Unmanned cluster cooperative combat digital twin deduction optimization method under complex weather conditions

Abstract

The present application relates to meteorological services, quality, safety and environmental testing services and digital twin and combat deduction technical field, specifically related to a kind of complex weather conditions under the cooperative combat digital twin deduction optimization method of unmanned cluster, the method includes: S1, complex weather conditions under the cooperative combat digital twin model of unmanned cluster is built;S2, cross-coupling term is introduced in Lorenz system, and nonlinear weather condition model is built;S3, dynamic attenuation factor is designed, and nonlinear attenuation effect is quantified, and cluster cooperative ability model is built;S4, particle swarm optimization algorithm is improved, fitness function is designed, and high matching degree task allocation strategy is obtained;S5, weather-capability-task-strategy multidimensional coupling dynamic Bayesian network is built, and task allocation strategy is accurately deduced;S6, according to deduction result, multidimensional evaluation system including task efficiency, resource consumption and system robustness is built, and task allocation strategy is iteratively optimized.

Description

Technical Field

[0001] This invention relates to the fields of meteorological services, quality, safety and environmental testing and inspection services, as well as digital twin and combat simulation technologies. Specifically, it relates to an optimization method for digital twin simulation of unmanned swarm collaborative combat under complex weather conditions. Background Technology

[0002] Digital twin technology virtually maps real-time physical entities, relies on sensors to achieve data synchronization, and constructs a virtual-real synchronized model to assist in decision optimization. In military simulations, this technology can achieve in-depth combat simulation through the dynamic mapping of meteorological parameters and equipment performance.

[0003] Traditional military simulations rely on historical data and empirical rules, which have three main limitations: First, in terms of environmental factors, they ignore the multidimensional impact of complex weather conditions on equipment effectiveness; second, in model construction, it is difficult to analyze the relationship between cluster characteristics and mission requirements, leading to suboptimal mission allocation; and third, in terms of accuracy, the use of discrete state models ignores dynamic environmental changes and equipment performance degradation, making it difficult for simulation results to truly reflect the continuous evolution of environmental parameters and the nonlinear degradation dynamic characteristics of equipment performance. Summary of the Invention

[0004] The present invention provides a digital twin simulation optimization method for unmanned swarm collaborative operations under complex weather conditions to improve the robustness of task allocation and the accuracy of simulation decision-making, which can at least solve one of the above-mentioned technical problems.

[0005] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0006] A digital twin simulation optimization method for unmanned swarm collaborative operations under complex weather conditions includes the following steps:

[0007] S1. Collect meteorological parameters from multiple sources and performance parameters of unmanned swarms to construct a digital twin model of unmanned swarm collaborative operations under complex meteorological conditions.

[0008] S2. Introduce cross-coupling terms into the Lorenz system and use the least squares method to dynamically optimize the parameters obtained from S1 to construct a nonlinear meteorological condition model.

[0009] S3. Based on the nonlinear meteorological condition model in S2, design a dynamic attenuation factor to quantify the nonlinear attenuation effect of meteorological conditions on the performance parameters of intelligent agents in the unmanned swarm. Combine geometric mean, variance constraint and cooperative gain mechanism to construct a swarm cooperative capability model.

[0010] S4. Improve the particle swarm optimization algorithm by designing a fitness function based on the cluster collaboration capability model in S3 and the actual task requirements, and dynamically adjusting the inertia weight, adaptive learning factor and global guidance factor to obtain a task allocation strategy with high matching degree.

[0011] S5. Construct a multi-dimensional coupled dynamic Bayesian network of meteorology, capability, task, and strategy, and combine forward prediction, posterior update, and online learning mechanisms to achieve accurate deduction of the task allocation strategy in S4.

[0012] S6. Based on the deduction results of S5, construct a multi-dimensional evaluation system including task efficiency, resource consumption and system robustness, and continuously improve the task allocation strategy in S4 through iterative optimization.

[0013] Furthermore, the multi-source meteorological parameters in S1 include at least temperature, humidity, and visibility, and the performance parameters of the unmanned swarm include at least the perception, communication, and decision-making capabilities data of each agent within the swarm.

[0014] Furthermore, the meteorological state defined in S2 includes:

[0015]

[0016] Where T is temperature, H is humidity, and V is visibility;

[0017] The nonlinear differential equation structure of the Lorenz system reflects the nonlinear dynamic characteristics of real meteorological systems. Cross-coupling terms are introduced into this nonlinear differential equation to extend the nonlinear interactions between different meteorological conditions, resulting in the following nonlinear meteorological condition model:

[0018]

[0019] in:

[0020] , and These are all parameters that are originally present in the Lorenz system, namely Prandtl's constant, Rayleigh's constant, and a space-dependent constant;

[0021] , and These are the coupling coefficients under different meteorological conditions, reflecting the nonlinear interactions between meteorological elements.

[0022] Furthermore, in S3, the three capabilities possessed by the k-th unmanned cluster are defined as follows: decision-making capability. , communication ability and perception ability The unmanned swarm has There are three intelligent agents, each possessing the three capabilities mentioned above, defined as follows: , and ;

[0023] Introducing dynamic decay factor , and The attenuation of sensing ability, communication ability, and decision-making ability under the influence of meteorological conditions is described, and the expressions for the dynamic attenuation factors of sensing ability, communication ability, and decision-making ability are obtained respectively:

[0024]

[0025]

[0026]

[0027] in:

[0028] This represents the temperature sensitivity coefficient, which is determined by the material properties of the optical device.

[0029] This indicates the absolute deviation between the current temperature and the sensor's calibrated temperature.

[0030] The minimum visibility threshold is determined by the specific optical system.

[0031] and These are the specific decay coefficient and the power-law exponent, respectively;

[0032] This represents the critical room temperature threshold that triggers dynamic frequency reduction.

[0033] Indicates the highest room temperature threshold allowed by the processor;

[0034] This represents the processor performance degradation factor;

[0035] This is the minimum performance guarantee threshold;

[0036] The better the heat dissipation design, the smoother the performance degradation curve will be, thus controlling the steepness of the curve.

[0037] Furthermore, in S3, the communication capabilities of the unmanned swarm are discussed. Perception ability and decision-making ability The modeling is as follows:

[0038] ①.Communication ability Modeling:

[0039] While introducing a geometric mean mechanism to reflect the limitation of communication link performance by the weakest node, this paper also considers the gains in communication capability brought by cluster cooperation, and based on this, gives... The expression:

[0040]

[0041] in:

[0042] This represents the geometric mean of the communication capabilities of each agent within the cluster, where... Represents the number of agents within the cluster;

[0043] It is a factor affecting communication stability, among which, It is the variance of the communication capabilities of all intelligent agents; the larger the variance, the worse the communication stability.

[0044] It is the degree of influence of the experimentally determined variance on the collaborative ability;

[0045] This represents the gain in communication capabilities that cluster collaboration brings, where... The communication capability cooperation coefficient is determined by the cluster hardware. For the number of times of communication and collaboration, For the number of nodes, Expanding the network will increase the complexity of the unmanned swarm communication network, causing the communication capacity to decrease logarithmically.

[0046] ② Perception ability Modeling:

[0047] In addition to considering the gain of sensing capability from cluster collaboration, the gain of sensing capability from cluster communication capability is also introduced, and based on this, the following is given: The expression:

[0048]

[0049] in:

[0050] It is an intelligent agent Neighbor intelligent agents in The sum of real-time communication capabilities;

[0051] It is all intelligent agents in The sum of communication capabilities at any given moment;

[0052] This is the gain coefficient of the cluster communication capability on the sensing capability, as determined in the experiment.

