Performance evaluation method for vehicle-mounted firepower auxiliary decision-making system
By introducing Gaussian mapping, JAVA algorithm and early warning mechanism, as well as bidirectional sine chaos mapping in the transient search optimization algorithm, the problem that the algorithm is prone to fall into local optimality is solved, and the accuracy of performance evaluation of on-board firepower assisted decision-making system is improved.
Patent Information
- Application Number
- CN202410164208.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-02-05
- Publication Date
- 2024-12-31
AI Technical Summary
Transient search optimization algorithms are prone to local optimality in performance evaluation of on-board firepower assist decision-making systems, with low convergence accuracy, resulting in unsatisfactory evaluation results.
Introduce Gaussian mapping to initialize the search agent population position, adopt an improved transient search optimization algorithm, optimize position updates through JAVA algorithm and early warning mechanism, and use bidirectional sine chaos mapping to perform optimal position mutation, avoid local optimization and improve global Search capability.
The adaptability and search range of the algorithm are improved, and more accurate performance evaluation results of the on-board firepower assist decision-making system are obtained, which improves the accuracy of the evaluation.
Smart Images
Figure CN119226750A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of vehicle-mounted firepower auxiliary decision-making systems, and in particular to a performance evaluation method for vehicle-mounted firepower auxiliary decision-making systems. Background Art
[0002] The vehicle-mounted firepower auxiliary decision-making system is one of the key technologies for the victory of land warfare platforms. Taking into account military needs, application environment, mission attributes, system portability and scalability and many other factors, in-depth research on the technical characteristics of the vehicle-mounted firepower auxiliary decision-making system and evaluation of its performance are the basic work for further design and optimization of the vehicle-mounted firepower auxiliary decision-making system.
[0003] As one of the research contents of machine learning, support vector machine (SVM) has been widely used in performance evaluation in the military field. During the SVM training process, the quality of the selection of the penalty factor C of SVM and the RBF kernel function parameter g directly affects the accuracy of the final performance evaluation results of the vehicle-mounted firepower auxiliary decision-making system. Intelligent optimization algorithm is an effective SVM model parameter optimization method. Transient search optimization algorithm (TSOA) is a new intelligent optimization algorithm that simulates the transient behavior of switching circuits containing storage elements such as inductors and capacitors. It can also be applied to the performance evaluation of vehicle-mounted firepower auxiliary decision-making systems. However, there are still some defects in the transient search optimization algorithm, which makes the algorithm easy to fall into local optimality and low convergence accuracy. When evaluating the performance of vehicle-mounted firepower auxiliary decision-making systems, it often fails to achieve ideal evaluation results. Summary of the invention
[0004] In order to improve the accuracy of the performance evaluation of the vehicle-mounted firepower auxiliary decision-making system, the present invention provides a performance evaluation method of the vehicle-mounted firepower auxiliary decision-making system:
[0005] The performance evaluation method of the vehicle-mounted firepower auxiliary decision-making system includes the following steps:
[0006] According to the evaluation indicators and original data of the vehicle-borne firepower auxiliary decision-making system performance, a data set for the performance evaluation of the vehicle-borne firepower auxiliary decision-making system is constructed;
[0007] The classification accuracy of SVM is calculated using the data set, the classification accuracy of SVM with 5-fold cross validation is used as the objective function, and the upper and lower limits of the penalty factor C and the RBF kernel function parameter g are determined;
[0008] According to the objective function, the penalty factor C and the RBF kernel function parameter g are searched by an improved transient search optimization algorithm to obtain the optimal parameters; wherein the improved transient search optimization algorithm introduces Gaussian mapping to initialize the position of the search agent population, and introduces the position update method and early warning mechanism of the JAVA algorithm to optimize the position update method of the original search agent; the improved transient search optimization algorithm also includes mutating the optimal search agent through a bidirectional sine chaotic mapping;
[0009] The optimal penalty factor C and RBF kernel function parameter g are used as SVM parameters to establish a performance evaluation model of the vehicle-mounted firepower auxiliary decision-making system based on SVM.
[0010] The original data of the vehicle-mounted firepower auxiliary decision-making system to be evaluated is input into the trained vehicle-mounted intelligent microgrid performance evaluation model to obtain the evaluation results.
