0044] Improved algorithm adds Choice, Represented as the position vector of the k-th group target rk, The velocity vector of the k-th group target rk and the process of allocating the clusters exactly according to the multi-task target through counters and potential functions:

[0045]

[0046] The above formula is the general form of the improved multi-task clustering algorithm. For ease of understanding, it can be expanded to the following form:

[0047]

[0048] ; The obstacle avoidance algorithm means that under the drag of the group target, the cluster is dragged along the edge of the obstacle to cross the obstacle. The obstacle avoidance algorithm belongs to the virtual force method, which realizes obstacle avoidance and advancement through gravity and repulsion;

[0049] In order to better combine the characteristics of the drone's obstacle avoidance sensor and to improve the speed of bypassing obstacles, the direction interval method is introduced on the basis of the original algorithm. The method divides the drone's motion space into several equally divided grid units. The two-dimensional space is a number of square grids. If the working space is a three-dimensional space, the grid unit is a number of cubes. In these grid units, the obstacle accumulation value is recorded, indicating the possibility of the existence of an obstacle in the grid unit. The specific calculation is as follows:

[0050] Assume there are k drones in the front, among which the position information of agent i is (x i ,y i ,θ), the distance obtained by the ranging sensor is Δl, then the obstacle point corresponding to the drone The location information is (x i +Δlcosθ,y i +Δlsinθ), Since θ is not necessarily 0, that is, the cluster does not necessarily move along the coordinate axis, the coordinate axis needs to be rotated, and the rotation angle is θ. The conversion formula is as follows:

[0051]

[0052] right Do smoothing (cftool) and get the function make Get the extreme point set Take this extreme point as the boundary of clustering, and pass a point outside the smooth curve Draw a tangent to the smooth curve. This is the route after clustering.

[0053]

[0054] The route Transform into group goals The conversion method is The function passes through the point of tangency and then passes through the point where Δx′=d as the group target, thus transforming the obstacle avoidance problem into a clustering problem under the attraction of multiple group targets.

[0055] When Δl>d:

[0056]

[0057] in, is the acceleration vector of the cluster term of the Flocking algorithm, The acceleration vector of the target item of the Flocking algorithm task, is the coefficient, φ β The action function of the set obstacle, To convert the obstacle position into a vector sign, is the coefficient, b i,k , The velocity vector of the kth obstacle

[0058] When Δl≤d:

[0059]

[0060] in, The group target after the obstacle is transformed. is the group objective function after the set obstacle transformation, The position vector of the kth obstacle, is the velocity vector of the kth obstacle.

[0061] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0062] 1. Compared with the existing UAV formation algorithm, the present invention supplements and improves the Flocking algorithm system. The improved algorithm introduces the characteristics of the biological mechanism in the Vicsek model, such as limited field of view and random line of sight, and further improves the convergence efficiency by improving the topological structure;

[0063] 2. The improved algorithm of the present invention introduces a multi-task target splitting algorithm, which allows the Flocking algorithm to achieve the transition from aggregation to splitting into multiple subsets;

[0064] 3. The improved algorithm of the present invention converts the obstacle avoidance problem into a task path planning problem by judging the conditions, so that the drawbacks of the original obstacle avoidance algorithm are solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is the algorithm flow chart of the present invention;

[0066] Figure 2 is a flow chart for introducing the biological mechanism of the present invention;

[0067] Figure 3 It is a schematic diagram of multi-task execution clustering of the present invention;

[0068] Figure 4 It is a multi-task target clustering flow chart of the present invention.

[0069] Figure 5 is a flow chart of the obstacle avoidance algorithm of the present invention;

[0070] Figure 6 It is a two-dimensional space simulation diagram of the improved algorithm of the present invention;

[0071] Figure 7 It is a three-dimensional space simulation diagram of the improved algorithm of the present invention. DETAILED DESCRIPTION

[0072] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings:

[0073] 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.

[0074] like Figures 1 to 7As shown, the present invention provides an improved method for UAV formation control based on flocking, comprising the following steps: the improved algorithm for UAV formation control based on flocking includes three parts: a V-Flocking algorithm, a multi-task clustering algorithm, and a VHF+Flocking obstacle avoidance algorithm; wherein the V-Flocking algorithm is a biological mechanism clustering algorithm, and the VHF+Flocking obstacle avoidance algorithm is a vector histogram clustering obstacle avoidance algorithm;

[0075] Among them, the V-Flocking algorithm is used to introduce the characteristics of the biological mechanism in the Vicsek model, consider the limited field of view and random sight lines, and further improve the convergence efficiency by improving the topological structure;

[0076] Multi-task clustering algorithm is used to make the Flocking algorithm realize the process of aggregation and splitting into multiple subsets, so that the drone formation can perform multiple subtasks;

[0077] The VHF+Flocking obstacle avoidance algorithm is used to transform the obstacle avoidance problem into a task path planning problem, effectively solving the problem that the original obstacle avoidance algorithm oscillates in a narrow space.

