Distributed unmanned aerial vehicle formation cooperative control method based on non-cooperative game

Through the distributed unmanned aerial vehicle formation collaborative control method based on non-cooperative games, the security risks and communication resources problems caused by central node dependence are solved, and the low-cost and high-privacy formation control of the drone group is realized, ensuring complexity Complete the task.

CN118938995APending Publication Date: 2024-11-12HARBIN INST OF TECH AT WEIHAI
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410978400.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing UAV fleet control technology is highly dependent on central nodes, and has operational interruptions and security risks during failure or attacks, and communication resources are tight and difficult to expand, so it is impossible to effectively optimize complex optimization and control problems.

Method used

Adopt a distributed drone formation collaborative control method based on non-cooperative games, establish a drone model through kinematic analysis, build a virtual leader using the leadership-follower model, design and optimize the objective function, and draw the goals of the drone individual Functions are converted into optimization problems based on local information, and a distributed drone formation algorithm is designed to realize the target formation of the drone group.

Benefits of technology

It realizes full distributed control, reduces costs, alleviates the shortage of communication resources, ensures the privacy of drone data, and ensures the completion of the formation tasks of drone groups in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118938995A_ABST
    Figure CN118938995A_ABST
Patent Text Reader

Abstract

The invention relates to the field of unmanned aerial vehicle flight formation control, in particular to a distributed unmanned aerial vehicle formation cooperative control method based on a non-cooperative game, which comprises the following steps: S1, establishing a four-rotor unmanned aerial vehicle kinematic model based on kinematic analysis; s2, constructing an unmanned aerial vehicle group leaded by a virtual leader by using a leader-follower model, and defining an expected relative position error between each unmanned aerial vehicle and the virtual leader; s3, based on the requirements of the formation task, designing an optimization objective function of the unmanned aerial vehicle group formation task; s4, based on the designed aggregation item, converting the target function of the unmanned aerial vehicle individual into an optimization problem based on local information; and S5, designing a game-based distributed unmanned aerial vehicle formation algorithm, and controlling the unmanned aerial vehicle group to achieve the target formation. The algorithm of the invention greatly alleviates the condition of tension of communication resources, ensures the privacy of unmanned aerial vehicle data, and can perform cooperative control on the unmanned aerial vehicle group at a lower cost compared with centralized control and distributed control.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of UAV flight formation control, and in particular to a distributed UAV formation collaborative control method based on non-cooperative game. Background Art

[0002] The coordinated operation of multiple drones is an important trend in the development of drone technology. With its scale advantage, multiple drone groups have more powerful data collection and analysis capabilities and are more capable of handling more complex and changeable tasks. Among them, drone formation is an important topic of multi-drone collaboration. At present, drone formation technology has shown wide application potential in many fields, including flight performances, express logistics, precision agriculture, urban traffic monitoring, military reconnaissance and combat, disaster relief, environmental monitoring, search and rescue, scientific data collection and other application scenarios. At present, multi-drone control includes centralized control and distributed control. Centralized control is currently more widely used. Centralized drone flight formation control technology usually relies on a strong central node. Point and high-precision positioning systems generally use high-performance computing equipment and combine real-time kinematic positioning (RTK) ground stations and communication facilities to coordinate control of drone groups. Its advantages are that the control method is simple and convenient for unified command and management, and it is easy to achieve precise control. However, due to its high dependence on the central node, once the central node fails or is attacked, the entire system may face operational interruptions and security risks, and communication resources are tight and difficult to expand. Introducing game control into distributed optimization can help optimize the decision-making process, formulate more effective strategies based on its own and other participants' information, and perform reasonable resource scheduling and allocation under limited resources to solve complex optimization control problems. To this end, the present invention proposes a distributed drone formation collaborative control method based on non-cooperative game. Summary of the invention

[0003] The purpose of the present invention is to provide a distributed UAV formation collaborative control method based on non-cooperative game to solve the problems raised in the above background technology.

[0004] To achieve the above purpose, a distributed UAV formation cooperative control method based on non-cooperative game includes the following steps:

[0005] Step S1, establishing a kinematic model of a quadrotor drone based on kinematic analysis;

[0006] Step S2, using the leader-follower model to construct a drone swarm led by a virtual leader, and defining the expected relative position error between each drone and the virtual navigator;

[0007] Step S3, based on the requirements of the formation task, designing the optimization objective function of the UAV group formation task;

[0008] Step S4, based on the designed aggregation term, the objective function of the individual drone is converted into an optimization problem based on local information;

[0009] Step S5, design a game-based distributed UAV formation algorithm to control the UAV swarm to achieve the target formation.

