A method and device for target allocation in a swarm attack and defense game.
By acquiring drone situational information, threat assessment, and hit rate calculation, a game theory model was established and optimized, solving the problem of suboptimal resource allocation in drone swarm offensive and defensive confrontations, and achieving more efficient decision-making and defense effects.
Patent Information
- Application Number
- CN202311691994.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-11
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-12-11
AI Technical Summary
In existing drone swarm offensive and defensive confrontations, the correlation of the opponent's strategies is ignored, resulting in suboptimal resource allocation and increased losses for the friendly side. Furthermore, existing research has not fully considered coordinated air combat missions.
By acquiring target information from both sides in the battle, conducting threat assessment and hit rate calculation, a static game model is established. Then, the rolling time-domain control method is used for optimization to solve the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy. Finally, the Nash equilibrium is solved by combining the minimax theorem or game learning algorithm.
It improves the real-time decision-making of drone offensive and defensive confrontations and the efficiency of defense systems, enhances strike capabilities and survivability, and optimizes resource allocation.
Smart Images

Figure CN117687436B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of weapon target allocation technology, and in particular to a method and device for target allocation in a game of offensive and defensive confrontation between unmanned aerial vehicle (UAV) swarms. Background Technology
[0002] In recent years, with the rapid development of miniaturization, intelligence, and swarm technology for unmanned aerial vehicles (UAVs), UAV swarm warfare has moved from theory to practical application, becoming the most active, innovative, and combat-oriented development direction in the military field, and an important innovative development path for generating new combat capabilities. In UAV swarm warfare, the allocation of multiple weapons and targets in UAV swarm offensive and defensive confrontations is a key issue that needs to be addressed.
[0003] The problem of multi-weapon, multi-target allocation in drone swarm offensive and defensive confrontation is to optimally achieve the combat intent by distributing multiple combat weapons to multiple targets, so as to maximize the damage effect of our drone weapon swarm on incoming targets and minimize our economic losses.
[0004] Current research on the multi-weapon, multi-target allocation problem in UAV swarm offensive and defensive confrontation mainly focuses on constructing nonlinear integer programming models to maximize the effectiveness of attacks on targets or minimize their survival value. The confrontation between multiple UAVs and an adversary is essentially a game-theoretic process, characterized by uncertainty, strong adversarial nature, and high dynamics. However, recent research on the multi-weapon, multi-target allocation problem in UAV swarm offensive and defensive confrontation has focused solely on the player's perspective, neglecting the correlation with the adversary's strategies. In UAV offensive and defensive confrontation, when facing the weapon target allocation problem, it is necessary to analyze the adversary's possible shifting behaviors under different strategies and their impact on the allocation of resources in different capability domains of our side. This helps to achieve optimal allocation of our resources, narrow the capability gap with the enemy, and reduce losses due to insufficient capabilities. Furthermore, existing UAV offensive and defensive confrontation strategies mainly revolve around offensive and defensive confrontation modeling and target allocation solutions, neglecting research on cooperative air combat missions. Summary of the Invention
[0005] This invention provides a method and apparatus for target allocation in a game of offensive and defensive combat involving unmanned aerial vehicle (UAV) swarms. The technical solution is as follows:
[0006] On the one hand, a target allocation method for drone swarm offensive and defensive game theory is provided. This method is implemented by electronic devices and includes:
[0007] S1. Acquire target information from both sides in the operation; target information from both sides includes: UAV attitude information, number of UAVs, UAV launch position, and UAV situational information;
[0008] S2. Conduct threat assessment on the UAV situational information to obtain the threat assessment sequence of both sides in the combat.
[0009] S3. Calculate the hit rate of both sides based on the number of drones, drone attitude information, and drone launch position;
[0010] S4. Based on the threat assessment sequence of both sides and the hit rate of both sides, establish a static game model;
[0011] S5. Optimize the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model.
[0012] S6. Solve the Nash equilibrium for the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides in the war; the optimal weapon target allocation strategy combination for both sides in the war includes: the optimal weapon target allocation strategy of each side in the war.
[0013] S7. Execute the optimal weapon target allocation strategy for each side in the combat.
[0014] Optionally, S2 performs a threat assessment on the target information of both sides in the operation, obtaining a threat assessment sequence for both sides, including:
[0015] S21. Select six threat indicators: superior-designated, relative angle, relative speed, relative altitude, key protected areas, and radar cross-section, and construct threat indicator models corresponding to the six threat indicators.
[0016] S22. Input the UAV situational information into the threat indicator model corresponding to the six threat indicators respectively to obtain six threat level values;
[0017] S23. Calculate the first weight of the six threat levels using the analytic hierarchy process (AHP), and calculate the second weight using the entropy method. Combine the first and second weights to obtain the threat assessment sequence for both sides in the conflict.
[0018] Optionally, S3 calculates the hit probability for both sides based on the number of drones, drone attitude information, and drone launch positions, including:
[0019] S31. Construct an initial hit rate function based on the azimuth angle in the UAV attitude information;
[0020] S32. Calculate the distance between the drones of the two combat sides based on the drone's starting position and the number of drones;
[0021] S33. Combine the initial hit rate function with the distance between the drones of both sides to construct the final hit rate function.
[0022] S34. Input the azimuth angle and the distance between the two combat drones from their attitude information into the final hit rate function to obtain the hit rate of both combat drones.