[0053] The overall representation indicates the gain of the communication network on sensing capabilities;

[0054] This represents the gain in perception capability brought by cluster collaboration, where... This is the sensing capability collaboration coefficient, and its value is determined by the characteristics of the cluster hardware. For the number of times communication and collaboration were conducted;

[0055] ③ Decision-making ability Modeling:

[0056] In addition to using the geometric mean mechanism to balance the decision weights of each agent, a fault-tolerant constraint mechanism needs to be constructed to prevent global decision failure due to local communication failures, while also taking into account the gains in decision-making capabilities brought by cluster collaboration. Based on this, the following is proposed: The expression:

[0057]

[0058] in:

[0059] The geometric mean of the decision-making capabilities of each agent within the cluster;

[0060] It is the minimum value of the product of the communication capabilities of all intelligent agents, reflecting that the decision-making process is limited by the weakest communication link;

[0061] These are experimentally determined coefficients used to adjust the degree of influence of communication capabilities on collaborative decision-making.

[0062] The benefits that cluster collaboration brings to decision-making capabilities, among which... This is the decision-making capability collaboration coefficient, and its value is determined by the characteristics of the cluster hardware. For the number of times communication and collaboration were conducted;

[0063] To reduce the time spent on excessive group negotiation, a hyperbolic tangent function is introduced to simulate the diminishing marginal effect. The nonlinear characteristics of the saturation region of the hyperbolic tangent function can accurately describe the relationship between the number of negotiations and the efficiency of decision-making.

[0064] Furthermore, in S4, an unmanned cluster set is set. Combat mission Divided into several sub-tasks Particle swarm optimization algorithm is used for different clusters Assign combat missions:

[0065] S4.1 Encode the particles using a three-dimensional array. To define the position of the particle, where, Indices representing particle indices Indicates the index of the task. This represents the index of an unmanned cluster, and At the same time satisfy For all and All are true;

[0066] S4.2 Construct the fitness function of the particle swarm optimization algorithm. This function guides the particle swarm optimization algorithm to search towards the optimal solution. Let... The set consisting of weights representing the different cluster capability requirements of different tasks is defined as follows:

[0067]

[0068] In this set: Indicates task At any moment The required weight of unmanned swarm perception capabilities Indicates task At any moment The weight of the demand for unmanned cluster communication capabilities Indicates task At any moment The required weight of unmanned swarm decision-making capabilities Traverse the task set All tasks in;

[0069] Based on this, the fitness function is given as follows:

[0070]

[0071] in:

[0072] The core of the task requirement-capability matching item is to sort the capability requirements by weight from high to low, and prioritize matching high-weight items with clusters that have better corresponding capabilities.

[0073] Here is the time decay term, where, It is the time decay coefficient, used to adjust the contribution of task completion speed to the fitness function, and its value depends on the actual problem scenario;

[0074] S4.3 Construct the velocity update equation for the particle swarm optimization algorithm. This equation should consider at least the following parameters: inertia weight. Individual learning factors Social learning factors and global learning guidance factors ;

[0075] ① Inertia weight For the exploration and development of balancing algorithms, the initial value of the inertia weight is set to... The inertial weight can be obtained. The expression is:

[0076]

[0077] in, These represent the influence coefficients of temperature, humidity, and visibility on the inertial weight, respectively.

[0078] ② Individual learning factors Used to represent particles The ability to learn from its own historical best position, with the initial value of the individual learning factor set at [value]. Individual learning factors can be obtained. The expression is:

[0079]

[0080] The calculation of the matching degree in this formula further includes the following steps:

[0081] S4.3.1 Normalize the capabilities of each cluster to make all capabilities comparable on the same scale. , , It is the maximum capacity obtained by traversing each cluster;

[0082] S4.3.2. Improve the sensitivity of cluster capabilities to demand weights through exponential functions, thereby improving the sensitivity of individual learning factors to changes.

[0083] S4.3.3, after summation, compare with the total number of tasks. Divide to ensure The adjustment range should remain within a reasonable range depending on the scale of the task;

[0084] S4.3.4, Matching Degree Influence Coefficient Calibrated Through Experiments Controlling the matching degree The adjustment range;

[0085] ③ Social learning factors This represents the ability of a particle to learn towards the global optimal position. Let the initial value of the individual learning factor be... The social learning factor can be obtained. The expression is:

[0086]

[0087] in:

[0088] It is the Sigmoid function, used to map changes in weather conditions to adjustments in social learning factors;

[0089] It is the minimum value of the social learning factor, used to ensure that particles have the ability to learn towards the global optimum under any circumstances;

[0090] It is the maximum value of the social learning factor, used to avoid excessive learning intensity that could lead to instability in the particle swarm optimization algorithm;

[0091] The influence coefficient of meteorological condition changes, as determined experimentally, is used to control the impact of meteorological condition changes on... The impact;

[0092] ④ Global guiding factor To respond to sudden environmental changes and changes in mission capability requirements, the initial value of the global guiding factor is set to... The global learning guidance factor can be obtained. The expression is:

[0093]

[0094] in, It is the adjustment coefficient measured experimentally, representing the degree of influence of the gradient term on the global guiding factor;

[0095] By solving the gradient of each condition Norms can characterize the drastic changes of various conditions over time. Under sudden conditions, an increase in the gradient value will affect the global guiding factor. This increases the value of the particle swarm optimization algorithm, thereby guiding it to focus more on the global optimum and preventing it from getting trapped in local optima.

[0096] S4.4. Combining the above parameters, the velocity update equation for the particle swarm optimization algorithm is obtained:

[0097]

[0098] in, It is a particle The best position it has found in its own history. It is in the interval Uniformly distributed random numbers within the range. It is the optimal position found by the entire particle swarm in history. It is a pre-defined global guidance location, determined based on prior knowledge of the problem or other information. and Each of these is a random number that is uniformly distributed within the interval;

[0099] For particles Perform the following multiple iterations:

[0100]

[0101] Let the particle be at the th The position of the wheel is Calculate the corresponding fitness value ,like Then update ,like If so, no action is taken;

[0102] S4.5 After all particle iterations are complete, compare all... Choose the highest fitness value. As At this time This is the optimal task allocation scheme, which will... Set as .

[0103] Furthermore, in S5, a dynamic Bayesian multidimensional coupled state space is defined, consisting of meteorological state, capability state, task state, and task allocation strategy. :

[0104]

[0105] in, Indicates task status:

[0106]

[0107] Based on this, the various meteorological parameters and cluster capabilities in the state space are discretized respectively;

[0108] set up The evolution coefficient matrix is ​​used for discretization.

[0109]

[0110] against Online updates are used: Continuous updates This allows for real-time reflection of the latest data distribution, improving the adaptability of the entire simulation process to dynamic environments. Online rule updates typically employ gradient descent.

[0111]

[0112] Parameters are dynamically adjusted using the gradient ascent method. To maximize the state transition probability:

[0113]

[0114] in:

[0115] An adaptive learning rate can be expressed as:

[0116]

[0117] When the historical gradient is large, reduce the learning rate. To avoid oscillations, when the historical gradient is small, maintain or increase the learning rate. This accelerates convergence;

[0118] The initial learning rate is set to a fixed value;

[0119] For a historic moment Bayesian gradient;

[0120] This is the average of the historical sum of squared gradients, used to quantify the cumulative fluctuations in parameter updates;

[0121] set up Observed variables in dynamic Bayesian networks:

[0122]

[0123] The observed variables are various data obtained from real-world scenarios, including measurements from weather sensors. The current capabilities obtained from the self-inspection of the unmanned cluster and task execution status. .