[0010] Preferably: the specific process of step S1 is as follows: the quad-rotor drone model is established according to the dynamic analysis as follows:

[0011]

[0012] The position of each drone i∈{1,2,...,N} is represented by the vector q in the Earth coordinate system. i =(x i ,y i ,z i ) T Given, where x i ,y i ,z i Represents its weight on the three coordinate axes respectively. The mass of the drone is m i , lift (control input) is u i , g is the acceleration due to gravity, θ, ψ represents the pitch angle, roll angle, and yaw angle respectively, and the above formula is rewritten into a compact matrix form:

[0013]

[0014] Among them, M i (q i )=m i I 3 , g i (q i )=(0,0,m i g) T , F i =R(0,0,u i ),

[0015]

[0016] Where c(·) represents cos(·), and s(·) represents sin(·).

[0017] Preferably: the specific process of step S2 is as follows, the position vector of each drone in the cluster is represented by q i =(x i ,y i ,z i ) T , the velocity vector is expressed as The position vector of the virtual navigator is denoted as q0 =(x 0 ,y 0 ,z 0 ) T , the velocity vector is expressed as The vector of the drone swarm is expressed in global form, and the group position vector of the entire drone swarm is The group velocity vector is Define the expected position error between UAV i and the virtual navigator as q i -q 0 =Δ i , when the formation is achieved:

[0018]

[0019] Preferably: the specific process of step S3 is as follows: the following objective function is designed for each drone:

[0020]

[0021] Each drone will minimize its own objective function. When the solution to the game, namely the Nash equilibrium, is reached, the change in the state of each drone will lead to an increase in the objective function of other participants. The design of the cost function satisfies that when the drone group reaches the Nash equilibrium, the drone group also meets the conditions for achieving formation and forms a target formation.

[0022] Preferably: the specific process of step S4 is as follows: design aggregation items The global information to be obtained is converted into an estimate of the aggregate term σ(q), and the objective function is transformed into

[0023] J i (q i ,q -i )=J i (q i ,σ(q));

[0024] The objective function subgradient is transformed into

[0025]

[0026] Preferably: Step S5 considers the designed objective function, solves the Nash equilibrium solution of the game problem, and designs a distributed control algorithm according to the gradient descent strategy.

[0027]

[0028] The algorithm consists of the following five main parts: i represents the decision state of UAV i, s i is the drone’s estimate of the aggregate term in the objective function, ri is the auxiliary variable for estimating the aggregation term of the drone pair, G i It is the gradient approximation of the drone’s search for the Nash equilibrium solution.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The algorithm of the present invention is based on non-cooperative game theory and can achieve fully distributed control without relying on a central node. Each drone independently runs the algorithm to reach a formation consensus. Compared with centralized control, distributed control can coordinate the control of drone groups at a lower cost.

[0031] 2. Each drone in the cluster only needs to exchange information with a limited number of drones, which greatly alleviates the shortage of communication resources.

[0032] 3. Moreover, the communication information is aggregated information and does not include direct speed and location information, which ensures the privacy of drone data. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of the present invention;

[0034] Figure 2 is a time-varying undirected graph of the communication topology of five quad-rotor drones of the present invention;

[0035] Figure 3 The (x, y) direction position change curves of the five UAVs and the virtual navigator set to move in a uniform linear motion according to the embodiment of the present invention;

[0036] Figure 4 The speed change curves of the five UAVs and the virtual navigator set to move in a uniform linear motion in the (x, y) direction according to the embodiment of the present invention;

[0037] Figure 5 It is a motion trajectory diagram of the (x, y) direction positions of five UAVs and a virtual navigator according to an embodiment of the present invention;

[0038] Figure 6 The (x, y) direction position change curves of five UAVs and a virtual navigator set to move at variable speeds according to an embodiment of the present invention;

[0039] Figure 7 The speed change curves of the five UAVs and the virtual navigator set to variable speed motion in the (x, y) direction according to the embodiment of the present invention;

[0040] Figure 8 1 is a motion trajectory diagram of the five UAVs and the virtual navigator in the (x, y) direction according to an embodiment of the present invention. DETAILED DESCRIPTION

[0041] 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 implementation regulations described 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.