[0023] Optionally, S4 establishes a static game model based on the threat assessment sequence of both sides and the hit rate of the two sides, including:
[0024] S41. Treat the drones of both sides in the combat as participants in the game;
[0025] S42. The weapon target allocation scheme of the drones of both sides in the combat is used as a game strategy;
[0026] S43. Calculate the attack value of both sides using the threat assessment sequence and the hit rate of both sides, and use the matrix constructed from the attack value as the game payoff matrix.
[0027] Optionally, S5 optimizes the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model, including:
[0028] S51. Preset the same time interval; use the time interval to divide the entire time phase of the attack and defense confrontation process to obtain the threat assessment time node;
[0029] S52. At each threat assessment time point, clear up the drone strike situation of both sides in the previous round, and update the number of drones, drone threat assessment sequence and hit rate of both sides in the new round.
[0030] S53. Based on the number of drones of both sides in the new round, the drone threat assessment sequence, and the hit rate, the static game model is optimized to obtain a multi-stage dynamic game model.
[0031] Optionally, S6 solves the Nash equilibrium for the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides, including:
[0032] The min-max theorem is used to determine whether the game model has a pure-strategy Nash equilibrium. If it does, the solution is obtained directly. If not, the mixed-strategy Nash equilibrium is solved using the payoff equivalence method; or a game learning algorithm is used to solve the Nash equilibrium of the game model to obtain the optimal weapon target allocation strategy combination for both sides.
[0033] Optionally, the minimax theorem method can be used to solve for the combination of objective allocation strategies in a multi-stage dynamic game model, including:
[0034] (1) Use the minimax theorem to obtain the pure strategy Nash equilibrium solution for both sides in the battle;
[0035] (2) When the strategic gains of both sides are equal, the combination of strategies of both sides is taken as the optimal combination of weapon target allocation strategies.
[0036] (3) If the model does not have a pure policy Nash equilibrium, use the payoff equivalence method to obtain the mixed policy Nash equilibrium solution of the two sides in the battle;
[0037] (4) Both sides choose pure strategies with a certain mixed strategy probability distribution. The expected payoff of any pure strategy is the same, and the sum of the mixed probability distributions is 1.
[0038] Optionally, a game learning algorithm is used to solve for the combination of objective allocation strategies in a multi-stage dynamic game model, including:
[0039] (1) Initialize the mixed strategies of the two combatants, i.e. the probability distribution of the strategy set;
[0040] (2) Calculate the expected gains of both sides under the current hybrid strategy;
[0041] (2) Both sides update their hybrid strategy according to the expected payoff based on a specific update rule, which is specified by the game learning algorithm.
[0042] (4) The algorithm converges to Nash equilibrium and obtains a stable policy probability distribution.
[0043] On the other hand, a target allocation device for drone swarm offensive and defensive game is provided. This device is applied to the target allocation method for drone swarm offensive and defensive game. The device includes:
[0044] The acquisition unit is used to acquire target information from both sides in the combat. The target information from both sides includes: UAV attitude information, number of UAVs, UAV launch position, UAV situational information, and UAV speed.
[0045] The assessment unit is used to assess the credibility of target information from both sides in the combat and obtain a threat assessment sequence for both sides.
[0046] The computing unit is used to calculate the hit rate of both sides in combat based on the number of drones, drone attitude information, drone launch position, and drone speed.
[0047] The building unit is used to establish a static and dynamic game model based on the threat assessment sequence and hit rate of both sides in the combat.
[0048] The optimization unit is used to optimize the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model.
[0049] The solution unit is used to solve the Nash equilibrium of a multi-stage dynamic game model and obtain the optimal weapon target allocation strategy combination.
[0050] The execution unit is used to execute the optimal weapon target allocation strategies of both sides in the combat.
[0051] Optionally, the evaluation unit is used for,
[0052] Six threat indicators were selected: superior-designated, relative angle, relative speed, relative altitude, key protected areas, and radar cross-section. Threat indicator models were constructed corresponding to these six threat indicators.
[0053] Input the UAV situational information into the threat indicator model corresponding to the six threat indicators to obtain six threat level values;
[0054] The first weight of the six threat levels is calculated using the analytic hierarchy process (AHP), and the second weight is calculated using the entropy method. The first and second weights are then fused to obtain the threat assessment sequence for both sides in the conflict.
[0055] Optionally, the computing unit is used for,
[0056] Construct an initial hit rate function based on the azimuth angle in the UAV attitude information;
[0057] Calculate the distance between the drones of the two sides based on their starting positions and the number of drones.
[0058] The initial hit rate function is combined with the distance between the drones of both sides to construct the final hit rate function.
[0059] The azimuth angle and the distance between the two combat drones are input into the final hit rate function to obtain the hit rate of both combat drones.
[0060] Optionally, building blocks are used for,
[0061] The drones of both sides in the battle are considered as participants in the game;
[0062] The weapon target allocation scheme of the drones of both sides in the combat is used as a game strategy;
[0063] The attack value of both sides is calculated by using the threat assessment sequence of both sides and the hit rate of both sides, and the matrix constructed by the attack value is used as the game payoff matrix.
[0064] Optionally, the optimization unit is used for,
[0065] Preset the same time interval; use the time interval to divide the entire time phase of the attack and defense confrontation process to obtain the time node for threat assessment;
[0066] At each threat assessment time point, the drone strike situation of both sides in the previous round is cleared, and the number of drones, drone threat assessment sequence and hit rate of both sides in the new round are updated.
[0067] Optionally, the solver element is used for,
[0068] The min-max theorem is used to determine whether the game model has a pure-strategy Nash equilibrium. If it does, the solution is obtained directly. If not, the mixed-strategy Nash equilibrium is solved using the payoff equivalence method; or a game learning algorithm is used to solve the Nash equilibrium of the game model to obtain the optimal weapon target allocation strategy combination for both sides.