[0011] Preferably, the Gaussian mapping is used to initialize the position of the search agent population, comprising the following steps:
[0012] Determine the size N of the population, search for the lower bound LB of the agent optimization and the upper bound UB of the agent optimization;
[0013] Generate the next random number x through Gaussian mapping t+1 :
[0014]
[0015] In the formula, mod(·) is the remainder function, x t is the current random number;
[0016] Initialize the search agent position using the generated Gaussian random number:
[0017] X(0) = (UB - LB) × x t +LB
[0018] Where X(0) is the position of the initial search agent.
[0019] Preferably, the location update of the search agent in the improved transient search optimization algorithm comprises the following steps:
[0020] By introducing the JAVA algorithm, the improved search agent position update formula is as follows:
[0021]
[0022] in:
[0023] C(t)=k·z·λ3+1
[0024] T=2·z·λ2-z
[0025]
[0026] Among them, α1 and α2 are random learning weights, which adjust the influence of the search agent's learning from the optimal search agent and learning from other search agents in the population, and the sum is 1; X wt (t) represents the worst position of the search agent at the tth iteration; X(t+1) is the position of the search agent at the t+1th iteration; X bs (t) represents the optimal position of the search agent at the tth iteration; X(t) is the position of the search agent at the current tth iteration; C(t) and T are random thermal resistance coefficients; z is the attenuation coefficient, which gradually decreases from 2 to 0 as the number of iterations increases; k is a constant, k = 0, 1, 2...; λ1, λ2 and λ3 are random numbers in [0, 1]; Miter is the number of iterations;
[0027] Introducing an early warning mechanism, assuming that the search agents that are aware of the danger account for 10% to 20% of the total number of search agents, and the initial positions of these search agents are randomly generated in the population, the position formula of the search agents that are aware of the danger is as follows:
[0028]
[0029] Among them, X i,j (t+1) is the i-th search agent position X at the t+1th iteration i The jth dimension of (t+1); X i,j (t) is the i-th search agent position X at the current t-th iteration i The jth dimension of (t); X bs(j) (t) represents the optimal position X of the tth iteration bs The jth dimension of (t); X wt(j) (t) represents the worst position X of the tth iteration wt (t) is the jth dimension; β is the step size control parameter, which is a random number that follows a normal distribution with a mean of 0 and a variance of 1; η is a random number in [0,1], which indicates the moving direction of the search agent and is also the step size control parameter; F i is the fitness value of the current search agent; F g and F w are the current global maximum and minimum fitness values respectively; ε is a small constant; when F i >F g When F i =F g When , it means that the search agent in the middle of the population is aware of the danger and needs to move to other search agents to reduce the risk;
[0030] Calculate the fitness value:
[0031] fitness(t)=F(X(t+1))
[0032] Where F(·) is the fitness value calculation function;
[0033] Record the best search agent in the iteration.
[0034] Preferably, in the improved transient search optimization algorithm, bidirectional sine mutation is performed on the optimal search agent position, comprising the following steps:
[0035] For dimension j, the sine chaos value is calculated according to the current number of iterations, and the positive and negative directions are switched with equal probability:
[0036] sinValue=sin(πx0)
[0037]
[0038] In the formula, rand is a random number between 0 and 1; x0 is the iteration sequence value;
[0039] Perform mutation perturbation on the optimal position:
[0040] X bs(j) (t+1)'=X bs(j) (t+1)+SinValue×X bs(j) (t+1)
[0041] Where: X bs(j) (t+1) represents the optimal position X of the t+1th iteration bs The jth dimension of (t+1); X bs(j) (t+1)' is the optimal position X after the mutation disturbance bs The jth dimension of (t+1);
[0042] Greedy Update:
[0043]
[0044] After each dimension has been mutated, stop mutation.