[0124] Furthermore, S5 further includes:

[0125] S5.1 Perform forward inference on the dynamic Bayesian network, with the goal of determining the current state. and task allocation scheme Predict the state distribution at the next time step. The specific steps are as follows:

[0126] S5.1.1, Set the state include:

[0127] Weather conditions:

[0128] Ability Status:

[0129] Task Status:

[0130] Each state component is conditionally independent, and the joint probability decomposition is as follows:

[0131]

[0132] Based on this, the transition probabilities of each meteorological state and cluster capability state are calculated respectively;

[0133] S5.1.2, Use the sigmoid function to calculate the task execution status at time t. Within the range, calculate the transition probability of the task state:

[0134] ;

[0135] S5.2. Perform a posteriori update on the dynamic Bayesian network, with the aim of updating the data based on all known historical observations. and the current task allocation scheme Under the condition of system state Optimal estimate Suppose that the transition of the observed variable follows:

[0136] Meteorological observation :

[0137] Self-assessment of capabilities :

[0138] Status Report :

[0139] The joint observational likelihood was obtained:

[0140]

[0141] The formula for posterior update is:

[0142]

[0143] in:

[0144] It can be represented as:

[0145]

[0146] In the above formula, The formula for calculating the posterior probability is:

[0147]

[0148] Let be the normalization constant, expressed as:

[0149] .

[0150] Furthermore, in S6, in After the time-sharing simulation is completed, the task allocation plan is determined. The inference results are evaluated, and an evaluation vector is constructed. :

[0151]

[0152] ① Vector middle, Represents task execution efficiency, expressed as:

[0153]

[0154] in, This is a time decay factor that penalizes delays in task completion time; the longer the delay, the lower the performance contribution. For the task The actual completion time The starting time of the simulation. For the task Expected value of completion The calculation expression is:

[0155]

[0156] Set the task to be executed The performance loss is The expression is:

[0157]

[0158] For tasks with high efficiency loss Reassign clusters based on capability matching. :

[0159]

[0160] In this formula, To the task The personalized learning rate can be expressed as: , The base learning rate needs to be determined through multiple trials based on experience.

[0161] ② Vector middle, Representing resource consumption, the expression is:

[0162]

[0163] in, Representing clusters Resource consumption for performing sensing, communication, and decision-making tasks;

[0164] Set up a cluster The energy consumption is For high-energy-consuming clusters, task load should be reduced and resource usage balanced:

[0165]

[0166] In this formula, This is the energy consumption penalty coefficient, used to control the intensity of the penalty; its value is determined by the actual battlefield environment.

[0167] The exponential term maps energy consumption to interval, when When the value approaches 1, it indicates that the cluster has high energy consumption. Approaching , Significantly reduced;

[0168] ③ Vector middle, Represents system robustness, used to measure the system's ability to maintain task performance in the face of sudden environmental changes and changes in task requirements. Let be the total number of time steps in the dynamic Bayesian extrapolation. The expression is:

[0169]

[0170] By averaging the cumulative changes in environmental mutations and task requirements over the entire time period, the impact of the randomness of a single mutation on the evaluation results can be avoided.

[0171] By normalizing the design, robustness R is mapped to... The larger the value in the range, the more stable the system.

[0172] Set up a cluster robustness is The expression is:

[0173]

[0174] in, In order to be in Time Cluster Task in progress A set;

[0175] For highly robust clusters, tasks should be assigned preferentially:

[0176]

[0177] In this formula, For robust linear enhancement terms, To enhance the coefficient, the value is determined by the actual battlefield environment;

[0178] Cluster The higher the robustness, the larger the linear enhancement term. Significant increase.

[0179] Furthermore, in S6, the combined... and This leads to a multi-objective joint optimization formula:

[0180]

[0181] in, These are the weighting coefficients for each target, and their specific values ​​need to be determined based on the actual mission requirements in the battlefield environment.

[0182] Adjusted Input is fed into a dynamic Bayesian network to obtain a new evaluation vector. :

[0183]

[0184] Let the optimization thresholds for each objective be as follows: If the following convergence condition is met:

[0185]

[0186] Then stop the optimization and obtain the optimal task allocation scheme. ;

[0187] Otherwise, repeat steps S5 and S6 until the above convergence condition is met.

[0188] The beneficial effects of this invention are reflected in:

[0189] 1. Based on the Lorenz system, a nonlinear meteorological evolution model is constructed, which integrates dynamic decay factor and digital twin technology to improve the simulation accuracy of unmanned swarm collaborative operations under complex weather conditions. At the same time, the nonlinear interaction between meteorological variables is strengthened by cross-coupling terms, and the dynamic evolution of temperature / humidity / visibility is accurately modeled, breaking through the limitations of traditional linear models.

[0190] 2. A dynamic attenuation factor is designed to quantify the nonlinear impact of meteorological conditions on cluster capabilities. The geometric mean, variance constraint, and collaborative gain mechanism are integrated, and the hyperbolic tangent function is combined to simulate the meteorological attenuation saturation effect, thus solving the problem that traditional methods ignore dynamic meteorological interference.

[0191] 3. An improved particle swarm optimization algorithm is proposed, which uses dynamic inertia weights and adaptive learning factors to adjust the task allocation strategy in real time by combining meteorological data, thereby generating a highly adaptable solution.

[0192] 4. Design a multi-source coupled dynamic Bayesian network for meteorological environment, cluster capability, combat mission, and mission allocation strategy. Construct a two-way inference closed loop of forward prediction and posterior update, and realize online strategy optimization under sudden environmental changes through an adaptive learning mechanism.

[0193] 5. Establish a joint evaluation system for task effectiveness, resource consumption, and robustness to achieve accurate evaluation of task allocation schemes, and continuously improve task allocation strategies through iterative optimization. Attached Figure Description

[0194] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application.

[0195] Figure 1 This is a schematic diagram of the overall process of the method according to an embodiment of the present invention.

[0196] Figure 2 This is a schematic diagram comparing the improved particle swarm optimization algorithm of this invention with the standard particle swarm optimization algorithm.

[0197] Figure 3 This is a schematic diagram comparing the improved dynamic Bayesian network of this invention with the standard dynamic Bayesian network.

[0198] Figure 4 This is a structural block diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0199] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0200] It should be noted that the meaning of "and / or" throughout the text includes three parallel solutions. Taking "A and / or B" as an example, it includes solution A, solution B, or a solution that simultaneously satisfies A and B. Furthermore, "multiple" refers to two or more. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0201] See Figure 1 This invention provides a digital twin simulation optimization method for unmanned swarm cooperative operations under complex weather conditions, comprising the following steps:

[0202] S1. Collect meteorological parameters from multiple sources and performance parameters of unmanned swarms to construct a digital twin model of unmanned swarm collaborative operations under complex meteorological conditions.

[0203] S2. Introduce cross-coupling terms into the Lorenz system and use the least squares method to dynamically optimize the parameters obtained from S1 to construct a nonlinear meteorological condition model.

[0204] S3. Based on the nonlinear meteorological condition model in S2, design a dynamic attenuation factor to quantify the nonlinear attenuation effect of meteorological conditions on the performance parameters of intelligent agents in the unmanned swarm. Combine geometric mean, variance constraint and cooperative gain mechanism to construct a swarm cooperative capability model.

[0205] S4. Improve the particle swarm optimization algorithm by designing a fitness function based on the cluster collaboration capability model in S3 and the actual task requirements, and dynamically adjusting the inertia weight, adaptive learning factor and global guidance factor to obtain a task allocation strategy with high matching degree.

[0206] S5. Construct a multi-dimensional coupled dynamic Bayesian network of meteorology, capability, task, and strategy, and combine forward prediction, posterior update, and online learning mechanisms to achieve accurate deduction of the task allocation strategy in S4.

