Method and system for evaluating equivalence of scale test of unmanned system

Abstract

The invention relates to an unmanned system scale test equivalence evaluation method and system. The method comprises the following steps: constructing a hierarchical evaluation index system of an unmanned system test; the index system comprises a plurality of top layer indexes and bottom layer indexes corresponding to the top layer indexes; respectively carrying out a small-scale scaling test and a large-scale prototype test, collecting original data corresponding to each underlying index, and carrying out quantitative processing on the original data to obtain a standardized value of each underlying index; constructing a capability value calculation model based on a linear weighting model, wherein the capability value calculation model is used for calculating the capability value of each top-layer index according to the standardized value of each bottom-layer index; calculating the combination weight of each bottom layer index by adopting a combination weighting method; based on the standardized values and the combined weights, calculating capability evaluation values of the small-scale scaling and the large-scale prototype test on each top index by using a capability value calculation model; and based on the deviation ratio between the capability evaluation values of the small-scale scaling test and the large-scale prototype test on each top index, judging the equivalent degree of the two tests.

Description

Technical Field

[0001] This invention relates to the field of unmanned system testing and evaluation technology, and in particular to a method and system for evaluating the equivalence of scaled-down unmanned system tests. Background Technology

[0002] With the rapid development of artificial intelligence and autonomous control technology, heterogeneous unmanned swarms composed of drones, unmanned ships, and other similar vessels are increasingly being used in reconnaissance, surveillance, and target acquisition missions. However, conducting large-scale, full-scale prototype tests during the research and development and verification of unmanned swarm systems is typically costly, time-consuming, and carries certain risks.

[0003] Therefore, conducting small-scale, low-cost scaled-down experiments to verify key algorithms and optimization strategies has become an efficient and necessary research method. However, scaled-down experiments and prototype experiments differ in terms of environment, scale, and platform performance. How to scientifically evaluate the equivalence of results from full-scale prototype experiments and small-scale scaled-down experiments, and ensure that the conclusions obtained from scaled-down experiments can effectively guide the design and optimization of prototype systems, is a key technical challenge currently facing the field of unmanned system testing and evaluation.

[0004] Existing evaluation methods mostly employ multi-indicator comprehensive evaluation systems, the core of which is determining the weight of each indicator. These methods heavily rely on data quality and sample size, and struggle to effectively handle some qualitative indicators that are difficult to quantify. Relying solely on any one method has significant limitations, making it difficult to comprehensively and scientifically assess the equivalence of two test scenarios for complex unmanned systems. Summary of the Invention

[0005] Based on the above analysis, the embodiments of the present invention aim to provide a method for evaluating the equivalence of scaled-down tests of unmanned systems, in order to solve the technical problem that existing evaluation methods are unable to scientifically quantify the equivalence between scaled-down tests and prototype tests.

[0006] This invention provides a method for evaluating the equivalence of scaled-down tests of unmanned systems, comprising the following steps:

[0007] A hierarchical evaluation index system for unmanned system testing is constructed; wherein, the hierarchical evaluation index system includes multiple top-level indicators and bottom-level indicators corresponding to each top-level indicator;

[0008] Small-scale scale-down experiments and large-scale prototype experiments were conducted to collect raw data corresponding to each underlying index. The raw data were then quantified to obtain the standardized values ​​of each underlying index.

[0009] A capability value calculation model is constructed based on a linear weighted model, which is used to calculate the capability value of each top-level indicator based on the standardized values ​​of each underlying indicator.

[0010] The combined weighting method is used to calculate the combined weights of each underlying indicator; based on the standardized values ​​and the combined weights, the capability value calculation model is used to calculate the capability evaluation values ​​of the small-scale scale-down test and the large-scale prototype test on each top-level indicator.

[0011] The degree of equivalence between the two tests is determined based on the deviation rate between the capability evaluation values ​​of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator.

[0012] Furthermore, the top-level indicators include top-level indicators for situational awareness, analysis, formation control, and target acquisition.

[0013] The situational awareness capability includes underlying metrics for target tracking error and sensor data fusion quality.

[0014] The analytical capabilities include underlying metrics such as data transmission latency and target recognition accuracy.

[0015] The formation control capability includes formation task completion rate and underlying indicators of formation stability;

[0016] The target capture capability includes underlying indicators such as regional exploration repetition rate, capture success rate, and capture efficiency.

[0017] Furthermore, the raw data undergoes quantification processing, including:

[0018] For raw data that can be directly measured for the corresponding underlying indicators, the ratio method is used to normalize the data to a quantized value between [0,1], which is then used as the standardized value of the underlying indicator.

[0019] For qualitative raw data that cannot be directly measured for the corresponding underlying indicators, quantitative values ​​between [0,1] are obtained through expert scoring and used as standardized values ​​for the underlying indicators.

[0020] Furthermore, the capability value calculation model is as follows:

[0021]

[0022] Among them, G a θ represents the capability value of the top-level indicator 'a', where the subscript 'a' is the index of the top-level indicator. i r is the combined weight of the i-th underlying indicator. i Let be the standardized value of the i-th bottom-level indicator, and n be the total number of bottom-level indicators contained in the top-level indicator.

[0023] Furthermore, the calculation of the combined weights of each underlying indicator using the combined weighting method includes:

[0024] The subjective weights of each bottom-level indicator are calculated based on the analytic hierarchy process (AHP).

[0025] The objective weights of each underlying indicator are calculated based on the CRITIC method;

[0026] Based on the principle of least squares, the subjective weights and objective weights are combined to calculate the combined weights of each underlying indicator.

[0027] Furthermore, the subjective weights of each underlying indicator calculated based on the analytic hierarchy process include:

[0028] The importance of indicators at the same level is compared pairwise using the 1-9 scaling method to construct the first judgment matrix V;

[0029] A consistency check is performed on the first judgment matrix V, and a second judgment matrix V' is obtained after the consistency ratio meets the preset consistency ratio threshold.

[0030] The elements of the second judgment matrix V' are column normalized to obtain the third judgment matrix V”;

[0031] After summing the rows of the third judgment matrix V”, normalization is performed to obtain the final subjective weight vector ω. i .

[0032] Furthermore, the calculation of the objective weights of each underlying indicator based on the CRITIC method includes:

[0033] The standardized values ​​of the underlying indicators are made dimensionless to distinguish between extremely large and extremely small indicators.

[0034] Calculate the extremely large and extremely small indicators, and calculate the standard deviation S of the indicators. j To characterize its variability;

[0035] Calculate the Pearson correlation coefficients among the indicators, and then calculate the conflict coefficients (T) for each indicator based on these coefficients. j ;

[0036] The information content C of each indicator is calculated based on the product of the variability and the conflict. j ;

[0037] The information is normalized to obtain the objective weights of each underlying indicator;

[0038] Among them, the underlying indicator whose attribute is defined as an extremely large indicator is based on the principle that the larger the value, the better; the underlying indicator whose attribute is defined as an extremely small indicator is based on the principle that the smaller the value, the better.

