Tracking escape coverage control method and device based on differential game, and storage device
Through a differential game-based method, the multi-agent system in the tracking and escape game is modeled, and the strategies of defenders and attackers are optimized, which solves the problem that defenders ignore area coverage and achieves more efficient defense and coverage performance. Ensure the safety of critical facilities.
Patent Information
- Application Number
- CN202411031961.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-15
AI Technical Summary
In the tracing and escape game, the defender only considers defense and ignores area coverage, resulting in the attacker's possible sneak attack and limited perception and movement.
Using a differential game-based approach, model multi-agent systems, considering the situation of defenders' perception and movement limitations, define control obstacle functions to optimize defense and attackers' strategies, and ensure that defenders achieve them when protecting critical facilities. Optimal coverage performance.
Improve the control accuracy and coverage performance of the defense system, ensure effective protection of critical facilities, and prevent attackers from successfully invading.
Smart Images

Figure CN118963130A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of multi-agents, and in particular to a tracking escape coverage control method, device and storage device based on differential game. Background Art
[0002] A multi-agent system is a system composed of a group of autonomous, interacting agents. Each agent can perceive the environment, make decisions and perform actions, and cooperate or compete with other agents to achieve the system's goals. Multi-agent systems are widely used in autonomous driving, drone formations, and robot cooperation. Non-cooperative game is an important model in game theory, which describes the situation when there is no cooperation between the participants and each participant makes his own decision. In recent years, non-cooperative game has been widely used in multi-agent engineering problems such as drone swarms, microgrid control, and wireless communications.
[0003] The Pursuit Escape Game is a non-cooperative game involving three players: a target, a defender, and an attacker. The defender aims to intercept the attacker before he reaches the target, while the attacker tries to avoid the defender's detection and attack the target. The target can be either stationary or mobile, the latter trying to evade the attack.
[0004] In previous studies, only the single-target strategy of the defender was considered, that is, the defender only considered defense and did not consider other things. In addition, previous studies often considered the situation where the perception of the intelligent agent is unlimited. However, in actual situations, the defender's action range and perception range are often limited. When protecting critical facilities, a defender who is too far away from the target may inadvertently allow the attacker to succeed. In order to reduce this risk, it becomes crucial to limit the defender's actions. In addition, in addition to considering defense against attackers, the defender must also consider coverage of the target area to prevent sneak attacks by attackers. Summary of the invention
[0005] In order to solve the problem that the defender in the tracking and escaping game only considers defense and misses the coverage of some areas, resulting in a sneak attack by the attacker, the present invention provides a tracking and escaping coverage control method, device and storage device based on differential game. A tracking and escaping coverage control method based on differential game mainly includes:
[0006] S1: For n defenders d i A multi-agent system consisting of a single attacker a, i = 1, 2, ..., n, uses a differential game model to model the pursuit escape game;
[0007] S2: Considering the situation where the defender's actions are restricted, we can obtain the objective functions of the defender and the attacker, and obtain the optimal strategies of the defender and the attacker under this objective function.
[0008] S3: When the optimal strategy of S2 is considered, the defender's perception is added to obtain the objective function of the defender and the attacker, and under this objective function, the optimal strategy of the defender and the attacker is obtained;
[0009] S4: When considering the optimal strategy under the conditions of defender perception and action restriction, the defender's coverage performance is considered, and the control obstacle function is defined to obtain the final optimal strategy;
[0010] S5: The multi-agent system uses the final optimal strategy to achieve optimal control of the automation project.
[0011] A storage device stores instructions and data for implementing a tracking escape coverage control method based on differential game.
[0012] A tracking and escaping coverage control device based on differential game includes: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement a tracking and escaping coverage control method based on differential game.
[0013] The technical solution provided by the present invention has the following beneficial effects: i The multi-agent system composed of a single attacker a and a single attacker a uses a differential game model to model the pursuit escape game; the pursuit escape game mainly simulates some real-life scenarios, such as the scenario of protecting key facilities from external invasion. The advantage is that when protecting key facilities, better and more efficient defense against external invasion can be guaranteed. The key facilities are the targets mentioned in the present invention. Defenders are set to protect key facilities and defend against invasions by external attackers. The present invention also takes into account the perception and action limitations of the defenders, and ensures that the coverage performance of the defenders obtains the final optimal strategy; the multi-agent system can achieve optimal control of the automation project using the final optimal strategy, greatly improving the control accuracy of the intelligent system. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:
[0015] Figure 1 It is a flow chart of a tracking and escape coverage control method based on differential game in an embodiment of the present invention.