[0207] S6. Based on the deduction results of S5, construct a multi-dimensional evaluation system including task efficiency, resource consumption and system robustness, and continuously improve the task allocation strategy in S4 through iterative optimization.

[0208] In this embodiment, the multi-source meteorological parameters in S1 include at least temperature, humidity and visibility, and the performance parameters of the unmanned swarm include at least the perception, communication and decision-making ability data of each intelligent agent in the swarm.

[0209] In this embodiment, the meteorological state defined in S2 includes:

[0210]

[0211] Where T is temperature, H is humidity, and V is visibility;

[0212] A nonlinear meteorological condition model is constructed using the Lorenz system from chaos theory. Compared with traditional or simplified models, the nonlinear differential equation structure of the Lorenz system can effectively reflect the nonlinear dynamic characteristics of real meteorological systems.

[0213] However, the interaction between the three variables of temperature, humidity, and visibility is more complex than that in Lorenz's original model. Therefore, a cross-coupling term was introduced into the equation to extend the nonlinear interaction between different meteorological conditions, resulting in the following nonlinear meteorological condition model:

[0214]

[0215] In this equation, , and These are all parameters that are originally present in the Lorenz system, namely Prandtl's constant, Rayleigh's constant, and a space-dependent constant;

[0216] Now, by employing a least squares optimization strategy, the solution process is restarted based on the system variables, in order to... For example, in humidity When stable, ignoring coupling terms, the above model equations are first simplified as follows:

[0217]

[0218] Next, the equation is discretized and linearly regressed:

[0219]

[0220] Rearranged into linear equations:

[0221]

[0222] Finally, the least squares solution is obtained:

[0223]

[0224] In this equation, , and These are the coupling coefficients under different meteorological conditions, reflecting the nonlinear interactions between meteorological elements. For example, let's first simplify the above model equations:

[0225]

[0226] Next, construct the objective function:

[0227]

[0228] Finally, take the derivative and let... The optimal solution is obtained as follows:

[0229]

[0230] In this embodiment, the three capabilities of the k-th unmanned cluster are defined in S3 as follows: decision-making capability. , communication ability and perception ability The unmanned swarm has There are three intelligent agents, each possessing the three capabilities mentioned above, defined as follows: , and ;

[0231] Introducing dynamic decay factor , and The following describes the attenuation of perception, communication, and decision-making capabilities under the influence of meteorological conditions:

[0232] The main meteorological conditions affecting sensing capabilities are temperature and visibility. Temperature causes the sensitivity of optical devices to decay exponentially, while visibility limits the measurement accuracy of optical devices to a threshold function.

[0233] The main meteorological condition affecting communication capabilities is humidity;

[0234] Temperature is a key meteorological condition that affects decision-making ability. In high-temperature environments, heat is conducted through the device casing to the internal chipset, causing the processor to reduce its frequency, thereby reducing the accuracy of decision-making and prolonging the decision-making time.

[0235] Therefore, the expressions for the dynamic attenuation factors of perception ability, communication ability, and decision-making ability are obtained as follows:

[0236]

[0237]

[0238]

[0239] in:

[0240] This represents the temperature sensitivity coefficient, which is determined by the material properties of the optical device.

[0241] This indicates the absolute deviation between the current temperature and the sensor's calibrated temperature.

[0242] The minimum visibility threshold is determined by the specific optical system.

[0243] and These are the specific attenuation coefficient and the power-law exponent, respectively. The values ​​can be found in the ITU-R recommendations for rainfall attenuation models.

[0244] This represents the critical room temperature threshold that triggers dynamic frequency reduction.

[0245] Indicates the highest room temperature threshold allowed by the processor;

[0246] This represents the processor performance degradation factor;

[0247] This is the minimum performance guarantee threshold;

[0248] The better the heat dissipation design, the smoother the performance degradation curve will be, thus controlling the steepness of the curve.

[0249] The above parameters were measured using different processor models under different room temperature conditions.

[0250] In this embodiment, in step S3, the communication capability of the unmanned cluster is discussed. Perception ability and decision-making ability The modeling is as follows:

[0251] ①.Communication ability Modeling:

[0252] While introducing a geometric mean mechanism to reflect the limitation of communication link performance by the weakest node, this paper also considers the gains in communication capability brought by cluster cooperation, and based on this, gives... The expression:

[0253]

[0254] in:

[0255] This represents the geometric mean of the communication capabilities of each agent within the cluster, where... Represents the number of agents within the cluster;

[0256] It is a factor affecting communication stability, among which, It is the variance of the communication capabilities of all intelligent agents; the larger the variance, the worse the communication stability.

[0257] It is the degree of influence of the experimentally determined variance on the collaborative ability;

[0258] This represents the gain in communication capabilities that cluster collaboration brings, where... The communication capability cooperation coefficient is determined by the cluster hardware. For the number of times of communication and collaboration, For the number of nodes, Expanding the network will increase the complexity of the unmanned swarm communication network, causing the communication capacity to decrease logarithmically.

[0259] ② Perception ability Modeling:

[0260] In addition to considering the gain of sensing capability from cluster collaboration, the gain of sensing capability from cluster communication capability is also introduced, and based on this, the following is given: The expression:

[0261]

[0262] in:

[0263] It is an intelligent agent Neighbor intelligent agents in The sum of real-time communication capabilities;

[0264] It is all intelligent agents in The sum of communication capabilities at any given moment;

[0265] This is the gain coefficient of the cluster communication capability on the sensing capability, as determined in the experiment.

[0266] The overall representation indicates the gain of the communication network on sensing capabilities;

[0267] This represents the gain in perception capability brought by cluster collaboration, where... This is the sensing capability collaboration coefficient, and its value is determined by the characteristics of the cluster hardware. For the number of times communication and collaboration were conducted;

[0268] ③ Decision-making ability Modeling:

[0269] In addition to using the geometric mean mechanism to balance the decision weights of each agent, a fault-tolerant constraint mechanism needs to be constructed to prevent global decision failure due to local communication failures, while also taking into account the gains in decision-making capabilities brought by cluster collaboration. Based on this, the following is proposed: The expression:

[0270]

[0271] in:

[0272] The geometric mean of the decision-making capabilities of each agent within the cluster;

[0273] It is the minimum value of the product of the communication capabilities of all intelligent agents, reflecting that the decision-making process is limited by the weakest communication link;

[0274] These are experimentally determined coefficients used to adjust the degree of influence of communication capabilities on collaborative decision-making.

[0275] The benefits that cluster collaboration brings to decision-making capabilities, among which... This is the decision-making capability collaboration coefficient, and its value is determined by the characteristics of the cluster hardware. For the number of times communication and collaboration were conducted;

[0276] To reduce the time spent on excessive group negotiation, a hyperbolic tangent function is introduced to simulate the diminishing marginal effect. The nonlinear characteristics of the saturation region of the hyperbolic tangent function can accurately describe the relationship between the number of negotiations and the efficiency of decision-making.

[0277] In this embodiment, in step S4, an unmanned cluster set is assumed. Combat mission Divided into several sub-tasks Particle swarm optimization algorithm is used for different clusters Assign combat missions:

[0278] S4.1 Encode the particles using a three-dimensional array. To define the particle's position, a three-dimensional array, compared to a one-dimensional vector representation, can clearly represent the probability of each task being assigned to each unmanned swarm. Indices representing particle indices Indicates the index of the task. This represents the index of an unmanned cluster, and At the same time satisfy For all and All are true;

[0279] S4.2 Construct the fitness function of the particle swarm optimization algorithm. This function guides the particle swarm optimization algorithm to search towards the optimal solution. Let... The set consisting of weights representing the different cluster capability requirements of different tasks is defined as follows:

[0280]

[0281] In this set: Indicates task At any moment The required weight of unmanned swarm perception capabilities Indicates task At any moment The weight of the demand for unmanned cluster communication capabilities Indicates task At any moment The required weight of unmanned swarm decision-making capabilities Traverse the task set All tasks in;

[0282] Compared to the static fitness function of general particle swarm optimization algorithms, this fitness function can dynamically adjust the weights and calculation methods of various factors based on real-time meteorological data and task progress, ensuring the timeliness and accuracy of the evaluation. Therefore, the fitness function is given as follows:

[0283]

[0284] in:

[0285] The core of the task requirement-capability matching item is to sort the capability requirements by weight from high to low, and prioritize matching high-weight items with clusters that have better corresponding capabilities.

