Cluster behavior prediction method based on potential function and Bayesian network and medium

Abstract

The invention discloses a fleet behavior prediction method based on a potential function and a Bayesian network and a medium. The fleet behavior prediction method comprises the following steps: establishing a topological relation of an unmanned aerial vehicle fleet; constructing an unmanned aerial vehicle potential energy field function and external regulation and control force; setting an adaptive obstacle avoidance distance, and establishing collision avoidance constraints, planning track continuity constraints and dynamics feasibility constraints; dividing the cluster by using a K-means algorithm; constructing a dynamic Bayesian network, and predicting a cluster behavior; setting a behavior prior probability; preprocessing the data to obtain required information; dynamically adjusting the weight of each behavior probability; and updating and normalizing the behavior probability. According to the method, cluster division and behavior identification results can be given under discrete data input (single moment) and continuous data input (multi-moment) conditions, good convergence time and stability are achieved, the requirement for stable output of a behavior identification algorithm in a flight scene is met, and meanwhile the behavior identification efficiency is improved. And algorithm identification results in a specific flight scene can be traced.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and in particular to a method and medium for predicting the behavior of a swarm of drones based on a potential function and a Bayesian network. Background Art

[0002] In recent years, the collaboration of drones, ground personnel and vehicles has been increasingly used in practical engineering, and aerial behavior has gradually become the protagonist. In the face of multi-formation collaborative aerial behavior scenarios, the group division and behavior prediction of target groups are crucial, but this process faces many research difficulties: the sharp increase in the number of flight units, the accelerated evolution of the situation, and massive inaccurate environmental information. In order to meet these challenges, it is necessary to study intelligent reasoning methods to achieve aerial target behavior prediction.

[0003] Traditional methods for cluster trajectory partitioning include Kalman filtering, Markov models, Bayesian theory, support vector machines, and manually designed models. Traditional cluster partitioning prediction methods have great limitations and have difficulties in dealing with large quantities of high-dimensional, multi-source raw data and the high nonlinearity and time-varying nature of the data. Therefore, some scholars have designed a method based on Monte Carlo deep policy gradient learning, using reinforcement learning for fast real-time decision-making. Based on the cluster partitioning research of many scholars, some scientists have further proposed an online trajectory planning algorithm, but when this method is based on nonlinear model predictive control (MPC), it requires the use of a special nonlinear solver to solve, which is relatively complex for solving and cluster classification.

[0004] The initial development of flight behavior prediction technology is mainly based on template matching. This method has a high time complexity and the prediction effect is not obvious. Therefore, many scholars have proposed improvement schemes for target behavior prediction algorithms based on template matching, but it still cannot change the large amount of calculations and complex template library construction and update compared with other target prediction algorithms. Therefore, some scholars have proposed target prediction algorithms based on neural networks and Bayesian theory. Target prediction based on neural networks faces problems such as complex parameters and long learning time. In the study of Bayesian classifiers, Godec M et al. proposed a random naive Bayesian classifier based on a random learning method, and Seok Pil Yoon et al. proposed a Bayesian classifier using normalized moment of inertia for feature training, and verified the efficiency and robustness of the algorithm through experiments, providing a new solution for the field of target behavior prediction. Summary of the invention

[0005] The present invention proposes a cluster behavior prediction method, device and storage medium based on potential function and Bayesian network to solve the cluster division and behavior prediction problems in complex dynamic environments where data has strong uncertainty.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A method for predicting the behavior of a fleet based on potential function and Bayesian network comprises the following steps: S1. Establish the topological relationship of the drone fleet; S2, construct the potential energy field function and external control force of the UAV; S3, set the adaptive obstacle avoidance distance, establish collision avoidance constraints, planning trajectory continuity constraints, and dynamic feasibility constraints; S4, use K-means algorithm to divide the cluster; S5. Build a dynamic Bayesian network to predict the behavior of the fleet; S6. Set the prior probability of the behavior; S7, data preprocessing, obtaining required information; S8, dynamically adjust the weight of each behavior probability; S9. Behavior probability update and normalization.