[0042] Example

[0043] See also Figure 1 , FIG. 1 is a preferred embodiment of the present invention, a distributed UAV formation cooperative control method based on non-cooperative game, comprising the following steps:

[0044] Step S1, establishing a kinematic model of a quadrotor drone based on kinematic analysis;

[0045] Step S2, using the leader-follower model to construct a drone swarm led by a virtual leader, and defining the expected relative position error between each drone and the virtual navigator;

[0046] Step S3, based on the requirements of the formation task, designing the optimization objective function of the UAV group formation task;

[0047] Step S4, based on the designed aggregation term, the objective function of the individual drone is converted into an optimization problem based on local information;

[0048] Step S5, design a game-based distributed UAV formation algorithm to control the UAV swarm to achieve the target formation.

[0049] Step S1 is as follows: a quadrotor drone model is established based on dynamic analysis as follows:

[0050]

[0051] The position of each drone i∈{1,2,...,N} is represented by the vector q in the Earth coordinate system. i =(x i ,y i ,z i ) T Given, where x i ,y i ,z i Represents its weight on the three coordinate axes respectively. The mass of the drone is m i , lift (control input) is u i , g is the acceleration due to gravity, θ, ψ represents the pitch angle, roll angle, and yaw angle respectively, and the above formula is rewritten into a compact matrix form:

[0052]

[0053] Among them, M i (q i )=m i I 3 , g i (q i )=(0,0,m i g) T , F i =R(0,0,u i ),

[0054]

[0055] Step S2 is as follows: initialize the initial speed and position states of N drones and the trajectory parameters of the virtual navigator. The position vector of each drone in the cluster is represented by q i =(x i ,y i ,z i ) T , the velocity vector is expressed as The position vector of the virtual navigator is denoted as q 0 =(x 0 ,y 0 ,z 0 ) T , the velocity vector is expressed as The vector of the drone swarm is expressed in global form, and the group position vector of the entire drone swarm is The group velocity vector is Set the expected position error between drone i and the virtual navigator

[0056] q i -q 0 =Δ i ,

[0057] When the formation is completed, the following conditions must be met:

[0058]

[0059] Step S3 is as follows: the following objective function is designed for each drone:

[0060]

[0061] Each drone will minimize its own objective function and optimize the objective function. When the drone group reaches a Nash equilibrium, the drone group simultaneously meets the conditions for achieving formation and forms a target formation.

[0062] When the above optimization model reaches the Nash equilibrium solution, it also meets the needs of the formation task as follows:

[0063] (1) Given a generalized Nash equilibrium problem, for each player i, the objective function J i (x i ,x -i ) is convex and second-order differentiable, the strategy set Ω is closed convex, and the definition And F(x) is ω-strongly monotone, then the game problem under consideration has a unique solution.

[0064] Based on the content given in (1), we can calculate F(q) for this problem as

[0065]

[0066] in

[0067]

[0068] Matrix [M] ij is a positive definite matrix, there exists a positive definite matrix Q such that M = Q T Q, so for There exists a positive value of ω such that

[0069]

[0070] We get F(q) to be ω-strongly monotonic, and the objective function J i (x i ,x -i ) is a convex function and second-order differentiable, and the strategy set Ω is a closed convex set. According to (1), this game problem has a unique Nash equilibrium solution.

[0071] Next, we explain that the unique Nash equilibrium solution mentioned above can simultaneously satisfy the formation task requirements. (2) The vector is a Nash equilibrium solution if and only if its objective function About q i The gradient satisfies

[0072] According to (2), when the Nash equilibrium is reached, the drone swarm satisfies

[0073]

[0074] From the above formula, we can get That is, when the Nash equilibrium is reached, the conditions for forming a formation are met.

[0075] Step S4 is as follows: In step S3, the objective function and its gradient need to obtain the global UAV position information. In order to build an optimization model based on local information, the aggregation term is designed. The global information to be obtained is converted into an estimate of the aggregate term σ(q), and the objective function is transformed into

[0076] J i (q i ,q -i )=J i (q i ,σ(q));

[0077] The objective function subgradient is transformed into

[0078]

[0079] Step S5 is as follows: considering the designed objective function, solving the Nash equilibrium solution of the game problem, and designing a distributed control algorithm based on the gradient descent strategy

[0080]

[0081] The above algorithm consists of the following five main parts: i represents the decision state of UAV i, s i is the drone’s estimate of the aggregate term in the objective function, r i is the auxiliary variable for estimating the aggregation term of the drone pair, G i is the gradient approximation of the drone’s search for the Nash equilibrium solution.