[0286] Here is the time decay term, where, It is the time decay coefficient, used to adjust the contribution of task completion speed to the fitness function, and its value depends on the actual problem scenario;

[0287] S4.3 Construct the velocity update equation for the particle swarm optimization algorithm. This equation should consider at least the following parameters: inertia weight. Individual learning factors Social learning factors and global learning guidance factors Compared with the fixed parameters used in general particle swarm optimization algorithms, the dynamic parameter adjustment strategy adopted can adaptively adjust the velocity update equation in different search stages and complex environments.

[0288] ① Inertia weight For the exploration and development of balancing algorithms, considering the current scenario of high temperature, high humidity, and low visibility, the inertia weight should be reduced to accelerate convergence, avoid local optima, and quickly adapt to environmental changes. The initial value of the inertia weight is set as follows: The inertial weight can be obtained. The expression is:

[0289]

[0290] in, These represent the influence coefficients of temperature, humidity, and visibility on the inertial weight, respectively. These coefficients need to be adjusted through experiments or according to the characteristics of the specific problem.

[0291] ② Individual learning factors Used to represent particles The ability to learn from its own historical best position, combined with the current scenario, should be considered. When the cluster's perception, communication, or decision-making capabilities are better matched with the needs of the current task, individual learning should be strengthened, and the use of its own strengths should be encouraged. Let the initial value of the individual learning factor be... Individual learning factors can be obtained. The expression is:

[0292]

[0293] The calculation of the matching degree in this formula further includes the following steps:

[0294] S4.3.1 Normalize the capabilities of each cluster to make all capabilities comparable on the same scale. , , It is the maximum capacity obtained by traversing each cluster;

[0295] S4.3.2. Improve the sensitivity of cluster capabilities to demand weights through exponential functions, thereby improving the sensitivity of individual learning factors to changes.

[0296] S4.3.3, after summation, compare with the total number of tasks. Divide to ensure The adjustment range should remain within a reasonable range depending on the scale of the task;

[0297] S4.3.4, Matching Degree Influence Coefficient Calibrated Through Experiments Controlling the matching degree The adjustment range;

[0298] ③ Social learning factors This represents the ability of a particle to learn towards the global optimal position. Considering the current scenario, under extreme weather conditions (such as high temperature or low visibility), individual performance may be unstable. In such cases, greater reliance on group experience is necessary. Let the initial value of the individual learning factor be [value missing]. The social learning factor can be obtained. The expression is:

[0299]

[0300] in:

[0301] It is the Sigmoid function, used to map changes in weather conditions to adjustments in social learning factors. The Sigmoid function makes... It will not fluctuate drastically due to minor changes in environmental factors, and can better reflect the relationship between environmental factors and... The complex nonlinear relationship between them;

[0302] It is the minimum value of the social learning factor, used to ensure that particles have the ability to learn towards the global optimum under any circumstances;

[0303] It is the maximum value of the social learning factor, used to avoid excessive learning intensity that could lead to instability in the particle swarm optimization algorithm;

[0304] The influence coefficient of meteorological condition changes, as determined experimentally, is used to control the impact of meteorological condition changes on... The impact;

[0305] ④ Global guiding factor To respond to sudden environmental changes and changes in mission capability requirements, the initial value of the global guiding factor is set to... The global learning guidance factor can be obtained. The expression is:

[0306]

[0307] in, It is the adjustment coefficient measured experimentally, representing the degree of influence of the gradient term on the global guiding factor;

[0308] By solving the gradient of each condition Norms can characterize the drastic changes of various conditions over time. Under sudden conditions, an increase in the gradient value will affect the global guiding factor. This increases the value of the particle swarm optimization algorithm, thereby guiding it to focus more on the global optimum and preventing it from getting trapped in local optima.

[0309] S4.4. Combining the above parameters, the velocity update equation for the particle swarm optimization algorithm is obtained:

[0310]

[0311] in, It is a particle The best position it has found in its own history. It is in the interval Uniformly distributed random numbers are introduced to increase the randomness and diversity of the search. It is the optimal position found by the entire particle swarm in history. It is a pre-defined global guidance location, determined based on prior knowledge of the problem or other information. and Each of these is a random number that is uniformly distributed within the interval;

[0312] For particles Perform the following multiple iterations:

[0313]

[0314] Let the particle be at the th The position of the wheel is Calculate the corresponding fitness value ,like Then update ,like If so, no action is taken;

[0315] S4.5 After all particle iterations are complete, compare all... Choose the highest fitness value. As At this time This is the optimal task allocation scheme, which will... Set as .

[0316] See Figure 2 By comparing the improved particle swarm optimization algorithm proposed in this invention with the standard particle swarm optimization algorithm, it can be seen that the improved particle swarm optimization algorithm proposed in this invention performs better in terms of fitness.

[0317] In this embodiment, step S5 defines a dynamic Bayesian multidimensional coupled state space composed of meteorological state, capability state, task state, and task allocation strategy. Compared to the single-type state modeling of ordinary dynamic Bayesian networks, it can achieve accurate simulation of complex task assignment scenarios, and its expression is:

[0318]

[0319] in, Indicates task status:

[0320]

[0321] Discretize each function in the state space, starting with temperature. For example, the meteorological conditions are discretized:

[0322]

[0323] in:

[0324] The temperature evolution coefficient under meteorological conditions where the initial values ​​are known;

[0325] To maintain environmental equilibrium temperature, the current temperature With rate Towards Approaching and simulating the natural thermal equilibrium process;

[0326] For temperature, there is a Gaussian noise term.

[0327] Similarly, regarding humidity and visibility A similar method is also used;

[0328] Secondly, let's consider perception. Discretize cluster capabilities as an example:

[0329]

[0330] in:

[0331] The coefficients for the evolution of perceptual ability, given their initial values;

[0332] The Gaussian noise term represents the perception capability;

[0333] Similarly, regarding communication capabilities and decision-making ability A similar method is also used;

[0334] set up The evolution coefficient matrix:

[0335]

[0336] against Online updates are employed: In typical dynamic Bayesian networks, parameters are usually updated... The fixed values, dependent on offline training data, are ill-suited to real-time changing environments. In contrast, this application utilizes an online mechanism for continuous updates. This allows for real-time reflection of the latest data distribution, improving the adaptability of the entire simulation process to dynamic environments. Online rule updates typically employ gradient descent.

[0337]

[0338] Parameters are dynamically adjusted using the gradient ascent method. To maximize the state transition probability:

[0339]

[0340] in:

[0341] An adaptive learning rate can be expressed as:

[0342]

[0343] When the historical gradient is large, reduce the learning rate. To avoid oscillations, when the historical gradient is small, maintain or increase the learning rate. This accelerates convergence;

[0344] The initial learning rate is set to a fixed value;

[0345] For a historic moment Bayesian gradient;

[0346] This is the average of the historical sum of squared gradients, used to quantify the cumulative fluctuations in parameter updates;

[0347] set up Observed variables in dynamic Bayesian networks:

[0348]

[0349] The observed variables are various data obtained from real-world scenarios, including measurements from weather sensors. The current capabilities obtained from the self-inspection of the unmanned cluster and task execution status. .