[0039] Furthermore, the combined weights of each underlying indicator are calculated, including:

[0040] For each underlying indicator corresponding to the top-level indicator, a deviation function is constructed based on the least squares principle, as follows:

[0041] h(θ i )=(w i -θ i ) 2 +(μ i -θ i ) 2

[0042] Differentiate the deviation function, set the derivative to 0, find the extreme points, and solve for the optimal combined weight θ. i ;

[0043] The optimal combination weight θ of each underlying indicator corresponding to the top-level indicator. i This yields the optimal combination weight vector θ = (θ1, θ2, ..., θ) of the underlying indicators corresponding to the top-level indicators. n ) T .

[0044] Furthermore, based on the capability evaluation values ​​of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator, the deviation rate between the corresponding large-scale scaled-down test and the large-scale prototype test is calculated based on the capability evaluation values ​​of each top-level indicator; wherein, the deviation rate includes the deviation rates of situational awareness capability, analysis capability, formation control capability, and target encirclement capability.

[0045] If the deviation rate is ≤ n1%, then the ability to evaluate the top-level metrics of small-scale scale-down tests and corresponding large-scale prototype tests is highly equivalent.

[0046] If n1% < deviation rate ≤ n2%, then the ability to evaluate the top-level indicators of small-scale scale-down tests and corresponding large-scale prototype tests is considered to be of medium equivalence.

[0047] If n2% < deviation rate ≤ n3%, then the ability to evaluate the top-level metrics of small-scale scale-down tests and corresponding large-scale prototype tests is poorly equivalent.

[0048] If the deviation rate is > n3%, then the ability to evaluate the top-level metrics of small-scale scale-down experiments and corresponding large-scale prototype experiments is not equivalent.

[0049] Where n1%, n2%, and n3% are the first, second, and third deviation rate thresholds, respectively, and 0 < n1 < n2 < n3;

[0050] If the deviation rates of situational awareness, analysis, formation control, and target acquisition capabilities are highly or moderately equivalent, then the overall results of this small-scale scaled-down test can effectively replace the results of the corresponding large-scale prototype test and be used for algorithm optimization and strategy verification of unmanned systems.

[0051] The present invention also discloses an equivalence evaluation system for scaled-down unmanned system tests. The system includes an evaluation index system construction module M1, an experimental data acquisition and standardization module M2, a capability calculation model construction module M3, a combined weight calculation and synthesis module M4, and an equivalence determination module M5.

[0052] The evaluation index system construction module M1 is used to construct a hierarchical evaluation index system for unmanned system experiments; wherein, the hierarchical evaluation index system includes multiple top-level indicators and bottom-level indicators corresponding to each top-level indicator.

[0053] The experimental data acquisition and standardization module M2 is used to conduct small-scale scale-down experiments and large-scale prototype experiments respectively, collect raw data corresponding to each underlying index, and quantify the raw data to obtain the standardized values ​​of each underlying index.

[0054] The capability calculation model construction module M3 is used to construct a capability value calculation model based on a linear weighted model, and to calculate the capability value of each top-level indicator based on the standardized value of each bottom-level indicator.

[0055] The combined weight calculation and synthesis module M4 is used to calculate the combined weight of each underlying indicator using the combined weighting method; based on the standardized value and the combined weight, the capability value calculation model is used to calculate the capability evaluation value of the small-scale scale-down test and the large-scale prototype test on each top-level indicator.

[0056] The equivalence determination module M5 is used to determine the degree of equivalence between the two tests based on the deviation rate between the capability evaluation values ​​of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator.

[0057] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0058] 1. This invention is the first to use the OODA loop theory as the core evaluation framework, deconstructing the complex collaborative capabilities of clusters into four logically clear stages: "Observation-Adjustment-Decision-Action," and establishing a hierarchical indicator system covering the entire task process based on this. The evaluation framework is more systematic, solving the problem of ambiguous evaluation dimensions. This makes the evaluation dimensions clearer and more systematic, ensuring the comprehensiveness and logic of the evaluation. It addresses the shortcomings of existing technologies for evaluating unmanned systems, especially cluster systems, which often suffer from scattered indicators and a lack of a unified theoretical framework.

[0059] 2. This invention innovatively proposes a combined weighting strategy of "AHP + CRITIC + Least Squares," which deeply integrates expert prior knowledge with the objective statistical characteristics of experimental data, and obtains the optimal balance point through mathematical optimization. This combined weighting effectively integrates subjective and objective information, significantly improving the scientificity and rationality of weight allocation, ensuring that the evaluation results are both consistent with domain knowledge and faithful to the data facts; the weight allocation is more scientific. It overcomes the shortcomings of existing technologies that often rely solely on subjective or objective weighting methods.

[0060] 3. This invention ultimately outputs quantified capability evaluation values, clear equivalence levels, and multi-dimensional spider diagram visualizations for comparison. This "numerical + level + graphical" output method makes the differences between small-scale scaled-down experiments and large-scale prototype experiments across various capability dimensions immediately apparent. This not only facilitates intuitive and clear comparative analysis but also provides R&D teams with precise decision-making basis for targeted algorithm optimization and iterative experimental scheme development, greatly enhancing the method's practical value. It addresses the shortcomings of existing technologies where evaluation results are often isolated scores or conclusions, hindering in-depth analysis and decision-making.

[0061] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0062] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0063] Figure 1 This is a flowchart of a method for evaluating the equivalence of a scaled-down test of an unmanned system according to an embodiment of the present invention;

[0064] Figure 2 This is a schematic diagram of a multi-dimensional spider web of the top-level indicator capability values ​​obtained according to an embodiment of the present invention.

[0065] Figure 3 This is a schematic diagram of the functional modules of an unmanned system scaled-down test equivalence evaluation system in an embodiment of the present invention. Detailed Implementation

[0066] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0067] Example 1:

[0068] The search phase of an unmanned swarm is highly consistent with the inherent logic of the OODA loop (Observe-Orient-Decide-Act). The unmanned swarm observes its environment through sensors, fuses and judges information to form situational awareness, conducts collaborative planning to make decisions, and finally, the control platform executes specific actions. The OODA loop provides an ideal theoretical framework for deconstructing and evaluating the complex collaborative capabilities of unmanned swarms. Based on the OODA framework, a hierarchical evaluation index system is constructed, including a top-level capability layer and bottom-level indicators. This system covers a comprehensive measurement from bottom-level technical parameters to top-level search capabilities, laying the foundation for equivalence quantitative evaluation.