[0069] Optionally, the minimax theorem method can be used to solve for the combination of objective allocation strategies in a multi-stage dynamic game model, including:
[0070] (1) Use the minimax theorem to obtain the pure strategy Nash equilibrium solution for both sides in the battle;
[0071] (2) When the strategic gains of both sides are equal, the combination of strategies of both sides is taken as the optimal combination of weapon target allocation strategies.
[0072] (3) If the model does not have a pure policy Nash equilibrium, use the payoff equivalence method to obtain the mixed policy Nash equilibrium solution of the two sides in the battle;
[0073] (4) Both sides choose pure strategies with a certain mixed strategy probability distribution. The expected payoff of any pure strategy is the same, and the sum of the mixed probability distributions is 1.
[0074] Optionally, a game learning algorithm is used to solve for the combination of objective allocation strategies in a multi-stage dynamic game model, including:
[0075] (1) Initialize the mixed strategies of the two combatants, i.e. the probability distribution of the strategy set;
[0076] (2) Calculate the expected gains of both sides under the current hybrid strategy;
[0077] (3) Both sides update their hybrid strategy according to the expected payoff based on a specific update rule, which is specified by the game learning algorithm.
[0078] (4) The algorithm converges to Nash equilibrium and obtains a stable policy probability distribution.
[0079] On the other hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-mentioned target allocation method for the offensive and defensive game of drone swarms.
[0080] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, and the at least one instruction is loaded and executed by a processor to implement the above-mentioned target allocation method for the offensive and defensive game of drone swarms.
[0081] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:
[0082] This invention first employs a reconnaissance and early warning system to monitor flying targets and obtain target information from both sides in combat. Threat assessments are then performed on this information to obtain a threat assessment sequence for both sides. Based on the number of UAVs, their attitude information, and their departure positions, the hit rate for both sides is calculated. A static game model is established using the hit rate and threat assessment sequence. This static game model is then optimized using a rolling time-domain control method to obtain a multi-stage dynamic game model. Finally, the Nash equilibrium is solved in the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides, and the optimal weapon target allocation strategies for each side are executed. This invention addresses the weapon target allocation problem in UAV offensive and defensive confrontation by proposing a method using the min-max theorem or game learning algorithm to solve the Nash equilibrium, thereby obtaining the optimal weapon target allocation strategy combination for both sides. Simultaneously, the rolling time-domain control method optimizes the static game model to obtain a multi-stage dynamic game model for weapon target allocation in UAV offensive and defensive confrontation, improving the real-time nature of decision-making, the efficiency of the defense system, and the success rate. Attached Figure Description
[0083] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0084] Figure 1 This is a flowchart of a target allocation method for a drone swarm offensive and defensive game provided by an embodiment of the present invention;
[0085] Figure 2 This is a schematic diagram of the target allocation structure in a drone swarm offensive and defensive game according to an embodiment of the present invention;
[0086] Figure 3 This is a block diagram of a target allocation device for a drone swarm offensive and defensive game provided in an embodiment of the present invention;
[0087] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation
[0088] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.
[0089] This invention provides a target allocation method for a drone swarm offensive and defensive game, which can be implemented by an electronic device, such as a terminal or a server. Figure 1 The flowchart shown illustrates the target allocation method for a drone swarm attack and defense game. The processing flow of this method may include the following steps:
[0090] S1. Obtain target information from both sides in the battle.
[0091] The target information for both sides in the operation may include: drone attitude information, number of drones, drone departure position, and drone situational information.
[0092] In one feasible implementation, target information of both sides in the combat can be obtained through monitoring by a reconnaissance and early warning system.
[0093] S2. Conduct threat assessment on the UAV situational information and obtain the threat assessment sequence of both sides in the operation.
[0094] Optionally, the specific operation of S2 may include the following steps S21-S23:
[0095] S21. Select six threat indicators: superior-designated, relative angle, relative speed, relative altitude, key protected areas, and radar cross-section, and construct threat indicator models corresponding to the six threat indicators.
[0096] The following section introduces the threat index model for protecting key locations and the radar cross-section threat index model:
[0097] In one feasible implementation, the threat index model for defending key locations can be represented by the following formula (1):
[0098]
[0099] Among them, Category 1 key locations indicate areas with a high probability of being attacked and extremely important strategic value; Category 2 key locations indicate areas with weak self-defense capabilities and a relatively high probability of being attacked; Category 3 key locations indicate military defense key locations with strong self-defense capabilities, a high probability of being attacked, and low tolerance for losses; Category 4 key locations indicate civilian defense key locations with strong self-defense capabilities, a low probability of being attacked, and low tolerance for losses; and Category 5 key locations can indicate defense key locations with weak self-defense capabilities, which may become targets of cluster attacks and have high tolerance for losses.
[0100] In one feasible implementation, the threat index model of radar cross-section can be expressed by the following formula (2):
[0101]
[0102] Where 'a' represents the radar cross-section, which varies for each viewpoint and different radar frequencies.
[0103] S22. Input the UAV situational information into the threat indicator model corresponding to the six threat indicators respectively to obtain six threat level values;
[0104] S23. Calculate the first weight of the six threat levels using the analytic hierarchy process (AHP), and calculate the second weight using the entropy method. Combine the first and second weights to obtain the threat assessment sequence for both sides in the conflict.