[0016] Figure 2 Schematic diagram of the coverage of the defender in the embodiment of the present invention.
[0017] Figure 3 1 is a schematic diagram of simulation results in an embodiment of the present invention, a is a simulation diagram when t=0, b is a simulation diagram when t=1, and c is a simulation diagram when t=1.4.
[0018] Figure 4 It is a schematic diagram of the operation of the hardware device in the embodiment of the present invention. DETAILED DESCRIPTION
[0019] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.
[0020] Please refer to Figure 1 , Figure 1 This is a flowchart of a tracking and escaping coverage control method based on differential game in an embodiment of the present invention. The present invention uses a differential game model to model the tracking and escaping game, considers the defender's perception and action restrictions in the game, and finds its optimal strategy. In addition, the defender's coverage performance of the target area is considered, and the control obstacle function is used to constrain it, ensuring the defender's coverage performance in the game, which is of great significance to real-life scenarios such as protecting key settings from external threats and protecting assets from intrusion. It includes:
[0021] S1: For n defenders d i A multi-agent system consisting of a single attacker a, i = 1, 2, ..., n, uses a differential game model to model the pursuit escape game;
[0022] S2: Considering the situation where the defender's actions are restricted, we can obtain the objective functions of the defender and the attacker, and obtain the optimal strategies of the defender and the attacker under this objective function.
[0023] S3: Add consideration of the defender's perception, obtain the objective function of the defender and the attacker, and obtain the optimal strategy of the defender and the attacker under the objective function;
[0024] S4: While considering the defender’s perception and action first, we add the defender’s coverage performance and define a control obstacle function to ensure the realization of this function, thus obtaining the final optimal strategy.
[0025] S5: The multi-agent system uses the final optimal strategy to achieve optimal control of the automation project. The optimal strategy is used to control the motion input of the agent. For example, the optimal strategy of the attacker is the input on the x and y axes, and the optimal strategy of the defender is the input on the angular velocity. Finally, the optimal action route is obtained according to the optimal strategy. Because after the coverage performance is increased, the constraint of the coverage performance is implemented by using an additional defined control obstacle function. At this time, the optimal strategy is obtained according to the original objective function. After obtaining the optimal strategy, a QP plan is required to re-solve the optimal strategy considering the coverage performance. This QP plan contains the control obstacle function.
[0026] The specific steps are as follows:
[0027] First, the system is modeled using differential game theory.
[0028] Consider a system with n defenders and a single attacker, where the defender group is denoted by d = {d1, d2, ... d n}, in order to limit the defender's range of action, it is restricted to the boundary of the circle, and its input is defined as use represents its position information, r represents the radius of the circle where the action is limited, and θ i (t) represents the angle of the ith defender. The position is expressed in polar coordinates as (r, θ). However, since the defender's actions are restricted to the boundary of the circle, the defender's r remains unchanged. Only the coverage angle changes. The position represented by P is the coordinate of the Cartesian coordinate system. The attacker's position and input are P a (t) = [x a (t)y a (t)] T and make Define the global state variable z(t)=col{z1(t),…,z n (t),P a (t)}∈R 2(n+1) , the state equation is:
[0029]
[0030] in, is the global input of the defender,
[0031]
[0032]
[0033] According to the mathematical model established above, the problem is modeled as a mathematical model that describes the problem. First, the situation where the defender's actions are restricted is considered, that is, its coverage performance and perception ability are not considered. The objective functions of the defender and the attacker are defined as follows:
[0034]