[0069] This invention first uses the OODA loop as its core framework to systematically construct a multi-level evaluation index system encompassing dimensions such as perception, analysis, decision-making, and action. To overcome the limitations of single weighting methods, a combined weighting algorithm integrating subjective analytic hierarchy process (AHP) with objective CRITIC method is employed to scientifically determine the weights of each underlying index, forming a complementary logic of empirical judgment and data verification. Finally, this evaluation framework is used to conduct equivalence evaluations of experiments of different scales. This method provides a systematic solution for validating the results of cross-scale experiments.

[0070] A specific embodiment of the present invention discloses a method for evaluating the equivalence of scaled-down tests of unmanned systems, such as... Figure 1 As shown, it includes the following steps:

[0071] Step S1: Construct a hierarchical evaluation index system for unmanned system testing; wherein, the hierarchical evaluation index system includes multiple top-level indicators and corresponding bottom-level indicators for each top-level indicator;

[0072] Step S2: Conduct small-scale scale-down tests and large-scale prototype tests respectively, collect raw data corresponding to each underlying index, and quantify the raw data to obtain the standardized values ​​of each underlying index.

[0073] Step S3: Construct a capability value calculation model based on a linear weighted model, which is used to calculate the capability value of each top-level indicator based on the standardized values ​​of each underlying indicator;

[0074] Step S4: Calculate the combined weights of each underlying indicator using the combined weighting method; based on the standardized values ​​and the combined weights, use the capability value calculation model to calculate the capability evaluation values ​​of the small-scale scale-down test and the large-scale prototype test on each top-level indicator.

[0075] Step S5: Based on the deviation rate between the capability evaluation values ​​of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator, determine the degree of equivalence between the two tests.

[0076] Step S1, specifically.

[0077] A hierarchical evaluation index system for scaled-down experiments of unmanned systems is constructed. Based on the OODA loop theory, the collaborative process of unmanned swarms executing encirclement tasks is decomposed into four stages: observation, adjustment, decision-making, and action. Using this framework, an evaluation system comprising four top-level capability indicators is constructed:

[0078] (1) Observe corresponds to situational awareness capability;

[0079] (2) Adjust (Orient) corresponding analysis capabilities;

[0080] (3) Decision-making corresponds to formation control capability;

[0081] (4) Action (Act) corresponds to the ability to capture targets.

[0082] Based on the OODA (Observe-Judgment-Decision-Action) cycle theory, the task process of unmanned systems is decomposed into four stages: observation, adjustment, decision-making, and action. Top-level indicators are then constructed for situational awareness, analysis, formation control, and target acquisition capabilities, respectively. Under each top-level indicator, corresponding bottom-level indicators are established step by step to form a hierarchical equivalence evaluation indicator system.

[0083] The top-level indicators include situational awareness, analysis, formation control, and target acquisition capabilities.

[0084] The situational awareness capability includes underlying metrics for target tracking error and sensor data fusion quality.

[0085] The analytical capabilities include underlying metrics such as data transmission latency and target recognition accuracy.

[0086] The formation control capability includes formation task completion rate and underlying indicators of formation stability;

[0087] The target capture capability includes underlying indicators such as regional exploration repetition rate, capture success rate, and capture efficiency.

[0088] The top-level indicators are further decomposed into lower-level indicators such as target tracking error, data transmission latency, formation task completion rate, and area exploration duplication rate, forming a complete hierarchical evaluation indicator system. Lower-level indicator attributes where larger values ​​are better are defined as "extremely large indicators," and lower-value lower-value lower-value lower-value indicators are defined as "extremely small indicators." The hierarchical evaluation indicators are shown in Table 1.

[0089] Table 1: Hierarchical Evaluation Metrics for Unmanned Systems Based on the OODA Cycle

[0090]

[0091] Step S1 aims to construct a hierarchical evaluation index system for unmanned system experiments based on the OODA loop theory. It clarifies the top-level capability indicators and corresponding bottom-level indicators for situational awareness, analysis, formation control, and target acquisition capabilities. Simultaneously, it defines the attributes of the bottom-level indicators, providing a clear framework for subsequent data quantification and weight calculation for equivalence evaluation. This addresses the technical problem of ambiguous evaluation dimensions in existing methods.

[0092] Step S2 includes steps S21-S22.

[0093] Step S21: Conduct small-scale scale-down tests and large-scale prototype tests respectively, and collect the original data corresponding to each of the underlying indicators.

[0094] Small-scale scale-down experiments and large-scale prototype experiments were conducted to collect raw data corresponding to each bottom-level indicator in the hierarchical evaluation indicator system. The collected raw data underwent quantitative processing.

[0095] For directly measurable indicators, normalization is performed using the ratio method;

[0096] For qualitative indicators that cannot be directly measured, quantitative values ​​are obtained through expert scoring.

[0097] In this example, the ratio of the feature length of a large-scale prototype experimental scenario to that of a small-scale scaled-down experimental scenario is defined as the scaling factor λ, as shown below:

[0098]

[0099] Based on the length scaling factor λ, and assuming that the physical processes in the two scenarios remain similar, the scaling relationships of other key physical quantities are defined according to the Froude similarity criterion, as follows:

[0100] The physical quantity is represented by a distance / length scaling factor λ;

[0101] The physical quantity is proportional to time.

[0102] The physical quantity represented by velocity is a scaled relationship as follows:

[0103] The physical quantity represented by the area is λ. 2 .

[0104] For example, λ = 10 is taken; in practical applications, it can be changed according to specific needs.

[0105] In this example, the small-scale scale-down experiment and the large-scale prototype experiment scenarios are as follows:

[0106] Small-scale scaled-down experiment: conducted in a 1×1 km lake environment, deploying a swarm of 10 drones and 4 unmanned boats to collaboratively search for a target unmanned boat;

[0107] Large-scale prototype test: conducted in a 10×10 km marine environment, a swarm of 10 drones and 4 unmanned vessels worked together to search for an unmanned vessel target.

[0108] The aerial drone swarm is responsible for reconnaissance, identification, and communication relay tasks, while the unmanned surface vessel (USV) swarm is responsible for collaboratively searching for and capturing the target. After the mission begins, the drones depart from the base and use search algorithms (e.g., distributed search algorithms based on dynamic collaborative search domains) to detect the target vessel. Once the drone swarm identifies a potential target, the USV departs from the base to collaborate with the drone swarm in capturing the target.

[0109] The raw experimental data came from real-time data transmitted back to the central control system by the GPS, electro-optical radar, communication modules, and control systems onboard the UAV and unmanned surface vessel. The acquired raw experimental data is shown in Table 2.

[0110] Table 2: Raw Data Acquired by the Unmanned System

[0111]

[0112] Step S22: Quantify the collected raw data to obtain the standardized values ​​of each underlying indicator.

[0113] Raw data corresponding to each underlying indicator were collected from small-scale and large-scale experiments, and the raw data were quantified.