[0007] Further, the details are as follows: S1. Establish the topological relationship of the drone fleet. The specific form is:

[0008] in, is a node set, , represent arrive An edge indicates that drones or aircraft can sense each other's presence.

[0009] According to the above relationship, the neighbors of the drone are expressed as:

[0010] At the same time, assume that the discrete linear model of each drone in the swarm is:

[0011] in, and is a constant matrix, For each Drones are in The state of the moment, and , and Represent the latitude and longitude and speed of the drone respectively. It is the control quantity used to generate trajectory planning and fleet division.

[0012] S2. Construct the potential energy field function of the drone and the external control force. The specific process is: Construct the repulsive potential field function :

[0013] in, is a constant coefficient, is the location of the obstacle, is the relative distance between the UAV and the obstacle, is the repulsive force of the obstacle, is the regulating factor, is a variable positive constant.

[0014] Taking the negative gradient of the repulsive field gives the repulsive function :

[0015] in, and are the two components of the repulsive force, and their specific forms are:

[0016]

[0017] in, is the adjustment factor.

[0018] Building external regulatory power :

[0019] in, is a new repulsive force, whose magnitude is: .

[0020] S3, set the adaptive obstacle avoidance distance, establish collision avoidance constraints, planning trajectory continuity constraints, and dynamic feasibility constraints. The specific process is: define the safe distance between the drone and obstacles , the specific form is:

[0021] in, and is the collision radius, For safety margin.

[0022] Set adaptive obstacle avoidance distance :

[0023] in, Dynamic obstacle speed, which can be based on the obstacle position information detected by the drone within a certain range Make an estimate.

[0024] Establish collision avoidance constraints between drones. The specific process is: treat the drone as a spherical rigid body with a collision radius of , assuming that the drone and A collision is about to occur, so constraints are applied at the collision point:

[0025] in, and are the positions of the two drones, is the safe flight distance between the two drones, This is the earliest moment when the two drones collided. It can be expressed as:

[0026] in, is a certain moment in the prediction time domain.

[0027] Introducing relaxation factors , establish continuous collision avoidance constraints:

[0028] Construct continuity constraints for the planned trajectory of the UAV:

[0029] in, Representative Drone The 0th derivative of the Bezier curve The control points are as follows:

[0030] The feasibility constraints of structural dynamics are:

[0031] in is the first At the control point, is the order of the Bezier curve, , , and are all constants, representing the minimum speed, maximum speed, minimum acceleration and maximum acceleration respectively.

[0032] S4. Divide the cluster based on K-means algorithm. The specific process is as follows: Assume The drone is the center of the fleet. For the The radius of the potential energy circle of the first UAV is calculated. The first drone The distance between drones :

[0033] in, is the radius of the Earth, and is the latitude of the two drones (in radians), is the difference in longitude between the two drones (in radians).

[0034] Judge all Is it established? If At this moment, the drone There is a high probability that The aircraft is the center of the fleet, calculating the aircraft In the plane The probability of a cluster centered on ;like Then judge that the drone Not in The center of the cluster.

[0035] S5. Build a dynamic Bayesian network to predict the behavior of the fleet. The specific process is: build a dynamic Bayesian network, and its specific structure is:

[0036] in, For the corresponding time point The state of the Bayesian model at time , the corresponding prior probability is , In the case of approaching time point, Model from state To status state migration process.

[0037] calculate The posterior probability is:

[0038] in, and Represent two adjacent moments The state confidence of the model, Indicates time Next The status of the drone, Representation Node The parent node of .

[0039] S6. Set a prior probability for each possible behavior. Based on historical data and expert knowledge, set a prior probability for each possible behavior. These probabilities reflect the relative likelihood of various behaviors occurring in the absence of any specific observation data.