[0105] Fusion computing aims to comprehensively map the relative importance of threat indicators.
[0106] In one feasible implementation, the fusion weight can be represented by the following formula (3):
[0107]
[0108] Among them, X j The fusion weights are represented by α, β, and Ω. j The second weighting represents the threat level value. This represents the first weight of each threat level value. j can take values of 1, 2, ..., J, where J represents the number of threat indicators.
[0109] The values of the fusion ratios of the first and second weights need to be determined comprehensively based on the quality and reliability of the information on the enemy situation obtained by the air defense weapon system and the reliability of the decision-makers, and must satisfy α+β=1.
[0110] In one feasible implementation, the threat assessment sequence of both combatants can be represented by the following formula (4):
[0111] TS = U·X j (4)
[0112] Among them, U=(u ij ) I×J This represents the threat indicator data matrix, where I represents the number of targets sorted, J represents the number of threat indicators, and TS represents the threat assessment sequence of both sides in the conflict.
[0113] S3. Calculate the hit rate of both sides based on the number of drones, drone attitude information, and drone departure position.
[0114] The drone attitude information includes the drone's position information and attitude angle information.
[0115] Optionally, the specific operation of S3 may include the following steps S31-S34:
[0116] S31. Construct an initial hit rate function based on the azimuth angle in the UAV attitude information;
[0117] In one feasible implementation, the relationship between the azimuth angles of the two combat parties' UAVs can be expressed by the following formula (5):
[0118]
[0119] Where η represents the azimuth angle between the drones of the two combat parties. This represents the position vector of the W-side UAV from the detection point O. This represents the position vector of the UAV D from the detection point.
[0120] Using the above azimuth formula, an initial hit rate function is constructed, which can be expressed by the following formula (6).
[0121]
[0122] in, η represents the initial hit rate. k Indicates the initial azimuth angle. p0 represents the hit probability of the i-th target located at point α1 attacking the j-th target located at point α2, where p0 indicates that η is greater than η. k The maximum hit rate that can be achieved at that time.
[0123] S32. Calculate the distance between the drones of the two combat sides based on the drone's starting position and the number of drones;
[0124] Wherein, the starting position of W's drone can be represented by cW = (x W ,y W ,z W The starting position of the drone of party D can be expressed as cD = (x D ,y D ,z D )express.
[0125] In one feasible implementation, the distance between the two combat drones can be obtained by the following formula (7):
[0126]
[0127] in, Let i represent the distance between the drones of the two combatants, i represent the i-th weapon of W (i can be 1,...,m), m represent the number of drones of W, and j represent the j-th weapon of D (j can be 1,...,n), n represent the number of drones of D.
[0128] S33. Combine the initial hit rate function with the distance between the drones of both sides to construct the final hit rate function.
[0129] In one feasible implementation, the final hit rate function can be expressed by the following formula (8):
[0130]
[0131] in, Indicates the distance between the drones of the two combatants. This represents the hit probability of the i-th target located at point α1 attacking the j-th target located at point α2, based on the azimuth angle η. ijmax This indicates the maximum distance between the drones of the two combat parties.
[0132] S34. Input the azimuth angle and the distance between the two combat drones from their attitude information into the final hit rate function to obtain the hit rate of both combat drones.
[0133] S4. Establish a static game model based on the threat assessment sequence of both sides and the hit rate of both sides.
[0134] The three elements for constructing a game theory model include: game participants, game strategies, and game payoff matrix.
[0135] Optionally, the specific operation of S4 may include the following steps S41-S43:
[0136] S41. Treat the drones of both sides in the combat as participants in the game;
[0137] Where N = {W, D} can represent the players in the game, W represents our side, and D represents the enemy side;
[0138] S42. The weapon target allocation scheme of the drones of both sides in the combat is used as a game strategy;
[0139] Where S = S W ×S D S can represent the policy action space. W and S D Let S represent the pure strategy combinations of W and D respectively. W The number of pure strategies is τ W ,S D The number of pure strategies is τ DThe number of all strategies can be expressed as |S| = τ W ×τ D .
[0140] S43. Calculate the attack value of both sides using the threat assessment sequence and the hit rate of both sides, and use the matrix constructed from the attack value as the game payoff matrix.
[0141] The payoff matrix for player W can be expressed as follows: The payoff matrix of player D's game strategy can be expressed as follows:
[0142] In one feasible implementation, the attack effectiveness of W and D in phase t can be obtained by the following formulas (9) and (10):
[0143]
[0144]
[0145] Where, x Wij ∈x W i This represents the allocation relationship between objective i of party W and objective j of party D, where w j x represents the threat value of the target on side W. Dji This represents the allocation relationship between target j of party D and target i of party W, where w i p represents the threat value of target D. Wij (t) represents W's hit rate in phase t, p Dji (t) represents the hit rate of D in stage t.
[0146] For a fixed phase, considering the many-to-many confrontation between UAVs, the attack effectiveness of both sides is normalized and expressed by the following formulas (11) and (12):
[0147]
[0148] in, This indicates W's weapon target allocation plan. This indicates the overall effectiveness of W's attack. j p represents the threat level of drones belonging to party D, N represents the number of drones belonging to party D, and p represents the threat level of drones belonging to party D. Wij nf represents the hit rate of drone i attacking drone j from side W. W (x W i ) indicates that W's total attack is invalid, N represents the number of D's drones at the current stage, and n is not, meaning invalid.