[0114] The raw data is quantized, including:

[0115] For raw data that can be directly measured for the corresponding underlying indicators, the ratio method is used to normalize the data to a quantized value between [0,1], which is then used as the standardized value of the underlying indicator.

[0116] For qualitative raw data that cannot be directly measured for the corresponding underlying indicators, quantitative values ​​between [0,1] are obtained through expert scoring and used as standardized values ​​for the underlying indicators.

[0117] For measurable values, a ratio method is used to unify the numerical values, normalizing them to the range [0,1]. Examples include detection rate, threat target identification accuracy, and area exploration rate, as shown below:

[0118]

[0119] Where P1 is the standardized value of the measurable value; x is the actual measured value; x mThe values ​​are standard values, total values, or expected values ​​(specific equipment performance indicators can be obtained from the technical manual).

[0120] P1 is the standardized value of the measurable indicator, a value between 0 and 1, representing the degree to which the actual performance of the indicator is achieved relative to the ideal value or the full score; the closer it is to 1, the better the performance.

[0121] x represents the raw data obtained directly from sensors, GPS, communication logs, etc., during the experiment. Examples include: the actual number of targets detected, the actual number of tasks completed, and the actual time consumed.

[0122] x m The preset maximum or upper limit reference value is usually derived from:

[0123] Technical manual: Best performance specifications of the equipment (e.g., maximum detection range of the sensor);

[0124] Task definition: The theoretically achievable perfect value (e.g., the target ship being searched is 1, and the accuracy in identification is x). m =1);

[0125] Physical constraints: The upper limit of scaling is calculated based on the similarity criterion (e.g., if the upper limit of the task time in the prototype scene is 30 minutes, then the scaled-down scene is 9.48 minutes).

[0126] For values ​​that cannot be measured, an expert scoring principle is adopted based on experiments, with a score range of [0,1]. For values ​​such as the stability of encirclement and tracking, and the effectiveness of decision-making, experts can give a score of 0 to 1 based on experimental results and their own experience. The values ​​are then quantified using the following formula:

[0127]

[0128] Where P2 is the standardized value of the unmeasurable value, S j Let k be the score given by the j-th expert; k is the number of experts who participated in the scoring.

[0129] The experimental data comes from the actual test results of the special test and expert evaluation, and covers all the underlying indicators of the evaluation index system.

[0130] In this example, the underlying metrics that cannot be measured include sensor data fusion quality and formation stability, whose values ​​are given by expert ratings. The remaining underlying metrics are measurable values.

[0131] The calculation method for the measurable underlying indicators given in Table 1 is as follows:

[0132] (1) Situational awareness capability

[0133] The target tracking error is calculated as follows:

[0134]

[0135] Among them, e 探测 Target tracking error, also known as radial error or Euclidean distance error, represents the actual straight-line distance in a two-dimensional plane between the target position reported by the sensor and the target's true position. The smaller this value, the higher the sensor's detection and tracking accuracy. 误差 y 误差 These are the position errors of the target in the x and y coordinate axes, respectively, which are the differences between the target's x and y coordinate values ​​measured by the sensor and its true x and y coordinate values; if the upper limit of the target tracking error in the prototype scene is 2m, then the upper limit of the target tracking error in the scaled-down scene is 0.2m.

[0136] (2) Analytical ability

[0137] Data transmission delay is calculated as follows: the time difference between when data is sent and when it is received. For the prototype scenario, the upper limit of the error is 1 second, and for the scaled-down scenario, the upper limit of the error is 0.32 seconds.

[0138] Target recognition accuracy is calculated as follows:

[0139]

[0140] (3) Formation control capability

[0141] The formation mission completion rate is calculated as follows:

[0142]

[0143] (4) Target capture capability

[0144] The region exploration repetition rate is calculated as follows:

[0145]

[0146] (5) Target capture capability

[0147] The success rate of the encirclement and capture is calculated as follows:

[0148]

[0149] Encirclement efficiency is calculated as follows:

[0150]

[0151] Experimental data were collected in two scenarios: small-scale scale-down experiments and large-scale prototype experiments. Five sets of valid raw test data were obtained for each scenario. The raw data obtained are shown in Table 3.

[0152] Table 3: Examples of the original data obtained

[0153]

[0154]

[0155] The standardized values ​​obtained by quantizing the original data are shown in Table 4.

[0156] Table 4: Standardized values ​​of underlying indicators after quantification of raw data

[0157]

[0158] Step 2 aims to collect experimental measured values ​​and expert evaluation results of the underlying indicators, clarify the measurable attributes of the indicators (directly measurable / indirectly measurable), quantify the two types of indicators respectively, form standardized data, and provide standardized and usable input data for subsequent evaluation stages.

[0159] Step S3, specifically.

[0160] Establish a capability value calculation model. Based on the relationship characteristics between indicators, establish a quantitative calculation model for the capability of each top-level indicator based on a linear weighted model. The top-level indicator is obtained by weighted summation of its corresponding lower-level indicators, and the quantitative indicator of the top-level capability is obtained by weighted summation of the lower-level indicators.

[0161] The capability value calculation model is as follows:

[0162]

[0163] Among them, G a The value represents the capability of the top-level indicator, with subscript 'a' indicating the index of the top-level indicator, and θ representing the capability value of the top-level indicator. i r is the combined weight of the i-th underlying indicator. i Let be the standardized value of the i-th bottom-level indicator, and n be the total number of bottom-level indicators contained in the top-level indicator.

[0164] G a , a = 1, 2, 3, 4 are the top-level indicator capability values, ∑θ i =1. When the underlying indicator is an extremely large indicator, r i =P i When the indicator is a very small indicator, r i =1-P i P i This refers to the standardized value of the underlying indicator itself.

[0165] The standardized values ​​of the underlying indicators are weighted to obtain the quantitative indicator value of the top-level capability.

[0166] Step S3 serves to aggregate the standardized values ​​of the underlying indicators into a quantitative evaluation value of the top-level capabilities through a linear weighted model, providing core data support for equivalence assessment.

[0167] Step S4 includes steps S41-S42.

[0168] Step S41: Calculate the combined weights of each underlying indicator.

[0169] The method of calculating the combined weights of each underlying indicator using a combined weighting approach includes:

[0170] The subjective weights of each bottom-level indicator are calculated based on the analytic hierarchy process (AHP).

[0171] The objective weights of each underlying indicator are calculated based on the CRITIC method;

[0172] Based on the principle of least squares, the subjective weights and objective weights are combined to calculate the combined weights of each underlying indicator.

[0173] The combined weighting method is used to calculate the combined weights of the indicators at each level, including:

[0174] (1) Calculating subjective weights based on the Analytic Hierarchy Process (AHP): Domain experts were invited to conduct pairwise comparisons of indicators at the same level in the indicator system, constructing a judgment matrix. After passing the consistency test, the subjective weights w of each bottom-level indicator were calculated. i .