[0040] S7, data preprocessing, obtain the required information. The specific process is: obtain the height, speed, electromagnetic state and longitude and latitude information from the original data, among which the height, speed and electromagnetic state can directly obtain the value, and the longitude and latitude calculation formula is:

[0041] in, is the latitude difference between the two drones (in degrees), (Unit: °).

[0042] S8. Dynamically adjust the weight of each behavior probability. The weight adjustment formula is:

[0043] Specific weight adjustments are based on the importance of specific aircraft types, while basic weights consider the contribution of each aircraft type to various behaviors under average conditions.

[0044] S9, behavior probability update and normalization. Analyze the flight status of each aircraft, use the Bayesian formula to update the probability of the behavior, and then normalize the probability after weight adjustment and multiple probability updates to ensure that the sum of the probabilities of all behaviors is 1. The specific form of probability update is:

[0045] in, For drones Behavior The probability of Based on drone Types For behavior The weight of is the total number of drones in the fleet.

[0046] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0047] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0048] It can be seen from the above technical scheme that the present invention proposes a K-means-based fleet division method and constructs a dynamic Bayesian network to predict fleet behavior in view of the high nonlinearity, time-varying and uncertainty of the data. The method can give the fleet division and behavior identification results under the conditions of discrete data input (single moment) and continuous data input (multiple moments), and has good convergence time and stability. In actual use scenarios, it can meet the requirements of fast convergence and smooth output of the behavior identification algorithm, and also has the advantages of strong adaptability, high interpretability, good robustness and stability, high efficiency and strong flexibility. By adjusting the algorithm parameters, it can be optimized for different flight scenarios to ensure that accurate fleet behavior prediction results can be obtained in various complex environments, and at the same time, the algorithm identification results in specific scenarios can be traced back and have interpretability. By introducing the K-means-based fleet division method, it can flexibly adapt to fleets of different sizes and types, enhancing the universality of the system. The dynamic Bayesian network can make full use of the advantages of time series data to improve the accuracy and reliability of behavior prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of an embodiment of the present invention; Figure 2 This is a flow chart of the K-means clustering algorithm according to an embodiment of the present invention; Figure 3 The figure is a flow chart of cluster behavior inference according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0051] like Figure 1 As shown, this embodiment discloses a cluster behavior prediction method based on potential function and Bayesian network of the present invention, comprising the following steps: S1. Establish the topological relationship of the drone fleet; S2, construct the potential energy field function and external control force of the UAV; S3, set the adaptive obstacle avoidance distance, establish collision avoidance constraints, planning trajectory continuity constraints, and dynamic feasibility constraints; S4, use K-means algorithm to divide the cluster; S5. Build a dynamic Bayesian network to predict the behavior of the fleet; S6. Set the prior probability of the behavior; S7, data preprocessing, obtaining required information; S8, dynamically adjust the weight of each behavior probability; S9. Behavior probability update and normalization.

[0052] The following are specific instructions: S1. Establish the topological relationship of the drone fleet. The specific form is:

[0053] in, is a node set, , represent arrive An edge indicates that drones or aircraft can sense each other's presence.

[0054] According to the above relationship, the neighbors of the drone are expressed as:

[0055] At the same time, assume that the discrete linear model of each drone in the swarm is:

[0056] in, and is a constant matrix, For each Drones are in The state of the moment, and , and Represent the latitude and longitude and speed of the drone respectively. It is the control quantity used to generate trajectory planning and fleet division.

[0057] S2. Construct the potential energy field function of the drone and the external control force. The specific process is: Construct the repulsive potential field function :

[0058] in, is a constant coefficient, is the location of the obstacle, is the relative distance between the UAV and the obstacle, is the repulsive force of the obstacle, is the regulating factor, is a variable positive constant.

