Unmanned equipment system task reliability evaluation method based on reconfigurable killing network

Abstract

The invention discloses an unmanned equipment system task reliability assessment method based on a reconfigurable killing net, and relates to the technical field of unmanned equipment task reliability assessment, and the method comprises the steps: collecting multi-modal data, and carrying out the preprocessing of the multi-modal data; extracting multi-modal features based on the pre-processed multi-modal data, and fusing the multi-modal features by using a multi-modal fusion network to generate a high-dimensional environment feature vector; predicting a reliability score of the task according to the high-dimensional environment feature vector; and according to the task reliability score, evaluating the overall reliability trend of the task, and generating a task reliability evaluation report. According to the method, the dynamic change of the task reliability is effectively captured by combining the depth time sequence prediction model of the LSTM and the Transform, accurate future trend prediction is provided, the accuracy and timeliness of unmanned equipment system task reliability evaluation are remarkably enhanced, and powerful support is provided for decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned equipment mission reliability assessment, and in particular to an unmanned equipment system mission reliability assessment method based on a reconfigurable kill network. Background Art

[0002] In recent years, with the widespread application of unmanned equipment systems in military and other fields, the complexity and intelligence level of their task execution have gradually increased. Especially in dynamic battlefield environments, unmanned equipment systems based on reconfigurable kill nets have become an important part of modern military equipment systems due to their flexible task allocation capabilities and modular reconstruction characteristics. Reconfigurable kill nets can achieve collaborative operations in complex mission environments by integrating multiple unmanned equipment in a networked form, greatly improving the efficiency and accuracy of task completion. However, since unmanned equipment systems usually need to perform tasks in dynamic, hostile or even extreme environments, their mission reliability assessment has become an important research direction for measuring system performance and ensuring mission success rate. Existing mission reliability assessment technologies usually rely on single-modal data or static analysis methods, and fail to fully consider the complex interactive characteristics of unmanned equipment systems in multi-modal dynamic environments. Especially in the mission reliability assessment based on reconfigurable kill net systems, existing technologies still have great limitations in data fusion, dynamic modeling and trend prediction.

[0003] The shortcomings of existing technologies in the task reliability assessment of unmanned equipment systems based on reconfigurable kill nets are mainly reflected in the following two points: First, the current reliability assessment methods are usually based on simple statistical analysis or rule derivation, which makes it difficult to capture the changing characteristics of the task reliability of unmanned equipment systems in dynamic environments. Second, in the process of multimodal data fusion, the existing methods fail to fully utilize the importance differences of multimodal features, resulting in poor fusion effects and difficulty in accurately reflecting the task reliability trends of unmanned equipment systems based on reconfigurable kill nets. In addition, there is also a large room for improvement in the existing methods in predicting the overall reliability trend of tasks and the automation capabilities of generating evaluation reports. For example, it is difficult to predict the changing trend of future task reliability through historical data of reliability scores, and it is also impossible to provide accurate decision-making assistance based on evaluation trends. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for evaluating the mission reliability of an unmanned equipment system based on a reconfigurable kill net to solve the problems in the prior art that changes in mission reliability in a dynamic environment are difficult to capture and the multi-modal feature fusion effect is insufficient.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for evaluating the mission reliability of an unmanned equipment system based on a reconfigurable kill net, comprising collecting multimodal data and preprocessing the multimodal data; extracting multimodal features based on the preprocessed multimodal data, and fusing the multimodal features using a multimodal fusion network to generate a high-dimensional environmental feature vector; predicting a reliability score of the task based on the high-dimensional environmental feature vector; evaluating the overall reliability trend of the task based on the task reliability score, and generating a task reliability evaluation report.

[0008] As a preferred solution of the unmanned equipment system mission reliability assessment method based on the reconfigurable kill net described in the present invention, the multimodal data includes terrain image data, weather data and enemy dynamic data.