[0175] (2) Calculate objective weights based on the CRITIC method: Using the quantified experimental data from step two, calculate the variability (standard deviation) and conflict (correlation) of each underlying indicator, and then calculate the objective weight μ of each underlying indicator. i .

[0176] (3) Calculate the combined weights: Based on the principle of least squares, the subjective weights w i With objective weight μ i By combining the results, we obtain the final combined weight θ. i .

[0177] Step S41 includes steps S411-S413.

[0178] Step S411: Calculate the subjective weights of each underlying indicator based on the analytic hierarchy process.

[0179] The subjective weights for calculating each bottom-level indicator based on the analytic hierarchy process include:

[0180] The importance of indicators at the same level is compared pairwise using the 1-9 scaling method to construct the first judgment matrix V;

[0181] A consistency check is performed on the first judgment matrix V, and a second judgment matrix V' is obtained after the consistency ratio meets the preset consistency ratio threshold.

[0182] The elements of the second judgment matrix V' are column normalized to obtain the third judgment matrix V”;

[0183] After summing the rows of the third judgment matrix V”, normalization is performed to obtain the final subjective weight vector ω. i .

[0184] Calculate the subjective weight ω. Invite 3-5 domain experts to conduct pairwise importance comparisons of all lower-level indicators under the same top-level indicator, based on the 1-9 scaling method shown in Table 5, and construct the judgment matrix V′.

[0185] Table 5: Importance Scale Table

[0186]

[0187] Define the underlying metric for which weights need to be calculated as V. i and V j The comparison value is v ij Construct the first judgment matrix V, with matrix elements v ij As shown below:

[0188]

[0189] Among them, v ii =1, v ij >0.

[0190] Since subjective judgments may contain logical contradictions, a consistency check is needed on the first judgment matrix V. The consistency index is calculated as follows:

[0191]

[0192] Where CI is the consistency index, and λ max Let m be the largest eigenvalue of the first judgment matrix, and m be the order of the first judgment matrix.

[0193] The consistency ratio is calculated as follows:

[0194]

[0195] Wherein, CR is the consistency ratio, and RI is the random consistency index, the value of which is determined by the order of the first judgment matrix V.

[0196] Once the consistency ratio meets the preset consistency ratio threshold, the second judgment matrix V' is obtained, with matrix elements v′.ij For example, the preset consistency ratio threshold is 0.1; in specific applications, it can be changed according to specific needs.

[0197] If CR≤0.1, the consistency of the first judgment matrix is ​​acceptable; otherwise, the first judgment matrix V needs to be adjusted until the consistency of the first judgment matrix V is acceptable.

[0198] This study uses the sum-product method to calculate the weight vector of the judgment matrix that passes the consistency test. The elements of the second judgment matrix V′ are column-normalized as follows:

[0199]

[0200] Normalize the columns of the second judgment matrix V' to obtain the third judgment matrix V”, whose elements are v″. ij .

[0201] The third judgment matrix V” is summed row by row, as shown below:

[0202]

[0203] The third judgment matrix V” is summed row-wise and then normalized to obtain the final weight vector ω. i As shown below:

[0204]

[0205] Multiply the single ranking weights calculated at each level to obtain the combined weight of the lowest-level indicator relative to the capability layer, so that it can be combined with the objective weighting results in the future.

[0206] By constructing first, second, and third judgment matrices, the weights of each underlying indicator are calculated step by step. To ensure the accuracy of the evaluation, experts are requested to evaluate each indicator according to the importance scaling table, and the third judgment matrix among the indicators is obtained by summarizing the results, as shown in Table 6-9.

[0207] Table 6: The Third Judgment Matrix of the Bottom-Level Indicators Corresponding to the Top-Level Indicator Situational Awareness Capability

[0208]

[0209] Table 7: The Third Judgment Matrix of the Underlying Indicators Corresponding to the Top-Level Indicator Analysis Capability

[0210]

[0211] Table 8: Third Judgment Matrix of Top-Level Indicator Formation Control Capability Corresponding to Bottom-Level Indicators

[0212]

[0213] Table 9: Third Judgment Matrix of Top-Level Indicator Target Encirclement Capability Corresponding to Bottom-Level Indicators

[0214]

[0215] The weights of the bottom-level indicators can be obtained by calculating using the analytic hierarchy process, as shown in Table 10.

[0216] Table 10: Weights of the bottom-level indicators for each indicator in the Analytic Hierarchy Process (AHP)

[0217]

[0218]

[0219] Step S412: Calculate the objective weights of each underlying indicator based on the CRITIC method.

[0220] The objective weights for each underlying indicator calculated using the CRITIC method include:

[0221] The standardized values ​​of the underlying indicators are made dimensionless to distinguish between extremely large and extremely small indicators.

[0222] Calculate the extremely large and extremely small indicators, and calculate the standard deviation S of the indicators. j To characterize its variability;

[0223] Calculate the Pearson correlation coefficients among the indicators, and then calculate the conflict coefficients (T) for each indicator based on these coefficients. j ;

[0224] The information content C of each indicator is calculated based on the product of the variability and the conflict. j ;

[0225] The information is normalized to obtain the objective weights of each underlying indicator;

[0226] Among them, the underlying indicator whose attribute is defined as an extremely large indicator is based on the principle that the larger the value, the better; the underlying indicator whose attribute is defined as an extremely small indicator is based on the principle that the smaller the value, the better.

[0227] Calculate the CRITIC objective weight μ. Use the data matrix of standardized values ​​from all the underlying indicators quantized in step S2 as input. First, perform dimensionless processing on the data (distinguishing between extremely large and extremely small indicators). To eliminate the influence of different indicator dimensions, standardize the data matrix.

[0228] For extremely large metrics where larger values ​​are better, see the following:

[0229]

[0230] Where K1 is the dimensionless value of the largest indicator in the underlying indicators; x i The actual value of the underlying indicator; x max This is the maximum value of this indicator; x min This is the minimum value of this indicator.

[0231] For extremely small metrics where smaller values ​​are better, see the following:

[0232]

[0233] Among them, the K2 minima index is a dimensionless value.

[0234] The variability of the indicator is expressed as standard deviation S. j The larger the standard deviation, the greater the data volatility, and the more important this indicator is in the evaluation. Variability is expressed as standard deviation S. j The calculation is as follows:

[0235]

[0236] Among them, S j The standard deviation, or variability, of the standardized value corresponding to the i-th underlying index; x ij The actual value of the j-th underlying index in the i-th test data; is the arithmetic mean of the underlying index across all experimental data; n is the number of trials.

[0237] The conflict of an indicator is reflected by its correlation coefficient with all other indicators. The conflict of the underlying indicator j (T) is shown below. j As shown below:

[0238]

[0239] Where, r ij This is the Pearson correlation coefficient between underlying indicators i and j, where n is the total number of indicators. The stronger the correlation between a given underlying indicator and other indicators (r...), the higher the correlation coefficient. ij The larger the absolute value, the lower the conflict and the less independent information it contains, and the lower the weight should be assigned.