[0350] In this embodiment, S5 further includes:

[0351] S5.1 Perform forward inference on the dynamic Bayesian network, with the goal of determining the current state. and task allocation scheme Predict the state distribution at the next time step. The specific steps are as follows:

[0352] S5.1.1, Set the state include:

[0353] Weather conditions:

[0354] Ability Status:

[0355] Task Status:

[0356] Each state component is conditionally independent, and the joint probability decomposition is as follows:

[0357]

[0358] Taking temperature T as an example, to calculate the transition probability of meteorological states, in dynamic Bayesian networks, the state transitions are usually assumed to follow a normal distribution:

[0359]

[0360] In the formula, the mean ,variance It needs to be measured separately through experiments or historical data to reflect the intensity of random disturbances under different meteorological conditions;

[0361] With perception For example, calculate the transition probability of the cluster's capability state:

[0362]

[0363] In the formula, the mean ,variance It also needs to be determined separately through experiments or historical data;

[0364] S5.1.2, Use the sigmoid function to calculate the task execution status at time t. Within the range, calculate the transition probability of the task state:

[0365] ;

[0366] S5.2. Perform a posteriori update on the dynamic Bayesian network, with the aim of updating the data based on all known historical observations. and the current task allocation scheme Under the condition of system state Optimal estimate Suppose that the transition of the observed variable follows:

[0367] Meteorological observation :

[0368] Self-assessment of capabilities :

[0369] Status Report :

[0370] The joint observational likelihood was obtained:

[0371]

[0372] The formula for posterior update is:

[0373]

[0374] in:

[0375] It can be represented as:

[0376]

[0377] In the above formula, The formula for calculating the posterior probability is:

[0378]

[0379] Let be the normalization constant, expressed as:

[0380] .

[0381] In this embodiment, in step S6... After the time-sharing simulation is completed, the task allocation plan is determined. The inference results are evaluated, and an evaluation vector is constructed. :

[0382]

[0383] ① Vector middle, Represents task execution efficiency, expressed as:

[0384]

[0385] in, This is a time decay factor that penalizes delays in task completion time; the longer the delay, the lower the performance contribution. For the task The actual completion time The starting time of the simulation. For the task Expected value of completion The calculation expression is:

[0386]

[0387] Set the task to be executed The performance loss is The expression is:

[0388]

[0389] For tasks with high efficiency loss Reassign clusters based on capability matching. :

[0390]

[0391] In this formula, To the task The personalized learning rate can be expressed as: , The base learning rate needs to be determined through multiple trials based on experience.

[0392] ② Vector middle, Representing resource consumption, the expression is:

[0393]

[0394] in, Representing clusters The resource consumption for performing sensing, communication, and decision-making tasks, taking sensing as an example, can be expressed as:

[0395]

[0396] In this formula:

[0397] The energy consumption coefficient per unit capacity is determined by the cluster hardware;

[0398] The expected value of perceptual ability is expressed mathematically as follows:

[0399]

[0400] In this formula, Represents a cluster The discretized value of the perceptual ability;

[0401] Similarly, the energy consumption calculation process for communication and decision-making tasks is similar;

[0402] Set up a cluster The energy consumption is For high-energy-consuming clusters, task load should be reduced and resource usage balanced:

[0403]

[0404] In this formula, This is the energy consumption penalty coefficient, used to control the intensity of the penalty; its value is determined by the actual battlefield environment.

[0405] The exponential term maps energy consumption to interval, when When the value approaches 1, it indicates that the cluster has high energy consumption. Approaching , Significantly reduced;

[0406] ③ Vector middle, Represents system robustness, used to measure the system's ability to maintain task performance in the face of sudden environmental changes and changes in task requirements. Let be the total number of time steps in the dynamic Bayesian extrapolation. The expression is:

[0407]

[0408] By averaging the cumulative changes in environmental mutations and task requirements over the entire time period, the impact of the randomness of a single mutation on the evaluation results can be avoided.

[0409] By normalizing the design, robustness R is mapped to... The larger the value in the range, the more stable the system.

[0410] Set up a cluster robustness is The expression is:

[0411]

[0412] in, In order to be in Time Cluster Task in progress A set;

[0413] For highly robust clusters, tasks should be assigned preferentially:

[0414]

[0415] In this formula, For robust linear enhancement terms, To enhance the coefficient, the value is determined by the actual battlefield environment;

[0416] Cluster The higher the robustness, the larger the linear enhancement term. Significant increase.

[0417] In this embodiment, in step S6, the combined... and This leads to a multi-objective joint optimization formula:

[0418]

[0419] in, These are the weighting coefficients for each target, and their specific values ​​need to be determined based on the actual mission requirements in the battlefield environment.

[0420] Adjusted Input is fed into a dynamic Bayesian network to obtain a new evaluation vector. :

[0421]

[0422] Let the optimization thresholds for each objective be as follows: If the following convergence condition is met:

[0423]

[0424] Then stop the optimization and obtain the optimal task allocation scheme. ;

[0425] Otherwise, repeat steps S5 and S6 until the above convergence condition is met.

[0426] See Figure 3 By comparing the improved dynamic Bayesian network proposed in this invention with the standard dynamic Bayesian network, it can be seen that the improved dynamic Bayesian network proposed in this invention performs better in all components of the evaluation vector.

[0427] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the above-described method for optimizing digital twin simulations of unmanned swarm collaborative operations under complex weather conditions.

[0428] See Figure 4 The present invention also provides a computer device, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the steps of the above-described method for optimizing digital twin simulations of unmanned swarm collaborative operations under complex weather conditions.

[0429] This invention also provides a computer program product containing instructions that, when run on a computer, causes the computer to execute the steps of the above-described method for optimizing digital twin simulations of unmanned swarm collaborative combat under complex weather conditions.

[0430] It is understood that the systems, devices and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention, and the explanations, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above-mentioned digital twin simulation optimization method for unmanned swarm collaborative combat under complex weather conditions.

[0431] It should be noted that those skilled in the art will understand that all or part of the steps implemented in the embodiments of the present invention can be implemented entirely or partially by software, hardware, firmware, or any combination thereof. When implemented in hardware, it can be implemented entirely or partially by purchasing standard parts or modifications. When implemented in software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid state disks (SSDs)).

[0432] In summary, this invention addresses the shortcomings of traditional simulation methods in meteorological dynamic modeling, quantification of swarm collaborative capabilities, and real-time strategy optimization. It provides a digital twin simulation optimization method for unmanned swarm collaborative operations under complex meteorological conditions. This method involves: first, constructing a battlefield digital twin model based on meteorological parameters and unmanned swarm performance parameters, and establishing a nonlinear evolution model of meteorological conditions by introducing cross-coupling terms into the Lorenz system; second, designing a dynamic attenuation factor to quantify the nonlinear attenuation effects of temperature, humidity, and visibility on the unmanned swarm's perception, communication, and decision-making capabilities, and constructing a swarm collaborative capability model by combining geometric mean and variance constraints; third, achieving a deep improvement to the particle swarm optimization algorithm by dynamically adjusting inertia weights, learning factors, and global guidance factors, thereby obtaining a task allocation strategy based on this algorithm; fourth, constructing a meteorological-capability-task-strategy state space using a multidimensional coupled dynamic Bayesian network, and combining forward prediction, posterior update, and online parameter optimization mechanisms to achieve accurate simulation of the task allocation strategy; and finally, establishing a multidimensional evaluation system based on the simulation results, continuously improving the scheme through iterative optimization, and ultimately forming an optimal task allocation strategy adapted to the battlefield situation. This invention significantly improves the robustness and simulation accuracy of task allocation in complex environments, providing highly adaptive decision support for unmanned swarm collaborative operations.