[0009] As a preferred solution of the unmanned equipment system task reliability assessment method based on the reconfigurable kill net of the present invention, wherein: the multimodal data is preprocessed, and the specific steps are as follows:

[0010] Gaussian filtering is used to remove noise from terrain images;

[0011] Use outlier detection to remove missing values ​​and abnormal data in weather data, and use time alignment algorithm to unify weather data to the timeline of task execution;

[0012] The normalization method is used to standardize the enemy's dynamic data.

[0013] As a preferred solution of the unmanned equipment system mission reliability assessment method based on the reconfigurable kill net of the present invention, wherein: the multimodal features are extracted based on the pre-processed multimodal data, and the specific steps are as follows:

[0014] Use convolutional neural network to perform convolution processing on the preprocessed terrain image to extract terrain features;

[0015] Extract weather characteristics from preprocessed weather data through statistical analysis;

[0016] Graph neural networks are used to extract spatial distribution features from preprocessed enemy dynamic data.

[0017] As a preferred solution of the unmanned equipment system task reliability assessment method based on the reconfigurable kill net of the present invention, wherein: the multimodal features are fused using a multimodal fusion network to generate a high-dimensional environmental feature vector, and the specific steps are as follows:

[0018] The zero-padding method is used to align the dimensions of multimodal features, and the self-attention mechanism is used to dynamically assign the importance weight of each modality feature. The expression is:

[0019]

[0020] Among them, α m is the attention weight of the mth modality feature, v is the attention score vector, U m is the attention weight matrix of the mth modality feature, V m is the feature vector after the alignment of the mth modal features, b m is the bias vector of the mth mode, j represents the index variable of all modal features in the normalization process, U j is the attention weight matrix of the j-th modality feature in the normalization process, V j is the feature vector of the jth modal feature after alignment in the normalization process, b j is the bias vector of the jth modal feature in the normalization process;

[0021] Based on the attention weights assigned to each modal feature, the attention mechanism is used to perform weighted fusion on multimodal features to generate a high-dimensional environment feature vector, which is expressed as:

[0022]

[0023] Among them, X e is a high-dimensional environment feature vector, A is a linear transformation matrix, X is a set of multimodal feature vectors, mean(X) represents the mean value of all elements in the set of multimodal feature vectors, and std(X) represents the standard deviation of all elements in the set of multimodal feature vectors.

[0024] As a preferred solution of the unmanned equipment system task reliability assessment method based on the reconfigurable kill net of the present invention, wherein: the reliability score of the task is predicted according to the high-dimensional environmental feature vector, and the specific steps are as follows:

[0025] A deep time series prediction model is built based on LSTM and Transformer. The high-dimensional environment feature vector is input into the deep time series prediction model and combined with the activation function to capture the dynamic change characteristics in the time dimension and predict the reliability score of the task. The expression is:

[0026] R(t)=σ(W o ReLU(W ff Transformer (LSTM (X e (tk:t))+d ff )+d o );

[0027] Among them, R(t) is the task reliability score at the current time t, X e(tk:t) represents the high-dimensional environment features in the time window from tk to t, σ is the Sigmoid activation function, W o is the weight matrix of the fully connected layer, W ff is the weight matrix of the feedforward neural network, d ff is the bias vector of the feedforward neural network, d o is the bias scalar of the fully connected layer, and k is the time window length.

[0028] As a preferred solution of the unmanned equipment system task reliability assessment method based on the reconfigurable kill net of the present invention, wherein: the deep time series prediction model is constructed based on LSTM and Transformer, and the specific steps are as follows:

[0029] Based on LSTM, it captures the dynamic change characteristics of task reliability in the time dimension, and combines the Transformer structure to extract global features in long time series;

[0030] Set up multiple layers of LSTM units, each unit is responsible for capturing the local time dependency within a time window, and the encoded time series is used as the output of the LSTM layer;

[0031] The Transformer layer uses a multi-head self-attention mechanism based on the output of the LSTM layer and extracts contextual information and long-term dependencies through a feed-forward neural network;

[0032] The outputs of the LSTM layer and the Transformer layer are fused through the fully connected layer to build a deep time series prediction model.