[0240] Information content C of indicator j j The product of its variability and conflict is as follows:

[0241]

[0242] By normalizing the information content of each indicator, the final objective weights can be obtained, as shown below:

[0243]

[0244] Where, μ j Let be the objective weight of the j-th indicator.

[0245] Based on the standardized values ​​of each underlying indicator obtained in step S2, the total weight of each underlying indicator can be calculated, as shown in Table 11.

[0246] Table 11: Total Weights of the CRITIC Weighting Method

[0247]

[0248] Step S413: Calculate the combined weights of each level of indicators using the combined weighting method.

[0249] Calculate the combined weights of each underlying indicator, including:

[0250] For each underlying indicator corresponding to the top-level indicator, a deviation function is constructed based on the least squares principle, as follows:

[0251] h(θ i )=(w i -θ i ) 2 +(μ i -θ i ) 2

[0252] Differentiate the deviation function, set the derivative to 0, find the extreme points, and solve for the optimal combined weight θ. i ;

[0253] The optimal combination weight θ of each underlying indicator corresponding to the top-level indicator. i This yields the optimal combination weight vector θ = (θ1, θ2, ..., θ) of the underlying indicators corresponding to the top-level indicators. n ) T .

[0254] Calculate the combined weights θ. To minimize the bias between subjective and objective weighting, a combined weighting model based on the least squares method is adopted. The weight vectors obtained by the AHP and CRITIC methods are ω=(ω1,ω2,…,ω…). m ) and μ=(μ1,μ2,…,μ m The combined weights are θ = (θ1, θ2, ..., θ). m ).

[0255] Based on the fundamental principles of the least squares method, a deviation function is constructed, and the weights for the combined weighting are set to θ. i As shown below:

[0256] h(θ i )=(w i -θ i) 2 +(μ i -θ i ) 2 Formula (23)

[0257] To achieve h(θ) in least squares i The minimum value for h(θ) i Take the derivative, as shown below:

[0258] h′(θ i )=4θ i -2w i -2μ i Formula (24)

[0259] Let h′(θ) i Let the derivative be 0 to find the extreme point, and solve for the optimal combination weight θ. i As shown below:

[0260]

[0261] Therefore, the weight θ obtained by the least squares combination weighting method can be expressed as:

[0262] θ=(θ1,θ2,...,θ n ) T Formula (26)

[0263] Where, ω i Let ∑ω be the weight vector of the analytic hierarchy process. i =1; μ i Assign weights to the CRITIC weighting method, ∑μ i =1; θ is the combined weight vector, ∑θ i =1.

[0264] Based on the combined weighting function, the weights obtained by the analytic hierarchy process (AHP) and the weights obtained by the CRITIC weighting method are combined, and the total weights of each bottom-level index can be obtained through calculation, as shown in Table 12.

[0265] Table 12: Total weight of each underlying indicator

[0266]

[0267] Step S42: Based on the standardized value and the combined weight, use the capability value calculation model to calculate the capability evaluation values ​​of the small-scale scale-down test and the large-scale prototype test on each top-level indicator.

[0268] The standardized value p of each underlying index obtained in step S2 i and the weight θ obtained in step S4 iSubstitute the values ​​into the capability value calculation model of formula (10) in step S3 to calculate the capability evaluation values ​​of small-scale scale-down tests and large-scale prototype tests on each top-level index.

[0269] For small-scale scaled-down experiments, the capability evaluation values ​​of four top-level indicators—situational awareness, analysis, formation control, and target acquisition—are calculated respectively.

[0270] For large-scale prototype tests, the capability evaluation values ​​of four top-level indicators—situational awareness, analysis, formation control, and target acquisition—are calculated respectively.

[0271] Step 4 is to first calculate subjective weights using the AHP method combined with expert experience, then calculate objective weights using the CRITIC method based on experimental data, and finally combine the two types of weights using the least squares method to obtain combined weights. This avoids the limitations of single weighting and allows the weights to have both expert experience and data objectivity, providing a reasonable weight basis for the weighting of indicators and accurate calculation of the top-level capability value in the capability value calculation model in step S3.

[0272] Step S5, specifically.

[0273] Based on the capability evaluation values ​​of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator, the deviation rate between the corresponding large-scale scaled-down test and the large-scale prototype test is calculated based on the capability evaluation values ​​of each top-level indicator; wherein, the deviation rate includes the deviation rate of situational awareness capability, analysis capability, formation control capability and target encirclement capability.

[0274] If the deviation rate is ≤ n1%, then the ability to evaluate the top-level metrics of small-scale scale-down tests and corresponding large-scale prototype tests is highly equivalent.

[0275] If n1% < deviation rate ≤ n2%, then the ability to evaluate the top-level indicators of small-scale scale-down tests and corresponding large-scale prototype tests is considered to be of medium equivalence.

[0276] If n2% < deviation rate ≤ n3%, then the ability to evaluate the top-level metrics of small-scale scale-down tests and corresponding large-scale prototype tests is poorly equivalent.

[0277] If the deviation rate is > n3%, then the ability to evaluate the top-level metrics of small-scale scale-down experiments and corresponding large-scale prototype experiments is not equivalent.

[0278] Where n1%, n2%, and n3% are the first, second, and third deviation rate thresholds, respectively, and 0 < n1 < n2 < n3;

[0279] If the deviation rates of situational awareness, analysis, formation control, and target acquisition capabilities are highly or moderately equivalent, then the overall results of this small-scale scaled-down test can effectively replace the results of the corresponding large-scale prototype test and be used for algorithm optimization and strategy verification of unmanned systems.

[0280] For example, the first, second, and third deviation rate thresholds are 5%, 10%, and 15%, respectively.

[0281] By comparing the capability evaluation values ​​of the small-scale scaled-down experiment and the large-scale prototype experiment on each top-level indicator, and determining their equivalence based on preset first, second, and third deviation rate thresholds, the capability evaluation values ​​are plotted as a multi-dimensional spider diagram, such as... Figure 2 As shown, a visual and intuitive comparative analysis is performed.

[0282] The deviation rate evaluation value of the two groups of experiments on a certain top-level capability is calculated as follows:

[0283]

[0284] Thus, four deviation rate values ​​were obtained for the top-level indicators: situational awareness, analysis, formation control, and target acquisition.

[0285] Plot the capability values ​​corresponding to the four top-level indicators on a multi-dimensional spider diagram, such as... Figure 2 As shown, by comparing the shapes and areas of the two polygons, the advantages and disadvantages of the two experimental modes in terms of various top-level capabilities and the degree of overall equivalence can be intuitively analyzed.