[0035] in, represents the position of the defender, P a (t) represents the attacker's position, z(t) represents the global state variable, represents the defender’s input, u d (t) represents the global input of the defender, represents the two-dimensional position coordinates of the i-th defender, P a(T) is the two-dimensional position coordinate of the attacker, z(T) represents the global variable at time T, T represents the terminal time, represents the terminal cost parameter related to the distance from the ith defender to the attacker, represents the instantaneous cost parameter related to the distance from the ith defender to the attacker, Represents attacker a to the i-th defender d i The control cost, R a represents the attacker's control cost parameter, represents the control cost parameter of the ith defender, is a parameter related to the terminal distance, Represents attacker a to the i-th defender d i The final cost parameter, represents the i-th defender d i To the final cost parameter of attacker a, represents the terminal cost parameter of the distance from the attacker to the target, I2 represents the two-dimensional unit matrix, and the parameters with f(F) are all terminal cost parameters. is a two-dimensional diagonal matrix, where the elements on the diagonal are Similar, among them The terminal cost parameter representing the distance from the attacker to the target, is a parameter related to the instantaneous cost, Represents attacker a to the i-th defender d i The instantaneous cost of represents the i-th defender d i The instantaneous cost to attacker a is represents the instantaneous cost of the target distance to attacker a, and R a =r a I2 is a parameter related to control cost. and r a is a parameter related to control cost, represents the i-th defender d i The control cost parameter, r a represents the control cost parameter of attacker a, since u a It is two-dimensional, that is, the input is in the x direction and the y direction respectively. In order to satisfy the matrix multiplication, there is R a =r a I2, but u d is one-dimensional, that is, angular velocity, so we use Convert it into two dimensions, and accordingly, we have C(t)=diag{C1(t),C2(t),…,Cn (t)},C i (t) = [-rsinθ i (t)rcosθ i (t)] T , Conclusion 1: Under the above objective function, the optimal strategies of the defender and the attacker are Where K d (t) represents the coefficient of the global variable in the defender’s optimal strategy, K a (t) represents the coefficient of the global variable in the attacker’s optimal strategy, K d (t) and K a (t) Satisfaction
[0036]
[0037] K d (T) = F d , K a (T) = F a .
[0038] By solving the above two differential equations, we can get K d (t) and K a (t).
[0039] Next, we consider the case where the defenders’ perception and actions are limited. We define the A matrix as the adjacency matrix of the network between defenders, which satisfies the following condition: i to d j When there is communication, a ij =1, otherwise 0. In addition, a ii = 0. Define the diagonal matrix of the defender's perception of the attacker When the defender i When an attacker is detected, s ii =1, otherwise 0. Next, define the following augmented matrix
[0040]
[0041] Defining Matrix Functions
[0042]
[0043] This function represents the defender d i Perceive the attacker's information at time t. In addition, define the defender d i The profit matrix
[0044]
[0045] in, represents the defender’s payoff matrix, Define the objective functions of the defender and attacker
[0046]
[0047] in
[0048]
[0049] ΔQ d (t) = Q d (t)-Q d , ΔQ a (t) = Q a (t)-Q a .
[0050] Conclusion 2: The optimal strategies for the defender and attacker are and in
[0051] Proof: For the defender, define the value function
[0052] V d (t,z(t))=z T (t)K d (t)z(t) (11)
[0053] but
[0054]
[0055] Integrating both sides of the equation from 0 to T, we have
[0056]
[0057] Since V a (T,z(T))=z T (T)K a (T)z(T),K a (T) = F a ,have
[0058]
[0059] Similarly, for attackers, there are
[0060]
[0061]
[0062] According to the above, are the optimal strategies for the defender and attacker, respectively.
[0063] Next, consider the defender's coverage performance, such as Figure 2 As shown, assuming that the defender's coverage radius is r d , its action is limited to a circle with a radius of r, and its coverage angle is 2β. According to the cosine theorem, Then β=arccos(1-0.5(r d / r) 2 ), and at least Only a defender can ensure coverage without blind spots.
[0064] Assuming the number of defenders That is, to ensure that the number of defenders can achieve full coverage of the coverage area. Since there are overlapping angles between defenders, the angle difference between the two farthest defenders is considered. The distance between the two defenders is The corresponding radius is R, that is, when the distance between the attacker and the target is R, at least one defender can sense the attacker and notify its neighbors to defend, where where r d ≤r, when r d >r, r≤R<r+r d By controlling The size of L, which controls the maximum distance L between two adjacent defenders max , we can control the defender's coverage performance and define the following control barrier function to ensure this performance
[0065]
[0066] is an n-dimensional row vector whose i-th element is 1 and j-th element is -1.