[0033] As a preferred solution of the unmanned equipment system mission reliability assessment method based on the reconfigurable kill net of the present invention, wherein: the overall reliability trend of the mission is assessed according to the mission reliability score, and a mission reliability assessment report is generated, and the specific steps are as follows:

[0034] According to the task reliability score R(t), the average score and extreme value of the task reliability score are obtained and combined with the angular frequency and phase offset to predict the overall reliability trend of the task. The expression is:

[0035]

[0036] Among them, F(t:t+k) is the overall reliability trend index of the task in the future time period t to t+k, σ R is the standard deviation of the reliability score, R max is the maximum reliability score of the task, R min is the minimum reliability score of the task, is the average reliability score of the task, γ is the weight coefficient of the trend change rate, δ is the weight coefficient of the periodic fluctuation, is the trend change between adjacent moments, ω is the angular frequency of the periodic fluctuation, φ is the phase offset of the periodic fluctuation, and t represents the current moment;

[0037] Based on historical task data, define a high reliability threshold H and a low reliability threshold L;

[0038] When F ≥ H, it is considered that the overall reliability of the task will show an upward trend in the future time period t to t + k;

[0039] When F < L, it is considered that the overall reliability of the task will show a downward trend in the future time period t to t + k;

[0040] When L≤F<H, it is considered that the overall reliability of the task tends to be stable in the future time period t to t+k;

[0041] Generate a mission reliability assessment report based on the assessment results of the overall reliability trend of the mission.

[0042] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: the computer program, when executed by the processor, implements any step of the method for assessing the mission reliability of an unmanned equipment system based on a reconfigurable kill net as described in the first aspect of the present invention.

[0043] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: the computer program, when executed by a processor, implements any step of the method for assessing the mission reliability of an unmanned equipment system based on a reconfigurable kill net as described in the first aspect of the present invention.

[0044] The beneficial effects of the present invention are: by intelligently generating high-dimensional environmental feature vectors through a multimodal fusion network and dynamically adjusting the importance of each mode using a self-attention mechanism, the present invention improves the accuracy of data expression and the adaptability of the model. The deep time series prediction model combining LSTM and Transformer effectively captures the dynamic changes of task reliability, provides accurate future trend predictions, significantly enhances the accuracy and timeliness of task reliability assessment of unmanned equipment systems, and provides strong support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0046] Figure 1 Flow chart of the method for evaluating the mission reliability of an unmanned equipment system based on a reconfigurable kill net in Example 1.

[0047] Figure 2 This is a flowchart of preprocessing multimodal data in Example 1. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0049] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0050] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0051] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for evaluating the mission reliability of an unmanned equipment system based on a reconfigurable kill net, comprising the following steps:

[0052] S1: Collect multimodal data and preprocess the multimodal data.

[0053] S1.1: Multimodal data includes terrain image data, weather data, and enemy dynamics data.

[0054] Furthermore, terrain image data are collected through satellite and UAV aerial photography, weather data are obtained from weather stations or satellite weather services, and enemy dynamic data are obtained in real time by reconnaissance equipment and intelligence systems.

[0055] S1.2: Use Gaussian filtering to remove noise from the terrain image.

[0056] For example, by applying a 3x3 Gaussian filter to perform a convolution operation on the terrain image, the image is smoothed and random noise is effectively reduced, thereby improving the image quality.

[0057] S1.3: Use outlier detection to remove missing values ​​and abnormal data in weather data, and use time alignment algorithm to unify weather data to the timeline of task execution.

[0058] For example, the box plot method is used to identify and eliminate outliers in weather data (such as data points whose temperature and humidity exceed 1.5 times the interquartile range). For missing values, linear interpolation is used to fill them based on data from adjacent time points. Finally, a time alignment algorithm is used to ensure that all weather data is consistent with the specific time point of task execution.

[0059] S1.4: Use normalization methods to standardize enemy dynamic data.

[0060] For example, each feature value in the enemy dynamic data is scaled to a range between 0 and 1 to ensure that data of different dimensions are comparable in subsequent analysis, which facilitates model training and prediction.