[0045] Beneficial effects of the present invention:
[0046] The present invention proposes a performance evaluation method for a vehicle-mounted firepower auxiliary decision-making system, adopts an improved search agent optimization algorithm to obtain the optimal parameters of the SVM, and evaluates through the optimized SVM to obtain accurate vehicle-mounted firepower auxiliary decision-making system performance evaluation results. The method improves the position update method of the search agent, introduces the JAVA algorithm and the early warning mechanism, comprehensively considers the optimal position of the search agent, references the worst position and updates away from the worst position, the early warning mechanism and other factors, avoids the local optimum in each iteration, and generates new solutions according to different conditions, realizes the increase of the algorithm search range, and enhances the adaptability of the algorithm, thereby improving the accuracy of the performance evaluation of the vehicle-mounted firepower auxiliary decision-making system. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is a flow chart of a method for evaluating the performance of a vehicle-mounted firepower auxiliary decision-making system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0048] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0049] Example 1
[0050] This embodiment proposes a method for evaluating the performance of a vehicle-mounted firepower auxiliary decision-making system. The specific flow chart is as follows: Figure 1 As shown, the following steps are included:
[0051] S1: Based on the evaluation indicators of the performance of the vehicle-borne firepower auxiliary decision-making system (for example, optimal damage effect, shortest task execution time, etc.) and the original data, a data set for the performance evaluation of the vehicle-borne firepower auxiliary decision-making system is constructed, and the data set is divided into a training data set and a test data set.
[0052] S2: Establish the objective function funtion of the performance evaluation of the vehicle-mounted firepower auxiliary decision-making system based on the improved transient search optimization algorithm. Because the classification accuracy of the SVM is calculated using the training data set, the classification accuracy of the SVM of the 5-fold cross-validation of the training data set can be used as the objective function, that is, the fitness function; at the same time, set the corresponding constraints: the upper and lower limits of the SVM penalty factor C and the RBF kernel function parameter g.
[0053] S3: Parameter setting, mainly including: the size of the search agent population (ie, the number of search agent individuals) N; the maximum number of iterations (ie, the condition for stopping iteration) Miter; the search agent optimization lower boundary LB; the search agent optimization upper boundary UB.
[0054] S4: Initialize the search agent population position using Gaussian mapping.
[0055] Gaussian map random number generation:
[0056]
[0057] Initialize the search agent position using the generated Gaussian random number:
[0058] X(0) = (UB - LB) × x t +LB
[0059] Where: X(0) is the position of the initial search agent.
[0060] S5: Search agent location update.
[0061] In the original transient search optimization algorithm, only the optimal search agent position is used for guidance to update the search agent position. In order to more effectively improve the global search capability of the search agent, the JAVA algorithm and early warning mechanism are introduced. The optimal position of the search agent, reference to the worst position and updating away from the worst position, early warning mechanism and other factors are comprehensively considered to avoid local optimality in each iteration, thereby improving the global search capability of the transient search optimization algorithm.
[0062] By introducing the JAVA algorithm, the improved search agent position update formula is as follows:
[0063]
[0064] in:
[0065] C(t)=k·z·λ3+1
[0066] T=2·z·λ2-z
[0067]
[0068] Among them, α1 and α2 are random learning weights, which adjust the influence of the search agent's learning from the optimal search agent and learning from other search agents in the population, and the sum is 1; X wt (t) represents the worst position of the search agent at the tth iteration; X(t+1) is the position of the search agent at the t+1th iteration; X bs (t) represents the optimal position of the search agent at the tth iteration; X(t) is the position of the search agent at the current tth iteration; C(t) and T are random thermal resistance coefficients; z is the attenuation coefficient, which gradually decreases from 2 to 0 as the number of iterations increases; k is a constant, k = 0, 1, 2...; λ1, λ2 and λ3 are random numbers in [0, 1]; Miter is the number of iterations;
[0069] Introducing an early warning mechanism, assuming that the search agents that are aware of the danger account for 10% to 20% of the total number of search agents, and the initial positions of these search agents are randomly generated in the population, the position formula of the search agents that are aware of the danger is as follows:
[0070]
[0071] Among them, X i,j (t+1) is the i-th search agent position X at the t+1th iteration i The jth dimension of (t+1); X i,j (t) is the i-th search agent position X at the current t-th iteration i The jth dimension of (t); X bs(j) (t) represents the optimal position X of the tth iteration bs The jth dimension of (t); X wt(j) (t) represents the worst position X of the tth iteration wt (t) is the jth dimension; β is the step size control parameter, which is a random number that follows a normal distribution with a mean of 0 and a variance of 1; η is a random number in [0,1], which indicates the moving direction of the search agent and is also the step size control parameter; F i is the fitness value of the current search agent; F g and F w are the current global maximum and minimum fitness values respectively; ε is a small constant; when F i >F g When F i =F g , it means that the search agent in the middle of the population is aware of the danger and needs to move to other search agents to reduce the risk.