[0433] It should be understood that the examples and embodiments described herein are for illustrative purposes only and are not intended to limit the invention. Those skilled in the art can make various modifications or changes based on them. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A digital twin simulation optimization method for unmanned swarm cooperative operations under complex weather conditions, characterized in that, Includes the following steps: S1. Collect meteorological parameters from multiple sources and performance parameters of unmanned swarms to construct a digital twin model of unmanned swarm collaborative operations under complex meteorological conditions. S2. Introduce cross-coupling terms into the Lorenz system and use the least squares method to dynamically optimize the parameters obtained from S1 to construct a nonlinear meteorological condition model. S3. Based on the nonlinear meteorological condition model in S2, design a dynamic attenuation factor to quantify the nonlinear attenuation effect of meteorological conditions on the performance parameters of intelligent agents in the unmanned swarm. Combine geometric mean, variance constraint and cooperative gain mechanism to construct a swarm cooperative capability model. S4. Improve the particle swarm optimization algorithm by designing a fitness function based on the cluster collaboration capability model in S3 and the actual task requirements, and dynamically adjusting the inertia weight, adaptive learning factor and global guidance factor to obtain a task allocation strategy with high matching degree. S5. Construct a multi-dimensional coupled dynamic Bayesian network of meteorology, capability, task, and strategy, and combine forward prediction, posterior update, and online learning mechanisms to achieve accurate deduction of the task allocation strategy in S4. S6. Based on the deduction results of S5, construct a multi-dimensional evaluation system including task efficiency, resource consumption and system robustness, and continuously improve the task allocation strategy in S4 through iterative optimization. The multi-source meteorological parameters in S1 include at least temperature, humidity, and visibility, and the performance parameters of the unmanned swarm include at least the perception, communication, and decision-making capabilities data of each agent within the swarm.

2. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 1, characterized in that, The meteorological states defined in S2 include: Where T is temperature, H is humidity, and V is visibility; The nonlinear differential equation structure of the Lorenz system reflects the nonlinear dynamic characteristics of real meteorological systems. Cross-coupling terms are introduced into this nonlinear differential equation to extend the nonlinear interactions between different meteorological conditions, resulting in the following nonlinear meteorological condition model: in: , and These are all parameters that are originally present in the Lorenz system, namely Prandtl's constant, Rayleigh's constant, and a space-dependent constant; , and These are the coupling coefficients under different meteorological conditions, reflecting the nonlinear interactions between meteorological elements.

3. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 1, characterized in that, In S3, the three capabilities of the k-th unmanned cluster are defined as follows: decision-making capability. , communication ability and perception ability The unmanned swarm has There are three intelligent agents, each possessing the three capabilities mentioned above, defined as follows: , and ; Introducing dynamic decay factor , and The attenuation of sensing ability, communication ability, and decision-making ability under the influence of meteorological conditions is described, and the expressions for the dynamic attenuation factors of sensing ability, communication ability, and decision-making ability are obtained respectively: in: This represents the temperature sensitivity coefficient, which is determined by the material properties of the optical device. This indicates the absolute deviation between the current temperature and the sensor's calibrated temperature. The minimum visibility threshold is determined by the specific optical system. and These are the specific decay coefficient and the power-law exponent, respectively; This represents the critical room temperature threshold that triggers dynamic frequency reduction. Indicates the highest room temperature threshold allowed by the processor; This represents the processor performance degradation factor; This is the minimum performance guarantee threshold; The better the heat dissipation design, the smoother the performance degradation curve will be, thus controlling the steepness of the curve.

4. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 3, characterized in that, In S3, the communication capabilities of the unmanned cluster are discussed. Perception ability and decision-making ability The modeling is as follows: ①.Communication ability Modeling: While introducing a geometric mean mechanism to reflect the limitation of communication link performance by the weakest node, this paper also considers the gains in communication capability brought by cluster cooperation, and based on this, gives... The expression: in: This represents the geometric mean of the communication capabilities of each agent within the cluster, where... Represents the number of agents within the cluster; It is a factor affecting communication stability, among which, It is the variance of the communication capabilities of all intelligent agents; the larger the variance, the worse the communication stability. It is the degree of influence of the experimentally determined variance on the collaborative ability; This represents the gain in communication capabilities that cluster collaboration brings, where... The communication capability cooperation coefficient is determined by the cluster hardware. For the number of times of communication and collaboration, For the number of nodes, Expanding the network will increase the complexity of the unmanned swarm communication network, causing the communication capacity to decrease logarithmically. ② Perception ability Modeling: In addition to considering the gain of sensing capability from cluster collaboration, the gain of sensing capability from cluster communication capability is also introduced, and based on this, the following is given: The expression: in: It is an intelligent agent Neighbor intelligent agents in The sum of real-time communication capabilities; It is all intelligent agents in The sum of communication capabilities at any given moment; This is the gain coefficient of the cluster communication capability on the sensing capability, as determined in the experiment. The overall representation indicates the gain of the communication network on sensing capabilities; This represents the gain in perception capability brought by cluster collaboration, where... This is the sensing capability collaboration coefficient, and its value is determined by the characteristics of the cluster hardware. For the number of times communication and collaboration were conducted; ③ Decision-making ability Modeling: In addition to using the geometric mean mechanism to balance the decision weights of each agent, a fault-tolerant constraint mechanism needs to be constructed to prevent global decision failure due to local communication failures, while also taking into account the gains in decision-making capabilities brought by cluster collaboration. Based on this, the following is proposed: The expression: in: The geometric mean of the decision-making capabilities of each agent within the cluster; It is the minimum value of the product of the communication capabilities of all intelligent agents, reflecting that the decision-making process is limited by the weakest communication link; These are experimentally determined coefficients used to adjust the degree of influence of communication capabilities on collaborative decision-making. The benefits that cluster collaboration brings to decision-making capabilities, among which... This is the decision-making capability collaboration coefficient, and its value is determined by the characteristics of the cluster hardware. For the number of times communication and collaboration were conducted; To reduce the time spent on excessive group negotiation, a hyperbolic tangent function is introduced to simulate the diminishing marginal effect. The nonlinear characteristics of the saturation region of the hyperbolic tangent function can accurately describe the relationship between the number of negotiations and the efficiency of decision-making.

5. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 1, characterized in that, In S4, let the unmanned cluster set be... Combat mission Divided into several sub-tasks Particle swarm optimization algorithm is used for different clusters Assign combat missions: S4.1 Encode the particles using a three-dimensional array. To define the position of the particle, where, Indices representing particle indices Indicates the index of the task. This represents the index of an unmanned cluster, and At the same time satisfy For all and All are true; S4.2 Construct the fitness function of the particle swarm optimization algorithm. This function guides the particle swarm optimization algorithm to search towards the optimal solution. Let... The set consisting of weights representing the different cluster capability requirements of different tasks is defined as follows: In this set: Indicates task At any moment The required weight of unmanned swarm perception capabilities Indicates task At any moment The weight of the demand for unmanned cluster communication capabilities Indicates task At any moment The required weight of unmanned swarm decision-making capabilities Traverse the task set All tasks in; Based on this, the fitness function is given as follows: in: The core of the task requirement-capability matching item is to sort the capability requirements by weight from high to low, and prioritize matching high-weight items with clusters that have better corresponding capabilities. Here is the time decay term, where, It is the time decay coefficient, used to adjust the contribution of task completion speed to the fitness function, and its value depends on the actual problem scenario; S4.3 Construct the velocity update equation for the particle swarm optimization algorithm. This equation should consider at least the following parameters: inertia weight. Individual learning factors Social learning factors and global learning guidance factors ; ① Inertia weight For the exploration and development of balancing algorithms, the initial value of the inertia weight is set to... The inertial weight can be obtained. The expression is: in, These represent the influence coefficients of temperature, humidity, and visibility on the inertial weight, respectively. ② Individual learning factors Used to represent particles The ability to learn from its own historical best position, with the initial value of the individual learning factor set at [value]. Individual learning factors can be obtained. The expression is: The calculation of the matching degree in this formula further includes the following steps: S4.3.1 Normalize the capabilities of each cluster to make all capabilities comparable on the same scale. , , It is the maximum capacity obtained by traversing each cluster; S4.3.