[0149]
[0150] in, This indicates the weapon target allocation scheme of Party D, f D (x D j ) represents the overall attack effectiveness of D side, w i p represents the threat level of W's drones. Dji This indicates the hit rate of drone j from party D attacking drone i from party W. This indicates that D's overall attack is invalid, M represents the number of drones W currently has, and n is "not" meaning invalid.
[0151] Based on the above solution to the effectiveness of the attack value of both sides in combat, in the strategy combination The following formulas (13) and (14) represent the payoff functions of both sides in the conflict, obtained by using the evidence theory fusion operator:
[0152]
[0153]
[0154] in, This represents W's profit. This represents the profit of party D.
[0155] The static game model can be constructed by following the above steps, which can be represented as G = (N, S, P). W ,P D )express.
[0156] S5. Optimize the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model.
[0157] The purpose of using the rolling time-domain control method is to promptly handle dynamic events that occur during the multi-stage dynamic weapon target allocation process, thereby improving the real-time performance of decision-making.
[0158] Optionally, the specific operation of S5 may include the following steps S51-S53:
[0159] S51. Preset the same time interval; use the time interval to divide the entire time phase of the attack and defense confrontation process to obtain the threat assessment time node;
[0160] The threat assessment time points can be obtained using {T0,…,T e}express.
[0161] S52. At each threat assessment time point, clear up the drone strike situation of both sides in the previous round, and update the number of drones, drone threat assessment sequence and hit rate of both sides in the new round.
[0162] In this context, a round refers to the period between time nodes T0 and T1 during which both sides engage in firepower strikes until one side runs out of weapons and all targets are destroyed, ending the combat. The number of rounds can be represented by t.
[0163] S53. Based on the number of drones of both sides in the new round, the drone threat assessment sequence, and the hit rate, the static game model is optimized to obtain a multi-stage dynamic game model.
[0164] In one feasible implementation, the number of drones for both W and D is updated at each threat assessment time point, including the number of drones from the previous round and the new round; the opponent's situational information is reacquired and a threat assessment is performed to obtain the threat assessment sequence for both sides; the drone hit rate for both sides is calculated; the above parameters are used to optimize the static game model, resulting in a multi-stage dynamic game model that can be represented as follows: S6. Solve for the Nash equilibrium in the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides. The optimal weapon target allocation strategy combination for both sides includes: the optimal weapon target allocation strategy of each side.
[0165] Optionally, S6 solves the Nash equilibrium for the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination, including:
[0166] The min-max theorem is used to determine whether the game model has a pure-strategy Nash equilibrium. If it does, the solution is obtained directly. If not, the mixed-strategy Nash equilibrium is solved using the payoff equivalence method; or a game learning algorithm is used to solve the Nash equilibrium of the game model to obtain the optimal weapon target allocation strategy combination for both sides.
[0167] Among them, the min-max theorem solution method and the payoff equivalence method are suitable for fast and accurate solutions to small-scale UAV attack and defense confrontation problems, while the game learning algorithm is suitable for solving large-scale complex game scenarios with uncertainties and dynamic factors.
[0168] Optionally, the Nash equilibrium of the multi-stage dynamic game model can be solved using the minimax theorem. The specific steps are as follows:
[0169] (1) Obtain the pure strategy Nash equilibrium solution for both sides using the minimax theorem:
[0170] Based on the fundamental characteristics of game theory, the strategy combination for objective allocation in dynamic game models has a Nash equilibrium solution. Nash equilibrium can be defined as follows: for an n-player game G = {τ, S, U}, the strategy combination s... * ={s1 * ,s2 * ,...,sn *} is a Nash equilibrium if and only if Then, we have expression (15):
[0171]
[0172] in, This represents the optimal strategy.
[0173] In the strategy combination of the above formula (15), when one side changes its strategy and the other side keeps its optimal strategy unchanged, the benefit of changing the strategy is less than that of the optimal strategy.
[0174] In one feasible implementation, the Nash equilibrium is solved using the minimax theorem. For a finite two-player zero-sum game, the strategy chosen by player W can be expressed by the following formula (16):
[0175]
[0176] The D-side selection strategy can be expressed by the following formula (17):
[0177]
[0178] Among them, S W Indicates the strategy combination of W side. S represents the i-th strategy in the strategy combination of W; D This represents the strategy combination of D. Let j represent the j-th strategy in the strategy combination of D.
[0179] (2) When the strategic gains of both sides are equal, the strategic combination of both sides is taken as the optimal weapon target allocation strategy combination.
[0180] In one feasible implementation, the combination of strategies of the two sides is the optimal strategy combination for objective allocation in the dynamic game model if and only if the minimum and maximum strategic payoffs of the two sides are equal, and can be expressed by the following formula (18):
[0181]
[0182] The pure strategy Nash equilibrium solution of the dynamic game model is obtained through the above formula.
[0183] (3) If the model does not have a pure policy Nash equilibrium, use the payoff equivalence method to obtain the mixed policy Nash equilibrium solution of the two sides.
[0184] In one feasible implementation, the hybrid policy represents the probability distribution of the pure policy, and the probability distributions of the W-side and D-side policies are respectively represented as follows: and The expected payoffs of W and D using pure strategies ki and kj are expressed as follows:
[0185]
[0186] Among them, Ez ki This represents the expected payoff for player W using pure strategy ki. This represents the probability that player W chooses pure strategy ki. and This indicates that W is implementing the strategy. D's execution strategy The profits of W and D.
[0187]
[0188] Among them, Ez kj This represents the expected payoff for player D using the pure strategy kj. This represents the probability that player D will adopt pure strategy kj. and This indicates that W is implementing the strategy. D's execution strategy The profits of W and D.