[0059] Taking the negative gradient of the repulsive field gives the repulsive function :

[0060] in, and are the two components of the repulsive force, and their specific forms are:

[0061]

[0062] in, is the adjustment factor. Considering the situation where there are obstacles near the target point, as the distance between the drone and the target point and obstacles gradually decreases, the repulsive force The proportion of weight gradually increases. The specific gravity gradually decreases, and the increase of repulsion gradually slows down, thus improving the situation where the drone hovers at a low speed near the target point due to excessive repulsion.

[0063] The movement of the drone in the artificial potential field is determined by the combined force of repulsion and gravity. If the repulsion and gravity are equal in magnitude and opposite in direction, the drone will enter a force equilibrium state and cannot move, that is, it will fall into a local minimum. To solve this problem, an external control force can be introduced to push the drone away from the force equilibrium point, thus constructing an external control force. :

[0064] in, is a new repulsive force, whose magnitude is:

[0065] S3, set the adaptive obstacle avoidance distance, establish collision avoidance constraints, planning trajectory continuity constraints, and dynamic feasibility constraints. The specific process is: define the safe distance between the drone and obstacles , the specific form is:

[0066] in, and is the collision radius, For safety margin.

[0067] Set adaptive obstacle avoidance distance :

[0068] in, Dynamic obstacle speed, which can be based on the obstacle position information detected by the drone within a certain range Make an estimate.

[0069] Establish collision avoidance constraints between drones. The specific process is: treat the drone as a spherical rigid body with a collision radius of , assuming that the drone and A collision is about to occur, so constraints are applied at the collision point:

[0070] in, and are the positions of the two drones, is the safe flight distance between the two drones, This is the earliest moment when the two drones collided. It can be expressed as:

[0071] in, is a certain moment in the prediction time domain.

[0072] Introducing relaxation factors , establish continuous collision avoidance constraints:

[0073] Drones in The trajectory within the time period can be formed by splicing multiple third-order Bezier curves. The trajectory splicing requires that the trajectory is smooth at the splicing point, that is, the second-order derivative between the head and tail connection points of the two Bezier curves is continuous. According to the continuity requirement, the continuity constraint of the UAV planning trajectory is constructed:

[0074] in, Representative Drone The 0th derivative of the Bezier curve The control points are as follows:

[0075] The feasibility constraints of structural dynamics are:

[0076] in is the first At the control point, is the order of the Bezier curve, , , and are all constants, representing the minimum speed, maximum speed, minimum acceleration and maximum acceleration respectively.

[0077] S4. Use K-means algorithm to divide the cluster. K-means clustering algorithm is as follows Figure 2 As shown, the specific process is as follows: Assume that The drone is the center of the fleet. For the The radius of the potential energy circle of the first UAV is calculated. The first drone The distance between drones :

[0078] in, is the radius of the Earth, and is the latitude of the two drones (in radians), is the difference in longitude between the two drones (in radians).

[0079] Judge all Is it established? If At this moment, the drone There is a high probability that The aircraft is the center of the fleet, calculating the aircraft In the plane The probability of a cluster centered on ;like Then judge that the drone Not in The center of the cluster.

[0080] Assume that at a certain moment the present invention needs to divide the aircraft into The present invention divides these aircraft into Class Cluster ( = natural number), the steps of the aircraft group division algorithm are as follows: Step 1. The present invention starts by selecting an aircraft in the fleet as the center of the first fleet, and after completing the above-mentioned condition determination, it then iterates in sequence to find out the probability that the aircraft belongs to a certain fleet under the communication condition.

[0081] Step 2. According to other aircraft The distance probability of the center of the fleet is calculated, and then the unassigned aircraft are selected and operated in sequence according to the above formula to see if the requirements are met, and the probability is calculated.

[0082] Step 3. Count the cluster probabilities between the aircraft and the center point of each newly generated cluster. Jump to Step 1 until the center aircraft of each cluster no longer changes.

[0083] S5. Build a dynamic Bayesian network to predict the behavior of the fleet. The specific process is: build a dynamic Bayesian network, and its specific structure is:

[0084] in, For the corresponding time point The state of the Bayesian model at time , the corresponding prior probability is , In the case of approaching time point, Model from state To status state migration process.