[0061] S2: Based on the preprocessed multimodal data, multimodal features are extracted and fused using a multimodal fusion network to generate a high-dimensional environmental feature vector.

[0062] S2.1: Use a convolutional neural network to perform convolution processing on the preprocessed terrain image to extract terrain features.

[0063] Specifically, the preprocessed terrain image is input to the input layer of CNN; then, through multiple layers of convolutional layers, each convolutional layer contains multiple convolutional kernels (or filters), which slide on the image and perform local weighted summation operations to extract low-level features such as edges and textures at different scales and directions; then, nonlinear transformations are introduced through activation functions (such as ReLU) to enhance the model's expressiveness; subsequently, pooling layers (such as maximum pooling) are used to reduce the spatial size of the feature map while retaining the most important feature information; finally, through a series of deep convolutional layers, more abstract and complex terrain structure features, such as peaks, valleys and other terrain elements, are gradually extracted, and finally a high-dimensional feature representation rich in terrain information is generated, providing a solid foundation for subsequent task reliability evaluation.

[0064] S2.2: Extract weather characteristics from the preprocessed weather data through statistical analysis.

[0065] Specifically: first, the preprocessed weather data is classified and summarized according to the name (such as temperature, humidity, wind speed, etc.); then, the basic statistics of each weather parameter, such as the mean, standard deviation, maximum and minimum values, are calculated to capture its distribution characteristics; then, the time series analysis method is used to identify the time dependence and periodic patterns of each weather parameter; finally, based on these statistical analysis results, key features that can reflect the weather conditions are extracted, such as the daily variation of temperature, seasonal fluctuations of humidity, and the average intensity of wind speed, to provide detailed weather information support for subsequent mission reliability assessment.

[0066] S2.3: Use graph neural network to extract spatial distribution features from preprocessed enemy dynamic data.

[0067] Specifically, the preprocessed enemy dynamic data is constructed into a graph structure according to the geographical location, in which each node represents a specific enemy unit or observation point, and the edge represents the spatial relationship or interaction between units; then, through the multi-layer propagation mechanism of GNN, message transmission is carried out on the graph, the information of adjacent nodes is aggregated, and the spatial correlation and local patterns between enemy units are captured; then, the node embedding vector is updated using node features and edge weights, and high-level spatial distribution features are gradually extracted; finally, a feature representation reflecting the overall spatial structure of the enemy dynamic data is generated, such as the cluster density, movement trend and strategic deployment of enemy units, providing detailed spatial distribution information for mission reliability assessment.

[0068] S2.4: Use zero padding to align the dimensions of multimodal features, and use the self-attention mechanism to dynamically assign the importance weight of each modality feature, expressed as:

[0069]

[0070] Among them, α m is the attention weight of the mth modality feature, v is the attention score vector, U m is the attention weight matrix of the mth modality feature, V m is the feature vector after the alignment of the mth modal features, b m is the bias vector of the mth mode, j represents the index variable of all modal features in the normalization process, U j is the attention weight matrix of the j-th modality feature in the normalization process, V j is the feature vector of the jth modal feature after alignment in the normalization process, b j is the bias vector of the jth modal feature in the normalization process.

[0071] It should be noted that the zero-filling method refers to a method of filling missing values ​​or data points that need to be aligned with zero values ​​in data processing.

[0072] It should also be noted that by using the zero-filling method to align the dimensions of multimodal features and combining the self-attention mechanism to dynamically assign the importance weight of each modal feature, this step ensures that the different modal data have a unified dimension before fusion, and automatically adjusts the importance of each modal feature according to task requirements. This method not only improves the accuracy and robustness of multimodal data fusion, but also enhances the model's ability to adapt to complex environmental changes, ultimately improving the accuracy and reliability of task reliability assessment.

[0073] S2.5: Based on the attention weights assigned to each modal feature, the attention mechanism is used to perform weighted fusion on multimodal features to generate a high-dimensional environment feature vector, which is expressed as:

[0074]

[0075] Among them, X e is a high-dimensional environment feature vector, A is a linear transformation matrix, X is a set of multimodal feature vectors, mean(X) represents the mean value of all elements in the set of multimodal feature vectors, and std(X) represents the standard deviation of all elements in the set of multimodal feature vectors.