[0067] In addition, in order to ensure the coverage performance of the defender in the process of defending against the attacker after sensing the attacker, the difference in the distance between the defender and its two neighbors is limited. The following control barrier function is defined
[0068]
[0069] There are two control stations and therefore two control barrier functions h, the first one has a subscript of 1 and the second one has a subscript of 2. The superscript indicates the i-th defender, so, represents the second control barrier function of the ith defender. a represents the parameter of the control barrier function. represents the first control barrier function of the ith defender, is an n-dimensional row vector whose i-th element is 1 and j-th element is -1. j1 and j2 represent the subscripts of the two neighbors of the i-th defender. j1, j2 represent the subscripts of the two neighbors of the ith defender, is an n-dimensional row vector whose i-th element is 1 and j1-th element is -1. is an n-dimensional row vector whose i-th element is 1 and j2-th element is -1. represents the transpose of the position of the i-th defender.
[0070] In summary, considering the situation where the defender's actions and perception are limited, in order to ensure its coverage performance, its optimal strategy can be obtained by solving the following QP planning:
[0071]
[0072] Among them, z(t) represents the global state variable, G d (t) represents the defender’s payoff matrix, I d (t) represents the information of the attacker perceived by the defender at time t, α1 and α2 are the parameters of the two control barrier functions, They represent the two control barrier functions of the ith defender, They represent the first-order derivatives of the two control barrier functions of the i-th defender respectively.
[0073] Finally, numerical simulation is used to verify the conclusions.
[0074] In the numerical simulation, a model with four defenders and a single attacker is considered, and the adjacency matrix between the defenders is
[0075]
[0076] Let r d =7, r=8, we can get β=51.89°, which can meet the requirement of no blind spot coverage of the target area. In addition, the capture radius of the defender and attacker are set to and In addition, the parameters of the control barrier function are defined as a=3,L max = 11. Let the initial positions of the defender and attacker be θ(0) = [0, 3π / 4, π, 5π / 4] T and P a (0) = [25, -25] T . Define relevant parameters and They are {2,2,2,2} and {2,2,2,2,4} respectively. and are {1,1,1,1} and {1,1,1,1,1} respectively. and r aare {1,1,1,1} and 0.8 respectively. Let the payoff matrix of all defenders be Consider a game in [0,5] time with a discrete step length of 0.01. The simulation results are as follows Figure 3 As shown in Figure 1, at t=0, as shown in Figure (a), each agent is at its initial position. At t=1, as shown in Figure (b), the defender senses the attacker. Before this moment, due to the effect of the control obstacle function, the defender gradually moves to the position that maximizes the coverage performance, and its coverage range increases compared to the initial moment. At t=1.4, as shown in Figure (c), the attacker is captured by the defender and the game ends. Figure 3 In the figure, the small circle with a black dot in the center represents the attacker, the large circle with a black dot in the center represents the circle where the defender's actions are restricted, and the concentric circles represent the defender.
[0077] See also Figure 4 , Figure 4 It is a schematic diagram of the working of the hardware device of an embodiment of the present invention, and the hardware device specifically includes: a tracking escape coverage control device 401 based on differential game, a processor 402 and a storage device 403.
[0078] A tracking and escaping coverage control device 401 based on differential game: The tracking and escaping coverage control device 401 based on differential game implements the tracking and escaping coverage control method based on differential game.
[0079] Processor 402: The processor 402 loads and executes instructions and data in the storage device 403 to implement the tracking and escaping coverage control method based on differential game.
[0080] Storage device 403: The storage device 403 stores instructions and data; the storage device 403 is used to implement the tracking and escape coverage control method based on differential game.
[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A tracking and escaping coverage control method based on differential game, characterized by: include: S1: For n defenders d i A multi-agent system consisting of a single attacker a, i = 1, 2, ..., n, uses a differential game model to model the pursuit escape game; S2: Considering the situation where the defender's actions are restricted, we can obtain the objective functions of the defender and the attacker, and obtain the optimal strategies of the defender and the attacker under this objective function. S3: When the optimal strategy of S2 is considered, the defender's perception is added to obtain the objective function of the defender and the attacker, and under this objective function, the optimal strategy of the defender and the attacker is obtained; S4: When considering the optimal strategy under the conditions of defender perception and action restriction, the defender's coverage performance is considered, and the control obstacle function is defined to obtain the final optimal strategy; S5: The multi-agent system uses the final optimal strategy to achieve optimal control of the automation project.
2. The tracking and escaping coverage control method based on differential game as claimed in claim 1, characterized in that: In S1, the state equation of the established model is: Among them, u d (t) represents the global input of the defender, u a (t) represents the attacker’s global input, B d (t) represents the coefficient of the defender’s global input, B a (t) represents the coefficient of the attacker’s global input.