[0078] The V-Flocking algorithm considers the directionality of UAV communication, adds a communication interval to the original Flocking algorithm, and evenly divides the communication interval into several domain intervals, and only communicates with the nearest neighbors in each domain interval;

[0079] In addition, a normal distribution probability is superimposed on each communication interval, which not only conforms to the characteristics of biological vision, but also takes into account the radiation range of the data transmission module antenna.

[0080] In order to prevent the drones from colliding with each other, the minimum range of the angle range θ that the drones need to communicate is (-90°, 90°). In a two-dimensional scene, with the drone individual as the center and R as the radius of the circular area, the individual's field of view will be changed from the global field of view to the fan-shaped limited field of view defined by the angle range θ and R, and the left and right boundaries of the field of view have an angle of θ / 2 with the individual's movement direction; among them, other drone individuals within the field of view of drone individual i are the real neighbors of the drone;

[0081] Divide the angle interval θ into m sub-intervals in equal proportion. According to the characteristics of the rod antenna, the closer to the side, the smaller the antenna gain. At the same time, considering that the need for the drone to communicate with the drones on the sides is much lower than that with the drone in front, the Gaussian distribution is used to represent the communication probability of each sub-interval. The specific calculation is as follows:

[0082]

[0083] Among them, f(θ1) is the communication probability of the sub-interval in the θ1 angle direction, and θ1 is the angle between the central axis of the angle interval and the forward direction;

[0084] The drone can obtain the nearest neighbor j in several surrounding communication intervals through location information and communicate with it according to the communication probability of each interval; the introduction of communication probability is to further improve the convergence efficiency, which is equivalent to setting the weight of the neighbor. The neighbors that individuals need to consider are a combination of direction and distance, so a discriminant function is introduced. When the distance is small and the direction angle is small, it should be considered most;

[0085]

[0086] Among them, f is the introduced discriminant function. The larger the value of this function is, the more the influence of the drone’s neighbors should be considered. j is the location of neighboring drone j, q i is the position of the drone individual i;

[0087] The discriminant function has a threshold f0. When the value is greater than the threshold, neighbor j is considered as a valid neighbor to be considered by the algorithm. Let the neighbor set interval after the improved topological structure be M i , j∈M i ;

[0088] After the introduction of the angle range of drone communication, the position update rule and direction update rule are also followed during movement; taking individual i as an example, its position update rule is the same as above, and the direction update rule becomes:

[0089]

[0090] Among them, θ1(t+1) is the movement direction that drone individual i should take at time t+1, and ω represents the number of valid neighbors;

[0091] Δθ is noise, which is an adjustable function. Angle represents the average of all individuals in the field of view of i at time t+1, including their own movement direction. is the unit motion direction complex vector of individual i, satisfying:

[0092]

[0093] Combine the above formula:

[0094]

[0095] in, To introduce the effect of biological mechanism on the acceleration of UAV, p i is the velocity vector of UAV individual i;

[0096] The acceleration term after considering the communication probability and changing the topology model is:

[0097]

[0098] in, is the acceleration vector of the drone under the Flocking algorithm, ▽V(q) is the gradient of the cluster potential function V(q), is the Laplace term of the position vector, and p is the velocity vector;

[0099] So the improved formula is as follows:

[0100]

[0101] In the formula, c1 and c2 are impact factors, c3 is also an impact factor, φ a is the action function defined, q j represents the position of UAV j, q i represents the position of UAV i, n ij represents the unit vector from drone i to drone j, is the adjustable coefficient;

[0102] The weight can be set by this influence factor.