[0286] The calculation results can be used to scientifically determine the capability matching degree between small-scale scaled-down tests and large-scale prototype tests, and clarify whether the results of scaled-down tests can replace prototype tests.

[0287] If the results are of high / medium equivalence, algorithms and verification strategies can be optimized based on low-cost scaled-down experiments, reducing the number of high-cost prototype tests. Simultaneously, this result provides a basis for decision-making in the development of unmanned systems.

[0288] If low equivalence or inequivalence occurs, the root causes of the differences can be investigated in a targeted manner (such as environmental simulation or platform performance matching issues) to guide the iterative improvement of the scaled-down test scheme and ensure its effective guiding value for the optimization of prototype system design.

[0289] The standardized values ​​of the underlying indicators from five tests under each working condition were quantified and used for calculation. Based on the capability scores of the underlying indicators in the sample data, and according to the capability layer evaluation model, the top-level capability value and deviation rate were obtained by weighted summation, as shown in Table 13.

[0290] Table 13: Results of Top-Level Indicator Capability Values ​​and Deviation Rates

[0291]

[0292]

[0293] Based on the calculation results of the top-level capability values ​​above, a spider diagram is drawn, such as... Figure 2 As shown in the figure. The results show that both experiments achieve medium to high equivalence in the four top-level capabilities of the entire OODA loop, indicating that small-scale scaled-down experiments can serve as a low-cost alternative for the early verification and optimization of key capabilities of unmanned systems, significantly reducing the need for high-cost prototype experiments.

[0294] The purpose of step S5 is to classify the equivalence between small-scale scaled-down tests and large-scale prototype tests through quantitative calculation and visual comparison, so as to provide a reliable decision-making basis for the development of unmanned systems.

[0295] Through the above steps, a systematic and quantitative evaluation of the equivalence of scaled-down tests of unmanned systems is completed. This method is logically rigorous, comprehensively considers expert experience and objective data, and the evaluation results are scientifically reliable, providing important decision support for the research and development and testing of unmanned systems.

[0296] Example 2:

[0297] A specific embodiment of the present invention discloses an equivalence evaluation system for scaled-down unmanned system tests, thereby implementing the equivalence evaluation method for scaled-down unmanned system tests described in Embodiment 1. The specific implementation of each module is described in the corresponding section of Embodiment 1.

[0298] like Figure 3 As shown, an equivalence evaluation system for scaled-down unmanned system tests is provided. The system includes an evaluation index system construction module M1, a test data acquisition and standardization module M2, a capability calculation model construction module M3, a combined weight calculation and synthesis module M4, and an equivalence determination module M5.

[0299] The evaluation index system construction module M1 is used to construct a hierarchical evaluation index system for unmanned system experiments; wherein, the hierarchical evaluation index system includes multiple top-level indicators and bottom-level indicators corresponding to each top-level indicator.

[0300] The experimental data acquisition and standardization module M2 is used to conduct small-scale scale-down experiments and large-scale prototype experiments respectively, collect raw data corresponding to each underlying index, and quantify the raw data to obtain the standardized values ​​of each underlying index.

[0301] The capability calculation model construction module M3 is used to construct a capability value calculation model based on a linear weighted model, and to calculate the capability value of each top-level indicator based on the standardized value of each bottom-level indicator.

[0302] The combined weight calculation and synthesis module M4 is used to calculate the combined weight of each underlying indicator using the combined weighting method; based on the standardized value and the combined weight, the capability value calculation model is used to calculate the capability evaluation value of the small-scale scale-down test and the large-scale prototype test on each top-level indicator.

[0303] The equivalence determination module M5 is used to determine the degree of equivalence between the two tests based on the deviation rate between the capability evaluation values ​​of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator.

[0304] Since the system in this embodiment and the method in Embodiment 1 are related and can be referenced from each other, this description is redundant and will not be repeated here. Because this system embodiment shares the same principle as the above method embodiment, it also possesses the corresponding technical effects of the above method embodiment.

[0305] In summary, the unmanned system scaled-down test equivalence evaluation method and system of the present invention have the following beneficial effects:

[0306] 1. This invention is the first to use the OODA loop theory as the core evaluation framework, deconstructing the complex collaborative capabilities of clusters into four logically clear stages: "Observation-Adjustment-Decision-Action," and establishing a hierarchical indicator system covering the entire task process based on this. The evaluation framework is more systematic, solving the problem of ambiguous evaluation dimensions. This makes the evaluation dimensions clearer and more systematic, ensuring the comprehensiveness and logic of the evaluation. It addresses the shortcomings of existing technologies for evaluating unmanned systems, especially cluster systems, which often suffer from scattered indicators and a lack of a unified theoretical framework.

[0307] 2. This invention innovatively proposes a combined weighting strategy of "AHP + CRITIC + Least Squares," which deeply integrates expert prior knowledge with the objective statistical characteristics of experimental data, and obtains the optimal balance point through mathematical optimization. This combined weighting effectively integrates subjective and objective information, significantly improving the scientificity and rationality of weight allocation, ensuring that the evaluation results are both consistent with domain knowledge and faithful to the data facts; the weight allocation is more scientific. It overcomes the shortcomings of existing technologies that often rely solely on subjective or objective weighting methods.

[0308] 3. This invention ultimately outputs quantified capability evaluation values, clear equivalence levels, and multi-dimensional spider diagram visualizations for comparison. This "numerical + level + graphical" output method makes the differences between small-scale scaled-down experiments and large-scale prototype experiments across various capability dimensions immediately apparent. This not only facilitates intuitive and clear comparative analysis but also provides R&D teams with precise decision-making basis for targeted algorithm optimization and iterative experimental scheme development, greatly enhancing the method's practical value. It addresses the shortcomings of existing technologies where evaluation results are often isolated scores or conclusions, hindering in-depth analysis and decision-making.

[0309] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0310] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the equivalence of scaled-down tests of unmanned systems, characterized in that, The application relates to a method for evaluating unmanned system tests. The method comprises the following steps: constructing a hierarchical evaluation index system for unmanned system tests; wherein the hierarchical evaluation index system comprises multiple top-level indexes and corresponding bottom-level indexes of the top-level indexes; respectively carrying out small-scale scaled-down tests and large-scale prototype tests, collecting original data corresponding to the bottom-level indexes, and quantitatively processing the original data to obtain standardized values of the bottom-level indexes; constructing a capability value calculation model based on a linear weighting model, and calculating capability values of the top-level indexes according to the standardized values of the bottom-level indexes; calculating combination weights of the bottom-level indexes by using a combination weighting method; calculating capability evaluation values of the small-scale scaled-down tests and the large-scale prototype tests on the top-level indexes by using the capability value calculation model based on the standardized values and the combination weights; and judging the equivalence of the two kinds of tests based on a deviation rate between the capability evaluation values of the small-scale scaled-down tests and the large-scale prototype tests on the top-level indexes. The top-level indexes comprise situation awareness capability, analysis capability, formation control capability and target encirclement capability top-level indexes. The situation awareness capability comprises target tracking error and sensor data fusion quality bottom-level indexes. The analysis capability comprises data transmission delay and target recognition accuracy bottom-level indexes. The formation control capability comprises formation task completion rate and formation stability bottom-level indexes.