[0082] The distributed algorithm of the present invention has two tasks, one of which is to i With r i Estimate the aggregated items, update the data through local information exchange, and finally reach the target value using s i Replace aggregate items The second is to independently update the control input of each UAV to achieve Nash equilibrium and complete the formation goal.

[0083] Reference Figure 2 For the adjacency matrix, a time-varying undirected graph is used to describe the communication topology of the five quadrotor drones, where the numbers in the circles represent the serial numbers of the drones, the horizontal lines represent the bidirectional propagation of information between the two drones, and the time-varying topology means that the structure of the communication topology changes with time. As shown in the figure, the communication topology was switched once in the second.

[0084] (3) If the communication topology remains undirected and the constant α satisfies the following conditions:

[0085]

[0086] Consider the following system, where r i (0) = 0 3 , the system dynamic equation is as follows:

[0087]

[0088] When t→+∞, holds and has an exponential convergence rate.

[0089] According to (3) above, even if the communication topology of the drone swarm is time-varying, as long as the undirected connectivity is maintained, the estimated value of the aggregation term can be ensured to converge to the actual value, which shows that in a complex environment, even if the communication topology changes over time, the drone swarm can achieve formation.

[0090] Bringing the drone model into the above algorithm, it can be rewritten as

[0091]

[0092] Then we prove that the equilibrium point of the algorithm is the Nash equilibrium point. When the algorithm is at the equilibrium point, the following equation holds:

[0093]

[0094] According to (4), as time approaches infinity, It will tend to the average value of the global position information of all drones, that is, From this we can conclude that:

[0095]

[0096] So we get

[0097]

[0098] According to (2), we know that q * It is the Nash equilibrium solution of the game.

[0099] On the contrary, when q * is the Nash equilibrium solution of the game, indicating Established, (3) ensures that the following equation holds:

[0100]

[0101] This makes Therefore, there exists r i * and Make It is the balance point of the algorithm, which shows that the balance point of the algorithm is consistent with the Nash equilibrium solution of the multi-UAV formation game problem.

[0102] The stability of the algorithm of the present invention is proved as follows:

[0103]

[0104] Specifically, the control law Perform identity transformation

[0105] where e i =F i (q i ,q -i )-G i (q i ,s i )

[0106] According to (3), e i The exponential converges to 0, defined

[0107] The control law is expressed as ε i Expressed as

[0108]

[0109] in Rewriting the distributed algorithm into a global form is

[0110]

[0111] represents the Kronecker product,

[0112] Consider the Lyapunov function

[0113] Obviously V(t)≥0, taking the derivative of V(t) yields

[0114]

[0115] in Due to q i Bounded and e i The exponential converges to 0, and we can get

[0116] Next, construct an expression

[0117]

[0118] right According to the Jacobian matrix, an upper bound of F(q) can be expressed as

[0119]

[0120] So From this we can get

[0121]

[0122] That is

[0123]

[0124] According to the algorithm, the conditions for reaching the equilibrium point are W(ε * ) can be obtained

[0125]

[0126] Right now By shifting the terms and integrating both sides of the equation, we can get

[0127]

[0128] Since t→+∞ It shows that ε→ε * , plus ε(t) is uniformly continuous, we get ε(t) converges show

[0129] The above has completed the proof that the algorithm converges to the equilibrium point. The drones in the drone swarm reach a consistent speed, maintain a given formation to fly, and complete the formation mission.