[0103] The multi-task clustering algorithm refers to adding multiple group targets when the drone formation needs to perform multiple tasks. The strategy adopted by the clustering method is the principle of proximity and selection according to weight, which specifically includes the following:

[0104] Assume the total number of drones is num, num1 drones go to q1 at speed p1, num2 drones go to q2 at speed p2, then nunm drones go to qn at speed pn;

[0105] num=num1+num2+……numn (10)

[0106] Define the potential function It indicates that the individual is attracted by the target, which is inversely proportional to the square of the distance between the group target and the drone, and directly proportional to nunm. The potential functions of each group target to the drone are compared, and the group target with the largest potential function is taken as the potential destination, which is divided into n groups. The number of drones required for each mission target is roughly achieved through this coefficient;

[0107]

[0108] in, is the position vector of the k-th group target rk, k is a parameter, representing 1, 2, 3, 4...,

[0109] The second round of screening is carried out, and the drone with the largest potential function in each group is selected as the head of the sub-cluster. By increasing the counter, each time interval, according to the size of the gravitational potential function, each target captures a drone. When the counter is equal to the number of drones required by the target, the remaining drones are stopped from being attracted. The remaining drones in the group become free groups again, and the potential function is calculated according to the remaining group targets until all drones are divided by their respective group targets. The cluster head broadcasts all individual numbers of the sub-cluster and moves towards the single group target again according to the formula;

[0110] Improved algorithm adds Choice, Represented as the position vector of the k-th group target rk, The velocity vector of the k-th group target rk and the process of allocating the clusters exactly according to the multi-task target through counters and potential functions:

[0111]

[0112] The above formula is the general form of the improved multi-task clustering algorithm. For ease of understanding, it can be expanded to the following form:

[0113]

[0114] The obstacle avoidance algorithm means that under the drag of the group target, the cluster is dragged along the edge of the obstacle to cross the obstacle. The obstacle avoidance algorithm belongs to the virtual force method, which realizes obstacle avoidance and advancement through gravity and repulsion;

[0115] In order to better combine the characteristics of the drone's obstacle avoidance sensor and to improve the speed of bypassing obstacles, the direction interval method is introduced on the basis of the original algorithm. The method divides the drone's motion space into several equally divided grid units. The two-dimensional space is a number of square grids. If the working space is a three-dimensional space, the grid unit is a number of cubes. In these grid units, the obstacle accumulation value is recorded, indicating the possibility of the existence of an obstacle in the grid unit. The specific calculation is as follows:

[0116] Assume there are k drones in the front, among which the position information of agent i is (x i ,y i ,θ), the distance obtained by the ranging sensor is Δl, then the obstacle point corresponding to the drone The location information is (x i +Δlcosθ,y i +Δlsinθ), Since θ is not necessarily 0, that is, the cluster does not necessarily move along the coordinate axis, the coordinate axis needs to be rotated, and the rotation angle is θ. The conversion formula is as follows:

[0117]

[0118] right Do smoothing (cftool) and get the function make Get the extreme point set Take this extreme point as the boundary of clustering, and pass a point outside the smooth curve Draw a tangent to the smooth curve. This is the route after clustering.

[0119]

[0120] The route Transform into group goals The conversion method is The function passes through the point of tangency and then passes through the point where Δx′=d as the group target, thus transforming the obstacle avoidance problem into a clustering problem under the attraction of multiple group targets.

[0121] When Δl>d:

[0122]

[0123] in, is the acceleration vector of the cluster term of the Flocking algorithm, The acceleration vector of the target item of the Flocking algorithm task, is the coefficient, φ β The action function of the set obstacle, To convert the obstacle position into a vector sign, is the coefficient, b i,k , The velocity vector of the kth obstacle

[0124] When Δl≤d:

[0125]

[0126] in, The group target after the obstacle is transformed. is the group objective function after the set obstacle transformation, The position vector of the kth obstacle, is the velocity vector of the kth obstacle.

[0127] The above algorithm is simulated by Matlab to evaluate its performance. Considering that there are both two-dimensional and three-dimensional scenes in the actual application of UAV formation, these two situations are simulated separately.

[0128] Table 1 is a table of 2D and 3D simulation parameters. The specific parameters of the simulation are shown in the following table:

[0129] Table 1

[0130]

[0131] The simulation scene was set with a width of 100 meters, 100 drones, an effective connection radius of 3 meters, and a speed of 5 meters per second. A point was selected as the destination in both the two-dimensional scene and the three-dimensional scene.

[0132] From the two-dimensional scene simulation diagram, we can see that 100 drones (particles) gradually change from the initial random scattered state to a group with the same moving direction and speed under the potential field force. The group also bypasses obstacles and reaches the target location under the action of the potential field force. Similar to the two-dimensional scene simulation diagram, the three-dimensional space simulation also successfully demonstrates the formation performance of drones.