3. The tracking and escaping coverage control method based on differential game as claimed in claim 1, characterized in that: In S1, when considering the situation where the defender's actions are restricted, the objective functions of the defender and the attacker are: in, represents the position of the defender, P a (t) represents the attacker's position, z(t) represents the global state variable, represents the defender’s input, u d (t) represents the global input of the defender, represents the two-dimensional position coordinates of the i-th defender, P a (T) is the two-dimensional position coordinate of the attacker, z(T) represents the global variable at time T, T represents the terminal time, represents the terminal cost parameter related to the distance from the ith defender to the attacker, represents the instantaneous cost parameter related to the distance from the ith defender to the attacker, Represents attacker a to the i-th defender d i The control cost, R a represents the control cost parameter of attacker a, represents the control cost parameter of the ith defender, is a parameter related to the terminal distance, Represents attacker a to the i-th defender d i The final cost parameter, represents the i-th defender d i To the final cost parameter of attacker a, represents the terminal cost parameter of the distance from the attacker to the target, I2 represents the two-dimensional unit matrix, is a two-dimensional diagonal matrix, and is a parameter related to the instantaneous cost, Represents attacker a to the i-th defender d i The instantaneous cost of represents the i-th defender d i The instantaneous cost to attacker a is represents the instantaneous cost of the target distance to attacker a, and R a =r a I2 is a parameter related to control cost, represents the i-th defender d i The control cost parameter, r a represents the control cost parameter of attacker a, C(t)=diag{C1(t),C2(t),...,C n (t)},C i (t) = [-rsinθ i (t)rcosθ i (t)] T , 4. The tracking and escaping coverage control method based on differential game as claimed in claim 3, characterized in that: In S2, the optimal strategies of the defender and attacker are: and B d (t) represents the coefficient of the defender’s global input, B a represents the coefficient of the attacker's global input, the superscript T represents the transpose, K d (t) represents the coefficient of the global variable in the defender’s optimal strategy, K a (t) represents the coefficient of the global variable in the attacker’s optimal strategy.
5. The tracking and escaping coverage control method based on differential game as claimed in claim 4, characterized in that: In S3, when considering the defender's perception and action constraints, the objective functions of the defender and the attacker are: in, G d (t) represents the defender’s payoff matrix, Q d (t) represents the instantaneous cost parameter of the defender, Q a (t) represents the instantaneous cost parameter of the attacker, u d (t) represents the global input of the defender, I d (t) represents the information of the attacker perceived by the defender at time t.
6. The tracking and escaping coverage control method based on differential game as claimed in claim 5, characterized in that: When considering the defender's perception and action constraints, the optimal strategies of the defender and attacker are: and Among them, G d (t) represents the defender's payoff matrix.
7. The tracking and escaping coverage control method based on differential game as claimed in claim 1, characterized in that: In S4, two control barrier functions are designed, namely: Among them, a represents the parameter of the control barrier function, represents the first control barrier function of the ith defender, represents the second control barrier function of the ith defender, I2 represents the two-dimensional identity matrix, is an n-dimensional row vector whose i-th element is 1 and j-th element is -1, j1, j2 represent the subscripts of the two neighbors of the ith defender, is an n-dimensional row vector whose i-th element is 1 and j1-th element is -1. is an n-dimensional row vector whose i-th element is 1 and j2-th element is -1. represents the transpose of the i-th defender's position.
8. The tracking and escaping coverage control method based on differential game as claimed in claim 7, characterized in that: In S4, the final optimal strategy is obtained by solving the following QP planning: Among them, z(t) represents the global state variable, G d (t) represents the defender’s payoff matrix, I d (t) represents the information of the attacker perceived by the defender at time t, α1 and α2 are the parameters of the two control barrier functions, They represent the two control barrier functions of the ith defender, They represent the first-order derivatives of the two control barrier functions of the i-th defender respectively.
9. A storage device, characterized in that: The storage device stores instructions and data for implementing the tracking escape coverage control method based on differential game as described in any one of claims 1 to 8.
10. A tracking and escaping coverage control device based on differential game, characterized in that: include: Processor and storage device; the processor loads and executes instructions and data in the storage device to implement the tracking escape coverage control method based on differential game as described in any one of claims 1 to 8.