[0130] The following is a simulation example of the method of the present invention. For intuitive demonstration, the (x, y) coordinates of the drone group on the same altitude plane are controlled. The formation example when the virtual navigator moves at a uniform speed is set. The parameter α in the algorithm is set to 10. The position change curves of the five drones and the virtual navigator set to uniform linear motion in the (x, y) direction in this embodiment are as follows: Figure 3 As shown in the figure, the velocity change curves of the five drones and the virtual navigator set to uniform linear motion in the (x, y) direction in this embodiment are as follows: Figure 4 As shown, Figure 5 The following is a motion trajectory diagram of the (x, y) direction positions of the five UAVs and the virtual navigator. The following is an example of a formation when the motion of the virtual navigator is a variable speed motion. In this embodiment, the (x, y) direction position change curves of the five UAVs and the virtual navigator set to a variable speed motion are shown in FIG. Figure 6 As shown; the speed change curves of the five drones in this embodiment and the virtual navigator set to variable speed motion in the (x, y) direction Figure 7 As shown; Figure 8 The motion trajectory diagram of the five UAVs and the virtual navigator in the (x, y) direction. The above results show that this method may meet the requirements of multi-UAV formation tasks. Each UAV communicates through local information, and the communication content is not direct speed and position information, which ensures a certain degree of data privacy. Each UAV independently updates its own algorithm to complete the formation task.

[0131] The above content is a further detailed description of the present invention in combination with specific implementation methods. It cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, some simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the scope of protection determined by the claims submitted for the present invention.

Claims

1. A distributed UAV formation cooperative control method based on non-cooperative game, characterized in that: The steps include: Step S1, establishing a kinematic model of a quadrotor drone based on kinematic analysis; Step S2, using the leader-follower model to construct a drone swarm led by a virtual leader, and defining the expected relative position error between each drone and the virtual navigator; Step S3, based on the requirements of the formation task, designing the optimization objective function of the UAV group formation task; Step S4, based on the designed aggregation term, the objective function of the individual drone is converted into an optimization problem based on local information; Step S5, design a game-based distributed UAV formation algorithm to control the UAV swarm to achieve the target formation.

2. According to claim 1, a distributed UAV formation cooperative control method based on non-cooperative game, characterized in that: The specific process of step S1 is as follows: the quadrotor drone model is established according to the dynamic analysis as follows: The position of each drone i∈{1,2,...,N} is represented by the vector q in the Earth coordinate system. i =(x i ,y i ,z i ) T Given, where x i ,y i ,z i Represents its weight on the three coordinate axes respectively. The mass of the drone is m i , lift force is u i , g is the acceleration due to gravity, θ, ψ represents the pitch angle, roll angle, and yaw angle respectively, and the above formula is rewritten into a compact matrix form: Where M i (q i ) = m i I3,g i (q i ) = (0,0,m i g) T , F i = R(0,0,u i ), 3. The distributed UAV formation cooperative control method based on non-cooperative game according to claim 1 is characterized by: The specific process of step S2 is as follows: the position vector of each drone in the cluster is represented by q i =(x i ,y i ,z i ) T , the velocity vector is expressed as The position vector of the virtual navigator is expressed as q0 = (x0, y0, z0) T , the velocity vector is expressed as The vector of the drone swarm is expressed in global form, and the group position vector of the entire drone swarm is The group velocity vector is Define the expected position error between UAV i and the virtual navigator as q i -q0=Δ i , when the formation is achieved:

4. The distributed UAV formation cooperative control method based on non-cooperative game according to claim 1 is characterized by: The specific process of step S3 is as follows: the following objective function is designed for each drone Each drone will minimize its own objective function. When the solution to the game, namely the Nash equilibrium, is reached, the change in the state of each drone will lead to an increase in the objective function of other participants. The design of the cost function satisfies that when the drone group reaches the Nash equilibrium, the drone group also meets the conditions for achieving formation and forms a target formation.

5. The distributed UAV formation cooperative control method based on non-cooperative game according to claim 1, characterized in that: The specific process of step S4 is as follows: Design aggregation items The global information to be obtained is converted into an estimate of the aggregate term σ(q), and the objective function is transformed into J i (q i ,q -i )=J i (q i ,σ(q)); The objective function subgradient is transformed into 6. The distributed UAV formation cooperative control method based on non-cooperative game according to claim 1, characterized in that: The step S5 considers the designed objective function, solves the Nash equilibrium solution of the game problem, and designs a distributed control algorithm according to the gradient descent strategy. The algorithm contains the following five main components: i represents the decision state of UAV i, s i is the drone’s estimate of the aggregate term in the objective function, r i is the auxiliary variable for estimating the aggregation term of the drone pair, G i is the gradient approximation of the drone’s search for the Nash equilibrium solution.

Citation Information

Cited By

  • Unmanned aerial vehicle multi-target task scheduling method based on distributed cooperative game

    CN120255577A

  • Composite game reinforcement learning method and device for quad-rotor unmanned aerial vehicle cluster

    CN121879395A