2. Improve the sensitivity of cluster capabilities to demand weights through exponential functions, thereby improving the sensitivity of individual learning factors to changes. S4.3.3, after summation, compare with the total number of tasks. Divide to ensure The adjustment range should remain within a reasonable range depending on the scale of the task; S4.3.4, Matching Degree Influence Coefficient Calibrated Through Experiments Controlling the matching degree The adjustment range; ③ Social learning factors This represents the ability of a particle to learn towards the global optimal position. Let the initial value of the individual learning factor be... The social learning factor can be obtained. The expression is: in: It is the Sigmoid function, used to map changes in weather conditions to adjustments in social learning factors; It is the minimum value of the social learning factor, used to ensure that particles have the ability to learn towards the global optimum under any circumstances; It is the maximum value of the social learning factor, used to avoid excessive learning intensity that could lead to instability in the particle swarm optimization algorithm; The influence coefficient of meteorological condition changes, as determined experimentally, is used to control the impact of meteorological condition changes on... The impact; ④ Global guiding factor To respond to sudden environmental changes and changes in mission capability requirements, the initial value of the global guiding factor is set to... The global learning guidance factor can be obtained. The expression is: in, It is the adjustment coefficient measured experimentally, representing the degree of influence of the gradient term on the global guiding factor; By solving the gradient of each condition Norms can characterize the drastic changes of various conditions over time. Under sudden conditions, an increase in the gradient value will affect the global guiding factor. This increases the value of the particle swarm optimization algorithm, thereby guiding it to focus more on the global optimum and preventing it from getting trapped in local optima. S4.

4. Combining the above parameters, the velocity update equation for the particle swarm optimization algorithm is obtained: in, It is a particle The best position it has found in its own history. It is in the interval Uniformly distributed random numbers within the range. It is the optimal position found by the entire particle swarm in history. It is a pre-defined global guidance location, determined based on prior knowledge of the problem or other information. and Each of these is a random number that is uniformly distributed within the interval; For particles Perform the following multiple iterations: Let the particle be at the th The position of the wheel is Calculate the corresponding fitness value ,like Then update ,like If so, no action is taken; S4.5 After all particle iterations are complete, compare all... Choose the highest fitness value. As At this time This is the optimal task allocation scheme, which will... Set as .

6. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 1, characterized in that, In S5, a dynamic Bayesian multidimensional coupled state space is defined, consisting of meteorological state, capability state, task state, and task allocation strategy. : in, Indicates task status: Based on this, the various meteorological parameters and cluster capabilities in the state space are discretized respectively; set up The evolution coefficient matrix is ​​used for discretization. against Online updates are used: Continuous updates This allows for real-time reflection of the latest data distribution, improving the adaptability of the entire simulation process to dynamic environments. Online rule updates typically employ gradient descent. Parameters are dynamically adjusted using the gradient ascent method. To maximize the state transition probability: in: An adaptive learning rate can be expressed as: When the historical gradient is large, reduce the learning rate. To avoid oscillations, when the historical gradient is small, maintain or increase the learning rate. This accelerates convergence; The initial learning rate is set to a fixed value; For a historic moment Bayesian gradient; This is the average of the historical sum of squared gradients, used to quantify the cumulative fluctuations in parameter updates; set up Observed variables in dynamic Bayesian networks: The observed variables are various data obtained from real-world scenarios, including measurements from weather sensors. The current capabilities obtained from the self-inspection of the unmanned cluster and task execution status. .

7. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 6, characterized in that, S5 further includes: S5.1 Perform forward inference on the dynamic Bayesian network, with the goal of determining the current state. and task allocation scheme Predict the state distribution at the next time step. The specific steps are as follows: S5.1.1, Set the state include: Weather conditions: Ability Status: Task Status: Each state component is conditionally independent, and the joint probability decomposition is as follows: Based on this, the transition probabilities of each meteorological state and cluster capability state are calculated respectively; S5.1.2, Use the sigmoid function to calculate the task execution status at time t. Within the range, calculate the transition probability of the task state: ； S5.

2. Perform a posteriori update on the dynamic Bayesian network, with the aim of updating the data based on all known historical observations. and the current task allocation scheme Under the condition of system state Optimal estimate Suppose that the transition of the observed variable follows: Meteorological observation : Self-assessment of capabilities : Status Report : The joint observational likelihood was obtained: The formula for posterior update is: in: It can be represented as: In the above formula, The formula for calculating the posterior probability is: Let be the normalization constant, expressed as: 。 8. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 1, characterized in that, In S6 After the time-sharing simulation is completed, the task allocation plan is determined. The inference results are evaluated, and an evaluation vector is constructed. : ① Vector middle, Represents task execution efficiency, expressed as: in, This is a time decay factor that penalizes delays in task completion time; the longer the delay, the lower the performance contribution. For the task The actual completion time The starting time of the simulation. For the task Expected value of completion The calculation expression is: Set the task to be executed The performance loss is The expression is: For tasks with high efficiency loss Reassign clusters based on capability matching. : In this formula, To the task The personalized learning rate can be expressed as: , The base learning rate needs to be determined through multiple trials based on experience. ② Vector middle, Representing resource consumption, the expression is: in, Representing clusters Resource consumption for performing sensing, communication, and decision-making tasks; Set up a cluster The energy consumption is For high-energy-consuming clusters, task load should be reduced and resource usage balanced: In this formula, This is the energy consumption penalty coefficient, used to control the intensity of the penalty; its value is determined by the actual battlefield environment. The exponential term maps energy consumption to interval, when When the value approaches 1, it indicates that the cluster has high energy consumption. Approaching , Significantly reduced; ③ Vector middle, Represents system robustness, used to measure the system's ability to maintain task performance in the face of sudden environmental changes and changes in task requirements. Let be the total number of time steps in the dynamic Bayesian extrapolation. The expression is: By averaging the cumulative changes in environmental mutations and task requirements over the entire time period, the impact of the randomness of a single mutation on the evaluation results can be avoided. By normalizing the design, robustness R is mapped to... The larger the value in the range, the more stable the system. Set up a cluster robustness is The expression is: in, In order to be in Time Cluster Task in progress A set; For highly robust clusters, tasks should be assigned preferentially: In this formula, For robust linear enhancement terms, To enhance the coefficient, the value is determined by the actual battlefield environment; Cluster The higher the robustness, the larger the linear enhancement term. Significant increase.

9. The digital twin simulation optimization method for unmanned swarm cooperative combat under complex weather conditions as described in claim 8, characterized in that, In S6, the combined and This leads to a multi-objective joint optimization formula: in, , and These are the weighting coefficients for each target, and their specific values ​​need to be determined based on the actual mission requirements in the battlefield environment. Adjusted Input is fed into a dynamic Bayesian network to obtain a new evaluation vector. : Let the optimization thresholds for each objective be as follows: If the following convergence condition is met: Then stop the optimization and obtain the optimal task allocation scheme. ; Otherwise, repeat steps S5 and S6 until the above convergence condition is met.

Images