[0085] calculate The posterior probability is:

[0086] in, and Represent two adjacent moments The state confidence of the model, Indicates time Next Node status, Representation Node The parent node of .

[0087] S6. Set prior probabilities. Based on historical data and expert knowledge, set a prior probability for each possible behavior. These probabilities reflect the relative likelihood of various behaviors occurring in the absence of any specific observation data.

[0088] According to the knowledge of experts in the field, when approaching a target, an aircraft usually flies at an altitude of about 4.2km and a speed of less than 1200km / h, with an equal probability of the radar being turned on or off, which is defined as mid-altitude cruising. Similarly, statistics are obtained for flight characteristic states under typical meta-behaviors such as mid-altitude accelerated cruising, mid-altitude hovering, high-altitude hovering, ultra-high-altitude hovering, and low-altitude accelerated cruising. According to comprehensive statistics, the flight altitude under various typical meta-behaviors of aircraft generally has four states: about 4.2km, above 10km, below 1km, and between 4~6km. Similarly, the flight speed values ​​under typical meta-behaviors of aircraft are generally below 1200km / h and between 1000~1800km / h. The present invention defines cruising when the distance between the aircraft and the target is greater than 120km, and hovering within 150km.

[0089] Therefore, the typical evidence state vectors of the aircraft are divided into: high: ;speed: , is Mach unit; distance: ,electromagnetic: To express affirmation, Indicates negation.

[0090] Assume that according to historical data statistics, for the state division space, when the aircraft is cruising at medium altitude, there is an 80% probability of flying at an altitude of 3-10km, a 5% probability of flying above 20km and below 3km, and a 10% probability of flying between 6-10km. Therefore, the prior state probability vector of the altitude evidence during medium altitude cruising is [0.05, 0.8, 0.1, 0.05]. Similarly, the prior probability table of the aircraft's typical meta-behavior evidence state is formed.

[0091] At the same time, according to the experience and knowledge of experts in the field, flight behaviors are divided. It is generally believed that the behaviors of mid-altitude cruising and mid-altitude accelerated cruising are behavior A; the behavior of low-altitude accelerated cruising is behavior B; the behavior of low-altitude accelerated wandering is behavior C; the behavior of high-altitude electromagnetic wandering is behavior D; and the behavior of ultra-high-altitude wandering is behavior E. Therefore, the prior probability of the typical meta-behavior evidence state is the prior probability of the above five terminal behaviors. The prior probability table of the five terminal behaviors is shown in Table 1 below.

[0092] Table 1 Behavior prior probability table

[0093] Based on the aircraft state-meta-behavior relationship table, it is assumed that the probability of various behaviors is equal, each is 1 / 5. In addition, it is assumed that the values ​​of each piece of evidence in its state domain are also equal, and the probability of each piece of evidence taking a value in the state set is also equal. The present invention assumes that an aircraft is currently discovered and detected to be flying forward at an altitude of 12Km. At this time, the speed is 0.7ma, the distance from us is 120Km, and the electromagnetic state is turned on.

[0094] Then the situational evidence is: , , , . Then, according to the behavior prior probability table, the joint probability of each behavior is calculated, and the sum of the joint probabilities of all behaviors is used as the denominator.

[0095] set up is the flight status of the aircraft, including altitude, speed, distance and electromagnetic state, expressed as The behavior is , expressed as According to Bayes' theorem, the present invention has:

[0096] in, is the conditional probability of the flight state given the behavior. is the prior probability of each behavior. is the sum of the joint probabilities of all possible states.

[0097] The present invention can Expanded into the product of the conditional probabilities of each flight state:

[0098] Now, the present invention substitutes these probabilities with specific conditional probabilities and prior probabilities to obtain the posterior probability formula for each behavior.