[0072] S6: Calculate the fitness value.
[0073] fitness(t)=F(X(t+1))
[0074] Where F(·) is the fitness function used to calculate the fitness value.
[0075] S7: Record information. Record the best search agent in the current iteration.
[0076] S8: Perform bidirectional sine mutation on the optimal search agent in each dimension. For dimension j, first calculate the sine chaos value according to the current number of iterations, and switch the positive and negative directions with equal probability.
[0077] SinValue=sin(πx0)
[0078]
[0079] Then perform mutation perturbation on the optimal position:
[0080] X bs(j) (t+1)'=X bs(j) (t+1)+SinValue×X bs(j) (t+1)
[0081] Where: X bs(j) (t+1) represents the optimal position X of the t+1th iteration bs The j-th dimension of (t+1).
[0082] Greedy Update:
[0083]
[0084] After each dimension has been mutated, stop mutation.
[0085] S9: Record information. Record the best search agent in the current iteration.
[0086] S10: Repeat steps S5 to S9. When the maximum number of iterations Miter is reached, the algorithm stops and the optimal search agent is output. When the maximum number of iterations Miter is reached, the algorithm stops and the optimal search agent is output, that is, the optimal parameter C of SVM is obtained. best and g best .
[0087] S11: A performance evaluation model of the vehicle-mounted firepower auxiliary decision-making system is established with the output optimal parameters as SVM parameters, and the test data set is input into the model to obtain the evaluation results and evaluation accuracy of the performance of the vehicle-mounted firepower auxiliary decision-making system.
[0088] The key points of the present invention mainly include 3:
[0089] (1) By introducing Gaussian mapping to initialize the positions of the search agent population, the uniformity and diversity of the population position distribution can be improved, thereby enhancing the stability of the algorithm.
[0090] (2) The position update method of the search agent was improved by introducing a JAVA algorithm and an early warning mechanism. The optimal position of the search agent, reference to the worst position and updating away from the worst position, and the early warning mechanism were comprehensively considered to avoid local optimality in each iteration and generate new solutions based on different conditions, thereby increasing the search range of the algorithm and enhancing its adaptability.
[0091] (3) The bidirectional sine chaotic map mutation is used to mutate the optimal search agent, which enables the algorithm to escape from the local optimal solution in the later stage.
[0092] In this embodiment:
[0093] 1000 groups of samples of the vehicle-mounted firepower auxiliary decision-making system are selected, and 800 groups are randomly selected as training samples, and the remaining 200 groups are test samples. TSOA-SVM and ITSOA-SVM are used to evaluate the performance of the vehicle-mounted firepower auxiliary decision-making system. Using MATLAB as the simulation platform, the parameters in the TSOA algorithm are: N = 100, Maxiter = 500, the search range of C and g are both between 0-100, that is, LB = 0, UB = 100; the parameters in the ITSOA algorithm are: N = 100, Maxiter = 500, the search range of C and g are both between 0-100, that is, LB = 0, UB = 100.
[0094] The evaluation indicators of TSOA-SVM model and ITSOA-SVM model can be selected as: mean absolute error (MAE), mean relative error (MRE) and root mean square error (RMSE). As shown in Table 1, compared with TSOA-SVM, ITSOA-SVM has a higher accuracy in performance evaluation of vehicle-mounted firepower auxiliary decision-making system, that is, the SVM parameters searched by ITSOA are better than those searched by TSOA. The simulation results show that the ITSOA algorithm has stronger search capability than the TSOA algorithm, and the evaluation accuracy of ITSOA-SVM is higher than that of TSOA-SVM, which verifies the effectiveness of the method.
[0095] Table 1. Comparison of evaluation methods
[0096]
[0097] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. The performance evaluation method of the vehicle-mounted firepower auxiliary decision-making system is characterized by: The following steps are involved: According to the evaluation indicators and original data of the vehicle-borne firepower auxiliary decision-making system performance, a data set for the performance evaluation of the vehicle-borne firepower auxiliary decision-making system is constructed; The classification accuracy of SVM is calculated using the data set, the classification accuracy of SVM with 5-fold cross validation is used as the objective function, and the upper and lower limits of the penalty factor C and the RBF kernel function parameter g are determined; According to the objective function, the penalty factor C and the RBF kernel function parameter g are searched by an improved transient search optimization algorithm to obtain the optimal parameters; wherein the improved transient search optimization algorithm introduces Gaussian mapping to initialize the position of the search agent population, and introduces the position update method and early warning mechanism of the JAVA algorithm to optimize the position update method of the original search agent; the improved transient search optimization algorithm also includes mutating the optimal search agent through a bidirectional sine chaotic mapping; The optimal penalty factor C and RBF kernel function parameter g are used as SVM parameters to establish a performance evaluation model of the vehicle-mounted firepower auxiliary decision-making system based on SVM. The original data of the vehicle-mounted firepower auxiliary decision-making system to be evaluated is input into the trained vehicle-mounted intelligent microgrid performance evaluation model to obtain the evaluation results.