[0076] It should be noted that the linear change matrix is ​​defined based on historical data and is used to linearly combine weighted multimodal feature vectors. The design of this matrix is ​​based on the understanding of the characteristics of different modal data, and its parameters are determined by mathematical methods to ensure that different modal features can be properly scaled and transformed during the fusion process. This ensures that the generated high-dimensional environmental feature vector has stronger expressive power and better structured information, thereby improving the effect of task reliability assessment.

[0077] S3: Predict the reliability score of the task based on the high-dimensional environment feature vector.

[0078] S3.1: Build a deep time series prediction model based on LSTM and Transformer. Input the high-dimensional environment feature vector into the deep time series prediction model and combine it with the activation function to capture the dynamic change characteristics in the time dimension and predict the reliability score of the task. The expression is:

[0079] R(t)=σ(W o ReLU(W ff Transformer (LSTM (X e (tk:t))+d ff )+d o );

[0080] Among them, R(t) is the task reliability score at the current time t, X e(tk:t) represents the high-dimensional environment features in the time window from tk to t, σ is the Sigmoid activation function, W o is the weight matrix of the fully connected layer, W ff is the weight matrix of the feedforward neural network, d ff is the bias vector of the feedforward neural network, d o is the bias scalar of the fully connected layer, and k is the time window length.

[0081] It should be noted that the reason for choosing LSTM is that it is good at capturing long-term dependencies in time series data and is very suitable for processing dynamically changing features in task reliability assessment.

[0082] The reason for choosing Transformer is that it can efficiently process global correlations in multimodal data through the self-attention mechanism, significantly improving the accuracy of feature extraction and fusion.

[0083] S3.2: Build a deep time series prediction model based on LSTM and Transformer. The specific steps are as follows:

[0084] S3.2.1: Based on LSTM, it captures the dynamic changing characteristics of task reliability in the time dimension, and combines the Transformer structure to extract global features in long time series.

[0085] Specifically, the high-dimensional environment feature vector is processed layer by layer through multi-layer long short-term memory network units. Each long short-term memory network unit is responsible for capturing local time dependencies within a time window and memorizing long-term historical information. Then, the output of the long short-term memory network layer is passed to the Transformer structure, and its multi-head self-attention mechanism is used to process all positions in the input sequence in parallel to extract global features and complex patterns in long time series. Finally, the local dynamic characteristics captured by the long short-term memory network and the global features extracted by the Transformer are fused through the fully connected layer.

[0086] S3.2.2: Set up multiple layers of LSTM units, each unit is responsible for capturing local temporal dependencies within a time window, and the encoded time series is used as the output of the LSTM layer.

[0087] For example, there are three layers of LSTM units. The first layer of LSTM units receives high-dimensional environment feature vectors as input, processes these feature vectors one by one through a sliding time window, and captures short-term time dependencies. The second layer of LSTM units is based on the output of the first layer to further capture dynamic changes in a longer time span and enhance the memory of historical information. The third layer of LSTM units finally integrates the information of the first two layers to generate a more abstract and high-level time series representation as the output of the entire LSTM layer.

[0088] S3.2.3: The Transformer layer uses a multi-head self-attention mechanism based on the output of the LSTM layer, and extracts contextual information and long-term dependencies through a feed-forward neural network.

[0089] Specifically, the time series representation generated by the LSTM layer is input into the encoder part of the Transformer, and the information of each time step is processed in parallel through the multi-head self-attention mechanism to capture the global correlation and contextual information between different time points. Then, the feedforward neural network is used to perform nonlinear transformation on the features of each time step to further extract and enhance the contextual information and long-term dependencies. This process can not only efficiently process complex patterns in long time series, but also ensure that the features of each time step are fully paid attention to and optimized, and finally generate a high-dimensional representation rich in global features, providing more accurate support for task reliability assessment.