[0099] For behavior C, suppose the aircraft is found, , the present invention has:

[0100] The posterior probabilities of other behaviors can be calculated in a similar way. After all the behavior probabilities are calculated, the probability components of the aircraft's behavior at this moment are [0.619, 0.019, 0.077, 0.006, 0.278], which means that the most likely behavior is attack, with a probability of 0.619.

[0101] S7, data preprocessing. The specific process is: obtain the altitude, speed, electromagnetic state and longitude and latitude information from the original data, among which the altitude, speed and electromagnetic state can directly obtain numerical values, and the longitude and latitude calculation formula is:

[0102] in, is the latitude difference between the two drones (in degrees), (unit: °), is the radius of the Earth.

[0103] S8. Weight adjustment. Dynamically adjust the weight of each behavior probability. The weight adjustment formula is:

[0104] Among them, specific weight adjustments are based on the importance of special aircraft types, and basic weights consider the contribution of each aircraft type to various behaviors under average conditions. Different types of aircraft have different initial basic weights according to their attributes. The adjustment factors are based on the performance and importance of the aircraft type in a specific mission. When this weight adjustment model is actually applied, simulation data is used for verification and adjustment. By simulating different flight scenarios, the impact of different weight settings on decision results can be observed, and the model can be further fine-tuned to ensure its effectiveness and accuracy in actual operations. At the same time, flight data is continuously collected and analyzed, and the weight model is updated in a timely manner according to new needs and technological developments.

[0105] In the embodiment of the present invention, the following seven types of aircraft are considered in the fleet: aircraft type 1, aircraft type 2, aircraft type 3, aircraft type 4, aircraft type 5, aircraft type 6, and aircraft type 7. In view of the influence of each type of aircraft in the fleet on the five behaviors, the relevant weights are set as shown in Table 2. The default weight is 0, which means that the aircraft type has basically no influence on the matching behavior. Table 2 Weight table of the influence of aircraft type on behavior

[0106]

[0107] S9, probability update and normalization. The specific process is: analyze the flight status of each aircraft, and use the Bayesian formula to update the probability of the behavior. The specific form is:

[0108] in, For drones Behavior The probability of Based on drone Types For behavior The weight of is the total number of drones in the fleet.

[0109] The probabilities after weight adjustment and multiple probability updates are normalized to ensure that the sum of the probabilities of all behaviors is 1. The above fleet behavior prediction process is as follows: Figure 3 shown.

[0110] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0111] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0112] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any of the cluster behavior prediction methods based on potential functions and Bayesian networks in the above embodiments.

[0113] It is understandable that the system, device and storage medium provided in the embodiments of the present invention correspond to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above methods.

[0114] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a server or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or a data center that includes one or more available media integrated. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0115] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0116] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0117] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting fleet behavior based on potential function and Bayesian network, characterized in that: The following steps are included: S1. Establish the topological relationship of the drone fleet; S2, construct the potential energy field function and external control force of the UAV; S3, set the adaptive obstacle avoidance distance, establish collision avoidance constraints, planning trajectory continuity constraints, and dynamic feasibility constraints; S4, use K-means algorithm to divide the cluster; S5. Build a dynamic Bayesian network to predict the behavior of the fleet; S6. Set the prior probability of the behavior; S7, data preprocessing, obtaining required information; S8, dynamically adjust the weight of each behavior probability; S9. Behavior probability update and normalization.

2. The method for predicting swarm behavior based on potential function and Bayesian network according to claim 1, characterized in that: S1, establishing the topological relationship of the drone fleet, specifically in the form of: in, is a node set, , represent arrive An edge of , indicating that drones or aircraft can sense each other's presence; According to the above relationship, the neighbors of the drone are expressed as: At the same time, assume that the discrete linear model of each drone in the swarm is: in, and is a constant matrix, For each Drones are in The state of the moment, and , and Represent the latitude and longitude and speed of the drone respectively. It is the control quantity used to generate trajectory planning and fleet division.