2. The vehicle-mounted firepower auxiliary decision-making system performance evaluation method according to claim 1 is characterized in that: The Gaussian mapping is used to initialize the position of the search agent population, including the following steps: Determine the size N of the population, search for the lower bound LB of the agent optimization and the upper bound UB of the agent optimization; Generate the next random number x through Gaussian mapping t+1 : In the formula, mod(·) is the remainder function, x t is the current random number; Initialize the search agent position using the generated Gaussian random number: X(0)=(UB-LB)×x t +LB Where X(0) is the position of the initial search agent.
3. The vehicle-mounted firepower auxiliary decision-making system performance evaluation method according to claim 1 is characterized in that: The position update of the search agent in the improved transient search optimization algorithm includes the following steps: By introducing the JAVA algorithm, the improved search agent position update formula is as follows: in: C(t)=k·z·λ3+1 T=2·z·λ2-z Among them, α1 and α2 are random learning weights, which adjust the influence of the search agent's learning from the optimal search agent and learning from other search agents in the population, and the sum is 1; X wt (t) represents the worst position of the search agent at the tth iteration; X(t+1) is the position of the search agent at the t+1th iteration; X bs (t) represents the optimal position of the search agent at the tth iteration; X(t) is the position of the search agent at the current tth iteration; C(t) and T are random thermal resistance coefficients; z is the attenuation coefficient, which gradually decreases from 2 to 0 as the number of iterations increases; k is a constant, k = 0, 1, 2...; λ1, λ2 and λ3 are random numbers in [0, 1]; Miter is the number of iterations; Introducing an early warning mechanism, assuming that the search agents that are aware of the danger account for 10% to 20% of the total number of search agents, and the initial positions of these search agents are randomly generated in the population, the position formula of the search agents that are aware of the danger is as follows: Among them, X i,j (t+1) is the i-th search agent position X at the t+1th iteration i The jth dimension of (t+1); X i,j (t) is the i-th search agent position X at the current t-th iteration i The jth dimension of (t); X bs(j) (t) represents the optimal position X of the tth iteration bs The jth dimension of (t); X wt(j) (t) represents the worst position X of the tth iteration wt (t) is the jth dimension; β is the step size control parameter, which is a random number that follows a normal distribution with a mean of 0 and a variance of 1; η is a random number in [0,1], which indicates the moving direction of the search agent and is also the step size control parameter; F i is the fitness value of the current search agent; F g and F w are the current global maximum and minimum fitness values respectively; ε is a small constant; when F i >F g When F i =F g When , it means that the search agent in the middle of the population is aware of the danger and needs to move to other search agents to reduce the risk; Calculate the fitness value: finess(t)=F(X(t+1)) Where F(·) is the fitness value calculation function; Record the best search agent in the iteration.
4. The vehicle-mounted firepower auxiliary decision-making system performance evaluation method according to claim 3 is characterized in that: In the improved transient search optimization algorithm, bidirectional sine mutation is performed on the optimal search agent position, including the following steps: For dimension j, the sine chaos value is calculated according to the current number of iterations, and the positive and negative directions are switched with equal probability: sinValue=sin(πx0) In the formula, rand is a random number between 0 and 1; x0 is the iteration sequence value; Perform mutation perturbation on the optimal position: X bs(j) (t+1)'=X bs(j) (t+1)+SinValue×X bs(j) (t+1) Where: X bs(j) (t+1) represents the optimal position X of the t+1th iteration bs The jth dimension of (t+1); X bs(j) (t+1)' is the optimal position X after the mutation disturbance bs The jth dimension of (t+1); Greedy Update: After each dimension has been mutated, stop mutation.