[0189] (4) Both sides choose pure strategies with a certain probability distribution. The expected payoff of any pure strategy is the same, and the sum of the probability distributions is 1.
[0190] Among them, both sides in the battle obtained τ W and τ D Solving a system of equations consisting of several equations is used to find the Nash equilibrium of the mixed strategy.
[0191] Optionally, a game learning algorithm can be used to solve for the target allocation strategy combination in the optimized dynamic game model. The specific steps are as follows:
[0192] (1) Initialize the mixed strategies of the two combatants, i.e. the probability distribution of the strategy set;
[0193] In one feasible implementation, taking W as an example, the initialization of the W-squared hybrid strategy is expressed as follows: Similarly, the initialization of the D-squared hybrid strategy is expressed as follows:
[0194] (2) The expected gains of both sides in the current mixed strategy can be calculated using the above formulas (19) and (20);
[0195] (3) Both sides update their hybrid strategy according to the expected payoff based on a specific update rule, which is specified by the specific game learning algorithm.
[0196] Among them, the game learning algorithm can be selected from Q-learning algorithm, polynomial weighting algorithm and gradient ascent algorithm. Taking Q-learning algorithm as an example to introduce the game process, taking W as an example, the update of Q value and mixed strategy can be expressed by the following formula (21) and formula (22):
[0197]
[0198] Among them, Q ki (n) represents the value brought by W's strategy ki in stage n, Ez ki Q represents the expected return of W by adopting strategy ki. ki' (n) represents the value brought by W's strategy ki' in stage n, Q ki (n+1) represents the value brought by taking strategy ki in stage n+1.
[0199]
[0200] Where α represents the step size parameter, γ represents the discount factor, τ represents the temperature coefficient, and n represents the iteration stage. The mathematical constant e is approximately 2.718282, and the exponential part is the product of the temperature coefficient τ and the Q value corresponding to the strategy ki adopted in stage n+1. S represents W The sum of the Q-value and temperature coefficient corresponding to all strategies.
[0201] 4) The algorithm converges to Nash equilibrium, obtaining a stable probability distribution. and Select the optimal weapon target assignment strategy combination For Q-learning, choosing suitable parameters can lead to convergence to Nash equilibrium.
[0202] S7. Execute the optimal weapon target allocation strategy for each side in the combat.
[0203] This invention first employs a reconnaissance and early warning system to monitor flying targets and obtain target information from both sides in combat. Threat assessments are then performed on this target information to obtain a threat assessment sequence for both sides. Based on the number of UAVs, their attitude information, and their departure positions, the hit rate for both sides is calculated. A static-dynamic game model is established using the hit rate and threat assessment sequence. The static game model is then optimized using a rolling time-domain control method to obtain a multi-stage dynamic game model. Finally, the Nash equilibrium is solved in the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides, and the optimal weapon target allocation strategies for each side are executed. This invention addresses the weapon target allocation problem in UAV offensive and defensive confrontation by proposing a method using the min-max theorem or game learning algorithm to solve the Nash equilibrium, thereby obtaining the optimal weapon target allocation strategy combination for both sides. Simultaneously, the rolling time-domain control method optimizes the dynamic game model for weapon target allocation in UAV offensive and defensive confrontation, improving the real-time decision-making, the survivability of the defense system, and its strike capability.
[0204] Figure 3 This is a block diagram of a target allocation device for a drone swarm attack and defense game, according to an exemplary embodiment. The device is used in a drone swarm attack and defense game target allocation method. (Refer to...) Figure 3 The device includes an acquisition unit 310, an evaluation unit 320, a calculation unit 330, a construction unit 340, an optimization unit 350, a solution unit 360, and an execution unit 370, wherein:
[0205] Acquisition unit 310 is used to acquire target information of both sides in the combat; the target information of both sides in the combat includes: UAV attitude information, number of UAVs, UAV launch position, UAV situation information and UAV speed;
[0206] The assessment unit 320 is used to assess the credibility of target information from both sides in the combat and obtain a threat assessment sequence for both sides.
[0207] The calculation unit 330 is used to calculate the hit rate of both sides in combat based on the number of drones, drone attitude information, drone launch position and drone speed.
[0208] Building unit 340 is used to establish a static game model based on the threat assessment sequence of both sides and the hit rate of both sides.
[0209] The optimization unit 350 is used to optimize the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model.
[0210] Solver 360 is used to solve the Nash equilibrium of a multi-stage dynamic game model and obtain the optimal weapon target allocation strategy combination for both sides in the war.
[0211] Execution unit 370 is used to execute the optimal weapon target allocation strategies of both sides in the combat.
[0212] Optionally, the evaluation unit 320 is used for,
[0213] Six threat indicators were selected: superior-designated, relative angle, relative speed, relative altitude, key protected areas, and radar cross-section. Threat indicator models were constructed corresponding to these six threat indicators.
[0214] Input the UAV situational information into the threat indicator model corresponding to the six threat indicators to obtain six threat level values;
[0215] The first weight of the six threat levels is calculated using the analytic hierarchy process (AHP), and the second weight is calculated using the entropy method. The first and second weights are then fused to obtain the threat assessment sequence for both sides in the conflict.
[0216] Optionally, the computing unit 330 is used for:
[0217] Construct an initial hit rate function based on the azimuth angle in the UAV attitude information;
[0218] Calculate the distance between the drones of the two sides based on their starting positions and the number of drones.
[0219] The initial hit rate function is combined with the distance between the drones of both sides to construct the final hit rate function.