[0133] In this patent, we propose an improved method for UAV formation control based on flocking. The algorithm includes three parts: V-Flocking algorithm, multi-task clustering algorithm, and VHF+Flocking obstacle avoidance algorithm. The V-Flocking algorithm introduces the characteristics of biological mechanisms in the Vicsek model, such as limited field of view and random line of sight, and further improves the convergence efficiency by improving the topological structure. The multi-task clustering algorithm allows the Flocking algorithm to achieve the transition from aggregation to splitting into multiple subsets, so that the UAV formation can perform multiple subtasks. The VHF+Flocking obstacle avoidance algorithm transforms the obstacle avoidance problem into a task path planning problem, which solves the problem of the original obstacle avoidance algorithm oscillating in a narrow space.

[0134] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless defined as herein.

[0135] The above embodiments are only for explaining the technical idea of ​​the present invention, and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention shall fall within the protection scope of the present invention. The above embodiments of the present invention are described in detail, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge scope of ordinary technicians in this field without departing from the purpose of the present invention.

Claims

1. An improved method for UAV formation control based on flocking, characterized by: It includes three parts: V-Flocking algorithm, multi-task clustering algorithm, and VHF+Flocking obstacle avoidance algorithm. The V-Flocking algorithm is a biological mechanism clustering algorithm, and the VHF+Flocking obstacle avoidance algorithm is a vector histogram clustering obstacle avoidance algorithm. Among them, the V-Flocking algorithm is used to introduce the characteristics of the biological mechanism in the Vicsek model, consider the limited field of view and random sight lines, and further improve the convergence efficiency by improving the topological structure; Multi-task clustering algorithm is used to make the Flocking algorithm realize the process of aggregation and splitting into multiple subsets, so that the drone formation can perform multiple subtasks; VHF+Flocking obstacle avoidance algorithm is used to transform the obstacle avoidance problem into a task path planning problem, effectively solving the problem of the original obstacle avoidance algorithm oscillating in a narrow space; The V-Flocking algorithm considers the directionality of UAV communication, adds a communication interval to the original Flocking algorithm, and evenly divides the communication interval into several domain intervals, and only communicates with the nearest neighbors in each domain interval; Each communication interval is superimposed with a normal distribution probability, which not only conforms to the characteristics of biological vision, but also takes into account the radiation range of the data transmission module antenna; In order to prevent the drones from colliding with each other, the minimum range of the angle range θ that the drones need to communicate is (-90°, 90°). In a two-dimensional scene, with the drone individual as the center and R as the radius of the circular area, the individual's field of view will be changed from the global field of view to the fan-shaped limited field of view defined by the angle range θ and R, and the left and right boundaries of the field of view have an angle of θ / 2 with the individual's movement direction; among them, other drone individuals within the field of view of drone individual i are the real neighbors of the drone; Divide the angle interval θ into m sub-intervals in equal proportion. According to the characteristics of the rod antenna, the closer to the side, the smaller the antenna gain. At the same time, considering that the need for the drone to communicate with the drones on the sides is much lower than that with the drone in front, the Gaussian distribution is used to represent the communication probability of each sub-interval. The specific calculation is as follows: Among them, f(θ1) is the communication probability of the sub-interval in the θ1 angle direction, and θ1 is the angle between the central axis of the angle interval and the forward direction; The drone can obtain the nearest neighbor j in several surrounding communication intervals through location information and communicate with it according to the communication probability of each interval; the introduction of communication probability is to further improve the convergence efficiency, which is equivalent to setting the weight of the neighbor; the neighbor that an individual needs to consider is a combination of direction and distance, so a discriminant function is introduced. When the distance is small and the direction angle is small, it should be considered most; Among them, f is the introduced discriminant function. The larger the value of this function is, the more the influence of the drone’s neighbors should be considered. j is the location of neighboring drone j, q i is the position of the drone individual i; The discriminant function has a threshold f0. When the value is greater than the threshold, neighbor j is considered as a valid neighbor to be considered by the algorithm. Let the neighbor set interval after the improved topological structure be M i , j∈M i ; After the introduction of the angle range of drone communication, the position update rule and direction update rule are also followed during movement; taking individual i as an example, its position update rule is the same as above, and the direction update rule becomes: Among them, θ1(t+1) is the movement direction that drone individual i should take at time t+1, and ω represents the number of