3. The method for predicting swarm behavior based on potential function and Bayesian network according to claim 2, characterized in that: The specific process of step S2 includes: Constructing the repulsive potential field function : in, is a constant coefficient, is the location of the obstacle, is the relative distance between the drone and the obstacle, is the repulsive force of the obstacle, is the regulating factor, is a variable positive constant; Taking the negative gradient of the repulsive field gives the repulsive function : in, and are the two components of the repulsive force, and their specific forms are: in, is the regulating factor; Building external regulatory power : in, is a new repulsive force, whose magnitude is: 。 4. The method for predicting swarm behavior based on potential function and Bayesian network according to claim 3 is characterized by: The specific process of step S3 includes defining the safe distance between the drone and obstacles. , the specific form is: in, and is the collision radius, is the safety margin; Set adaptive obstacle avoidance distance : in, Dynamic obstacle speed, which can be based on the obstacle position information detected by the drone within a certain range Make an estimate; Establish collision avoidance constraints between drones. The specific process is: treat the drone as a spherical rigid body with a collision radius of , assuming that the drone and A collision is about to occur, so constraints are applied at the collision point: in, and are the positions of the two drones, is the safe flight distance between the two drones, This is the earliest moment when the two drones collided. It is expressed as: in, To predict a certain moment in the time domain; Introducing relaxation factors , establish continuous collision avoidance constraints: Construct the continuity constraint for the UAV planning trajectory: in, Representative Drone The 0th derivative of the Bezier curve The control points are as follows: The feasibility constraints of structural dynamics are: in is the first At the control point, is the order of the Bezier curve, , , and are all constants, representing the minimum speed, maximum speed, minimum acceleration and maximum acceleration respectively.

5. The method for predicting swarm behavior based on potential function and Bayesian network according to claim 4 is characterized by: The specific process of step S4 includes: Assume The drone is the center of the fleet. For the The radius of the potential energy circle of the first UAV is calculated. The first drone The distance between drones : in, is the radius of the Earth, and is the latitude of the two drones (in radians), is the difference in longitude between the two drones, in radians; Judge all Is it established? If At this moment, the drone There is a high probability that The aircraft is the center of the fleet, calculating the aircraft In the plane The probability of a cluster centered on ;like Then judge that the drone Not in The center of the cluster.

6. The method for predicting the behavior of a fleet of aircraft based on potential function and Bayesian network according to claim 5, characterized in that: The specific process of step S5 includes: Construct a dynamic Bayesian network, whose specific structure is: in, For the corresponding time point The state of the Bayesian model at time , the corresponding prior probability is , In the case of approaching time point, Model from state To status The state migration process; calculate The posterior probability is: in, and Represent two adjacent moments The state confidence of the model, Indicates time Next The status of the drone, Representation Node The parent node of .

7. The method for predicting swarm behavior based on potential function and Bayesian network according to claim 6 is characterized by: The specific process of step S7 includes: The altitude, speed, electromagnetic state, and longitude and latitude information are obtained from the original data. The altitude, speed, and electromagnetic state can directly obtain numerical values, and the longitude and latitude calculation formula is: in, is the latitude difference between the two drones, in degrees, The unit is °.

8. The method for predicting the behavior of a fleet of aircraft based on potential function and Bayesian network according to claim 7, characterized in that: The specific process of step S8 includes: The weight adjustment formula is: Specific weight adjustments are based on the importance of specific aircraft types, while basic weights consider the contribution of each aircraft type to various behaviors under average conditions.

9. The method for predicting the behavior of a fleet of aircraft based on potential function and Bayesian network according to claim 8, characterized in that: The specific process of step S9 includes: The flight status of each aircraft is analyzed, and the Bayesian formula is used to update the probability of the behavior. Then, the probability after weight adjustment and multiple probability updates is normalized to ensure that the sum of the probabilities of all behaviors is 1. The specific form of probability update is: in, For drones Behavior The probability of Based on drone Types For behavior The weight of is the total number of drones in the fleet.

10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 9.

Images