[0220] The azimuth angle and the distance between the two combat drones are input into the final hit rate function to obtain the hit rate of both combat drones.
[0221] Optionally, building unit 340 is used for:
[0222] The drones of both sides in the battle are considered as participants in the game;
[0223] The weapon target allocation scheme of the drones of both sides in the combat is used as a game strategy;
[0224] The attack value of both sides is calculated by using the threat assessment sequence of both sides and the hit rate of both sides, and the matrix constructed by the attack value is used as the game payoff matrix.
[0225] Optionally, the optimization unit 350 is used for:
[0226] Preset the same time interval; use the time interval to divide the entire time phase of the attack and defense confrontation process to obtain the time node for threat assessment;
[0227] At each threat assessment time point, the drone strike situation of both sides in the previous round is cleared, and the number of drones, drone threat assessment sequence and hit rate of both sides in the new round are updated.
[0228] The static game model is optimized based on the number of drones of both sides in the new round, the drone threat assessment sequence, and the hit rate to obtain a multi-stage dynamic game model.
[0229] Optionally, the solver element 360 is used for:
[0230] The min-max theorem is used to determine whether the game model has a pure-strategy Nash equilibrium. If it does, the solution is obtained directly. If not, the mixed-strategy Nash equilibrium is solved using the payoff equivalence method; or a game learning algorithm is used to solve the Nash equilibrium of the game model to obtain the optimal weapon target allocation strategy combination for both sides.
[0231] Optionally, the minimax theorem method can be used to solve for the combination of objective allocation strategies in a multi-stage dynamic game model, including:
[0232] (1) Use the minimax theorem to obtain the pure strategy Nash equilibrium solution for both sides in the battle;
[0233] (2) When the strategic gains of both sides are equal, the combination of strategies of both sides is taken as the optimal combination of weapon target allocation strategies.
[0234] (3) If the model does not have a pure policy Nash equilibrium, use the payoff equivalence method to obtain the mixed policy Nash equilibrium solution of the two sides in the battle;
[0235] (4) Both sides choose pure strategies with a certain mixed strategy probability distribution. The expected payoff of any pure strategy is the same, and the sum of the mixed probability distributions is 1.
[0236] Optionally, a game learning algorithm is used to solve for the combination of objective allocation strategies in a multi-stage dynamic game model, including:
[0237] (1) Initialize the mixed strategies of the two combatants, i.e. the probability distribution of the strategy set;
[0238] (2) Calculate the expected gains of both sides under the current hybrid strategy;
[0239] (3) Both sides update their hybrid strategy according to the expected payoff based on a specific update rule, which is specified by the game learning algorithm.
[0240] (4) The algorithm converges to Nash equilibrium and obtains a stable policy probability distribution.
[0241] This invention first employs a reconnaissance and early warning system to monitor flying targets and obtain target information from both sides in combat. Threat assessments are then performed on this target information to obtain threat assessment sequences for both sides. Based on the number of UAVs, their attitude information, and their departure positions, the hit rate for both sides is calculated. A static game model is established using the hit rates and threat assessment sequences. This static game model is then optimized using a rolling time-domain control method to obtain a multi-stage dynamic game model. Finally, the Nash equilibrium is solved in the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides, and each side's optimal weapon target allocation strategy is executed. This invention addresses the weapon target allocation problem in UAV offensive and defensive confrontation by proposing a method using the minimax theorem or game learning algorithms to solve the Nash equilibrium, thereby obtaining the optimal weapon target allocation strategy combination for both sides. Simultaneously, the rolling time-domain control method optimizes the dynamic game model for weapon target allocation in UAV offensive and defensive confrontation, improving the real-time decision-making, the survivability of the defense system, and its strike capability.
[0242] Figure 4 This is a schematic diagram of the structure of an electronic device 400 provided in an embodiment of the present invention. The electronic device 400 may vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 401 and one or more memories 402. The memory 402 stores at least one instruction, which is loaded and executed by the processor 401 to implement the steps of the above-mentioned Chinese text spelling check method.
[0243] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including instructions that can be executed by a processor in a terminal to complete the aforementioned Chinese text spelling check method. For example, the computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device.
[0244] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.
[0245] 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 principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for target allocation in a game of offensive and defensive combat involving unmanned aerial vehicle (UAV) swarms, characterized in that: The method includes: S1. Obtain target information from both sides in the combat; the target information from both sides includes: UAV attitude information, number of UAVs, UAV launch position, and UAV situational information; S2. Perform a threat assessment on the UAV situational information to obtain a threat assessment sequence for both sides in the combat. Specifically, S2 performs a threat assessment on the target information of both combatants to obtain a threat assessment sequence for both combatants, including: S21. Select six threat indicators: superior-designated, relative angle, relative speed, relative altitude, key protected areas, and radar cross-section, and construct threat indicator models corresponding to the six threat indicators. S22. Input the UAV situational information into the threat indicator model corresponding to the six threat indicators respectively to obtain six threat level values; S23. Calculate the first weight of the six threat levels using the analytic hierarchy process (AHP), and calculate the second weight using the entropy method; combine the first weight and the second weight to obtain the threat level assessment sequence for both sides in the conflict. S3. Calculate the hit rate of both sides based on the number of drones, drone attitude information, and drone departure position; Specifically, S3 calculates the hit rate of both sides based on the number of drones, drone attitude information, and drone launch positions, including: S31. Construct an initial hit rate function based on the azimuth angle in the UAV attitude information; S32. Based on the departure position and number of drones, calculate the number of drones between the two combat sides. The distance between; S33. Combine the initial hit rate function with the distance between the drones of the two combat parties to construct the final hit rate function; S34. Input the azimuth angle and the distance between the two combat drones from their attitude information into the final hit rate function to obtain the hit rate of both combat drones. S4. Based on the threat assessment sequence of the two combatants and the hit rate of the two combatants, establish a static game model; S5. Optimize the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model. S6. Solve the Nash equilibrium for the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides in the war; the optimal weapon target allocation strategy combination for both sides in the war includes: the optimal weapon target allocation strategy of each side in the war. S7. Execute the optimal weapon target allocation strategy for each side in the combat.