valid neighbors; Δθ is noise, which is an adjustable function. Angle represents the average of all individuals in the field of view of i at time t+1, including their own movement direction. is the unit motion direction complex vector of individual i, satisfying: Combine the above formula: in, To introduce the effect of biological mechanism on the acceleration of UAV, p i is the velocity vector of UAV individual i; The acceleration term after considering the communication probability and changing the topology model is: in, is the acceleration vector of the drone under the Flocking algorithm, is the gradient of the cluster potential function V(q), is the Laplace term of the position vector, and p is the velocity vector; So the improved formula is as follows: In the formula, c1 and c2 are impact factors, c3 is also an impact factor, φ a is the action function defined, q j represents the position of UAV j, q i represents the position of UAV i, n ij represents the unit vector from drone i to drone j, is the adjustable coefficient; Set the weight through this impact factor; The multi-task clustering algorithm refers to adding multiple group targets when the drone formation needs to perform multiple tasks. The strategy adopted by the clustering method is the principle of proximity and selection according to weight, which specifically includes the following: Assume the total number of drones is num, num1 drones go to q1 at speed p1, num2 drones go to q2 at speed p2, then nunm drones go to qn at speed pn; num=num1+num2+……numn (10) Define the potential function It indicates that the individual is attracted by the target, which is inversely proportional to the square of the distance between the group target and the drone, and directly proportional to nunm. The potential functions of each group target to the drone are compared, and the group target with the largest potential function is taken as the potential destination, which is divided into n groups. The number of drones required for each mission target is roughly achieved through this coefficient; in, is the position vector of the kth group target rk, k is a parameter, representing 1, 2, 3, 4..., The second round of screening is carried out, and the drone with the largest potential function in each group is selected as the head of the sub-cluster. By increasing the counter, each time interval, according to the size of the gravitational potential function, each target captures a drone. When the counter is equal to the number of drones required by the target, the remaining drones are stopped from being attracted. The remaining drones in the group become free groups again, and the potential function is calculated according to the remaining group targets until all drones are divided by their respective group targets. The cluster head broadcasts all individual numbers of the sub-cluster and moves towards the single group target again according to the formula; Improved algorithm adds Choice, Represented as the position vector of the k-th group target rk, The velocity vector of the k-th group target rk and the process of allocating the clusters exactly according to the multi-task target through counters and potential functions: The above formula is the general form of the improved multi-task clustering algorithm. For ease of understanding, it can be expanded to the following form: ； The obstacle avoidance algorithm means that under the drag of the group target, the cluster is dragged along the edge of the obstacle to cross the obstacle. The obstacle avoidance algorithm belongs to the virtual force method, which realizes obstacle avoidance and advancement through gravity and repulsion; In order to better combine the characteristics of the drone's obstacle avoidance sensor and to improve the speed of bypassing obstacles, the direction interval method is introduced on the basis of the original algorithm. The method divides the drone's motion space into several equally divided grid units. The two-dimensional space is a number of square grids. If the working space is a three-dimensional space, the grid unit is a number of cubes. In these grid units, the obstacle accumulation value is recorded, indicating the possibility of the existence of an obstacle in the grid unit. The specific calculation is as follows: Assume there are k drones in the front, among which the position information of agent i is (x i ,y i ,θ), the distance obtained by the ranging sensor is Δl, then the obstacle point corresponding to the drone The location information is (x i +Δlcosθ,y i +Δlsinθ), Since θ is not necessarily 0, that is, the cluster does not necessarily move along the coordinate axis, the coordinate axis needs to be rotated, and the rotation angle is θ. The conversion formula is as follows: right Do smoothing (cftool) and get the function make Get the extreme point set Take this extreme point as the boundary of clustering, and pass a point outside the smooth curve Draw a tangent to the smooth curve. This is the route after clustering. The route Transform into group goals The conversion method is The function passes through the point of tangency and then passes through the point where Δx′=d as the group target, thus transforming the obstacle avoidance problem into a clustering problem under the attraction of multiple group targets. When Δl>d: in, is the acceleration vector of the cluster term of the Flocking algorithm, The acceleration vector of the target item of the Flocking algorithm task, is the coefficient, φ β The action function of the set obstacle, To convert the obstacle position into a vector sign, is the coefficient, b i,k , The velocity vector of the kth obstacle When Δl≤d: in, The group target after the obstacle is transformed. is the group objective function after the set obstacle transformation, The position vector of the kth obstacle, is the velocity vector of the kth obstacle.

Images