2. The method of claim 1, wherein, The target encirclement capability comprises area exploration repetition rate, encirclement success rate and encirclement efficiency bottom-level indexes. The quantitatively processing of the original data comprises the following steps: for original data of the bottom-level indexes that can be directly measured, a ratio method is used to normalize the original data to quantified values between 0 and 1, which are taken as the standardized values of the bottom-level indexes; and for qualitative original data of the bottom-level indexes that cannot be directly measured, an expert scoring method is used to obtain quantified values between 0 and 1, which are taken as the standardized values of the bottom-level indexes. The capability value calculation model is as follows: The combination weighting method comprises the following steps: calculating subjective weights of the bottom-level indexes based on an analytic hierarchy process; calculating objective weights of the bottom-level indexes based on a CRITIC method; and combining the subjective weights and the objective weights to calculate the combination weights of the bottom-level indexes based on a least square method principle. The calculation of the subjective weights of the bottom-level indexes based on the analytic hierarchy process comprises the following steps: comparing the importance of indexes at the same level in pairs based on a 1-9 scale method to construct a first judgment matrix V; performing column normalization on elements of the second judgment matrix V' to obtain a third judgment matrix V''; and calculating the subjective weights of the bottom-level indexes based on the third judgment matrix V''.

3. The method of claim 1, wherein, The calculation of the objective weights of the bottom-level indexes based on the CRITIC method comprises the following steps: performing dimensionless processing on the standardized values of the bottom-level indexes to distinguish between maximum indexes and minimum indexes; and performing normalization processing on the information amount to obtain the objective weights of the bottom-level indexes; wherein the attribute of a bottom-level index is defined as a maximum index based on the fact that a larger value is better, and the attribute of a bottom-level index is defined as a minimum index based on the fact that a smaller value is better. The calculation of the combination weights of the bottom-level indexes comprises the following steps: ​ 4. The unmanned system scale model test equivalency assessment method of claim 1, wherein, ​ wherein G a is the capability value of the top-level indicator a, subscript a is the serial number of the top-level indicator, θ i is the combined weight of the i-th bottom-level indicator, r i is the normalized value of the i-th bottom-level indicator, and n is the total number of bottom-level indicators included in the top-level indicator.

5. The unmanned system scale model test equivalency assessment method of claim 1, wherein, ​ ​ ​ ​ 6. The unmanned system scale test equivalency assessment method of claim 5, wherein, ​ ​ The first judgment matrix V is subjected to consistency check, and a second judgment matrix V is obtained after the consistency ratio meets a preset consistency ratio threshold ' ; ​ After summing up the third judgment matrix V" by row, normalization is performed to obtain the final subjective weight vector ω i .

7. The unmanned system scale test equivalency assessment method of claim 5, wherein, ​ ​ Calculate the maximum and minimum indicators, calculate the standard deviation S of the indicators j to characterize their variability; Pearson correlation coefficients between the indicators are calculated, and based on this, the conflictive T of each indicator is calculated j ; The information content C of each indicator is calculated according to the product of the variability and the conflictivity j ; ​ ​ 8. The unmanned system scale test equivalency assessment method of claim 7, wherein, ​ For each bottom-level indicator corresponding to a top-level indicator, a bias function is constructed based on the least square method principle, as follows: h(θ i ) = (w i - θ i ) 2 + (μ i - θ i ) 2 Derivate the bias function, find the extreme point by setting the derivative to 0, and solve to get the optimal combination weight θ i ; Based on each bottom-level index corresponding to the top-level index Optimal combination weight θ i , get the optimal combination weight vector θ of the bottom-level index corresponding to the top-level index θ=(θ1, θ2,..., θ n ) T .

9. The method of evaluating the scale model test equivalency of unmanned systems of any one of claims 1-8, wherein, Based on the capability evaluation values of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator, the bias rate between the small-scale scaled-down test and the large-scale prototype test is calculated based on the capability evaluation values on the top-level indicators; wherein the bias rate includes the situation awareness capability, the analysis capability, the formation control capability and the target encirclement capability bias rate; If the bias rate is ≤n1%, it is evaluated that the capabilities on the top-level indicators of the small-scale scaled-down test and the corresponding large-scale prototype test are highly equivalent; If n1% < the bias rate ≤n2%, it is evaluated that the capabilities on the top-level indicators of the small-scale scaled-down test and the corresponding large-scale prototype test are moderately equivalent; If n2% < the bias rate ≤n3%, it is evaluated that the capabilities on the top-level indicators of the small-scale scaled-down test and the corresponding large-scale prototype test are lowly equivalent; If the bias rate > n3%, it is evaluated that the capabilities on the top-level indicators of the small-scale scaled-down test and the corresponding large-scale prototype test are not equivalent; Wherein n1%, n2% and n3% are the first, second and third bias rate thresholds respectively, and 0 < n1 < n2 < n3; If the situation awareness capability, the analysis capability, the formation control capability and the target encirclement capability bias rate are highly equivalent or moderately equivalent, it is determined that the overall result of the small-scale scaled-down test can effectively replace the result of the corresponding large-scale prototype test, and is used for algorithm optimization and strategy verification of the unmanned system.

10. An unmanned system scale model test equivalency assessment system, comprising: The system comprises an evaluation index system construction module M1, a test data acquisition and standardization module M2, a capability calculation model construction module M3, a combined weight calculation and synthesis module M4 and an equivalence determination module M5; The evaluation index system construction module M1 is used to construct a hierarchical evaluation index system of the unmanned system test; wherein the hierarchical evaluation index system comprises a plurality of top-level indicators and bottom-level indicators corresponding to each top-level indicator; The test data acquisition and standardization module M2 is used to respectively carry out a small-scale scaled-down test and a large-scale prototype test, acquire original data corresponding to each bottom-level indicator, and quantitatively process the original data to obtain standardized values of each bottom-level indicator; The capability calculation model construction module M3 is used to construct a capability value calculation model based on a linear weighting model, and is used to calculate capability values of each top-level indicator according to the standardized values of each bottom-level indicator; The combined weight calculation and synthesis module M4 is used to calculate combined weights of each bottom-level indicator by using a combined weighting method; based on the standardized values and the combined weights, the capability value calculation model is used to calculate capability evaluation values of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator; The equivalence determination module M5 is used to determine the equivalence degree of the two tests based on the bias rate between the capability evaluation values of the small-scale scaled-down test and the large-scale prototype test on each top-level indicator.

Images