2. The method according to claim 1, characterized in that, The S4 method establishes a static game model based on the threat assessment sequences of the two combatants and the hit rates of the two combatants, including: S41. Treat the drones of both sides in the combat as participants in the game; S42. The weapon target allocation scheme of the drones of both sides in the combat is used as a game strategy; S43. Calculate the attack value of both sides using the threat assessment sequence of the two sides and the hit rate of the two sides, and use the matrix constructed from the attack value as the game payoff matrix.
3. The method according to claim 1, characterized in that, The optimization of the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model includes: S51. Preset the same time interval; use the time interval to divide the entire time phase of the attack and defense confrontation process to obtain the threat assessment time node; S52. At each threat assessment time point, clear up the drone strike situation of both sides in the previous round, and update the calculation of the number of drones of both sides in the previous round, the drone threat assessment sequence and the hit rate. S53. Based on the number of drones of both sides in the new round, the drone threat assessment sequence, and the hit rate, the static game model is optimized to obtain a multi-stage dynamic game model.
4. The method according to claim 1, characterized in that, The S6 method solves for the Nash equilibrium in the multi-stage dynamic game model to obtain the optimal weapon target allocation strategy combination for both sides in the combat, including: The min-max theorem is used to determine whether the game model has a pure-strategy Nash equilibrium. If a pure-strategy Nash equilibrium exists, it is solved directly. Otherwise, the payoff equivalence method is used to solve for the mixed-strategy Nash equilibrium, or the game learning algorithm is used to solve for the Nash equilibrium of the game model to obtain the optimal weapon target allocation strategy combination for both sides.
5. The method according to claim 4, characterized in that, The method of solving for the Nash equilibrium of a multi-stage dynamic game model using the minimax theorem includes: S71. Use the minimax theorem to obtain the pure strategy Nash equilibrium solution for both sides in the battle; S72. When the strategic gains of both sides are equal, the strategic combination of both sides shall be taken as the optimal weapon target allocation strategy combination. S73. If the model does not have a pure policy Nash equilibrium, use the payoff equivalence method to obtain the mixed policy Nash equilibrium solution for both sides in the battle. S74. Both sides in the battle choose pure strategies according to a certain mixed strategy probability distribution. The expected payoff of any pure strategy is the same, and the sum of the mixed probability distributions is 1.
6. The method according to claim 4, characterized in that, The method of using game learning algorithms to solve for the Nash equilibrium of a multi-stage dynamic game model includes: S75. Initialize the mixed strategies of the two combatants, i.e., the probability distribution of the strategy set; S76. Calculate the expected gains of both sides in the current hybrid strategy; S77. Both sides in the combat use expected gains to update their hybrid strategies according to specific update rules; S78. The algorithm converges to Nash equilibrium, obtaining a stable policy probability distribution.
7. A target allocation device for a drone swarm offensive and defensive game, characterized in that, The device includes: The acquisition unit is used to acquire target information from both sides in the combat. The target information from both sides includes: UAV attitude information, number of UAVs, UAV launch position, UAV situational information, and UAV speed. The assessment unit is used to assess the credibility of target information from both sides in the combat and obtain a threat assessment sequence for both sides. The evaluation unit is used for, Six threat indicators were selected: superior-designated, relative angle, relative speed, relative altitude, key protected areas, and radar cross-section. Threat indicator models were constructed corresponding to these six threat indicators. Input the UAV situational information into the threat indicator model corresponding to the six threat indicators to obtain six threat level values; The first weight of the six threat levels is calculated using the analytic hierarchy process (AHP), and the second weight is calculated using the entropy method. The first and second weights are then fused to obtain the threat assessment sequence for both sides in the conflict. The computing unit is used to calculate the hit rate of both sides in combat based on the number of drones, drone attitude information, drone launch position, and drone speed. The computing unit is used for: Based on the azimuth angle in the UAV attitude information, an initial hit rate function is constructed; Based on the drone's departure location and number, calculate the drone count between the two combat sides. The distance between; The initial hit rate function is combined with the distance between the drones of the two combat parties to construct the final hit rate function; The azimuth angle and the distance between the two combat drones are input into the final hit rate function to obtain the hit rate of the two combat parties. The building unit is used to establish a static game model based on the threat assessment sequence and hit rate of both sides in the combat. The optimization unit is used to optimize the static game model using the rolling time-domain control method to obtain a multi-stage dynamic game model. The solution unit is used to solve the Nash equilibrium in the multi-stage dynamic game model and obtain the optimal weapon target allocation strategy combination for both sides in the battle. The execution unit is used to execute the optimal weapon target allocation strategies of both sides in the combat.
Citation Information
Patent Citations
Multi-aircraft cooperative capture space division method and device, and electronic equipment
CN114415723A
Unmanned aerial vehicle group collaborative autonomous decision-making method and device based on multi-body game
CN114460959A