[0090] S3.2.4: The outputs of the LSTM layer and the Transformer layer are fused through the fully connected layer to build a deep time series prediction model.

[0091] It should be noted that the fully connected layer integrates the different time scale features extracted by the two layers, and generates the final task reliability score through nonlinear transformation, thereby achieving accurate prediction and evaluation of task reliability trends.

[0092] S4: Based on the task reliability score, evaluate the overall reliability trend of the task and generate a task reliability assessment report.

[0093] S4.1: According to the task reliability score R(t), the average score and extreme value of the task reliability score are obtained and combined with the angular frequency and phase offset to predict the overall reliability trend of the task. The expression is:

[0094]

[0095] Among them, F(t:t+k) is the overall reliability trend index of the task in the future time period t to t+k, σ R is the standard deviation of the reliability score, R max is the maximum reliability score of the task, R min is the minimum reliability score of the task, is the average reliability score of the task, γ is the weight coefficient of the trend change rate, δ is the weight coefficient of the periodic fluctuation, is the trend change between adjacent moments, ω is the angular frequency of the periodic fluctuation, φ is the phase offset of the periodic fluctuation, and t represents the current moment.

[0096] S4.2: Based on historical task data, define a high reliability threshold H and a low reliability threshold L.

[0097] When F ≥ H, it is considered that the overall reliability of the task will show an upward trend in the future time period t to t+k.

[0098] When F<L, it is considered that the overall reliability of the task will show a downward trend in the future time period t to t+k.

[0099] When L≤F<H, it is considered that the overall reliability of the task tends to be stable in the future time period t to t+k.

[0100] For example, suppose that historical data shows that the average value of task reliability score is 0.75 and the standard deviation is 0.1. Based on these statistical data, combined with expert experience and specific task requirements, the high reliability threshold H is set to the average value plus one standard deviation (i.e. 0.85) to indicate that the task is in a highly reliable state; and the low reliability threshold L is set to the average value minus one standard deviation (i.e. 0.65) to indicate that the task has a high risk. In this way, it is ensured that the threshold can not only reflect the actual distribution of historical data, but also adapt to the specific needs of different tasks, providing a scientific and reasonable evaluation standard.

[0101] S4.3: Generate a mission reliability assessment report based on the assessment results of the overall reliability trend of the mission.

[0102] Furthermore, the mission reliability assessment report includes an analysis of the overall reliability trend of the mission, prediction results of future reliability, and assessment conclusions and recommendations based on high and low reliability thresholds set based on historical data and expert experience.

[0103] This embodiment also provides a computer device, which is suitable for the case of an unmanned equipment system mission reliability assessment method based on a reconfigurable kill net, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the unmanned equipment system mission reliability assessment method based on a reconfigurable kill net as proposed in the above embodiment.

[0104] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0105] The present embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for evaluating the mission reliability of an unmanned equipment system based on a reconfigurable kill network as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a disk or an optical disk.

[0106] In summary, the present invention improves the accuracy of data expression and the adaptability of the model by: intelligently generating high-dimensional environmental feature vectors through a multimodal fusion network, and dynamically adjusting the importance of each mode using a self-attention mechanism. The deep time series prediction model combining LSTM and Transformer effectively captures the dynamic changes of task reliability, provides accurate future trend predictions, significantly enhances the accuracy and timeliness of task reliability assessment of unmanned equipment systems, and provides strong support for decision-making.

[0107] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a task reliability assessment method for an unmanned equipment system based on a reconfigurable kill net is provided.

[0108] In order to verify the effectiveness of the mission reliability assessment method for unmanned equipment systems based on reconfigurable kill nets, a simulated military exercise was designed with three experimental groups to demonstrate the invention through objective data analysis.

[0109] First, terrain images, weather data, and enemy dynamics data are collected from multiple sources to ensure data diversity and authenticity. The collected data is preprocessed, including denoising, outlier processing, and time alignment.

[0110] Then, a deep time series prediction model is built based on LSTM and Transformer to capture the temporal dependencies in multimodal data.

[0111] Next, the fused features are input into the deep time series prediction model constructed by LSTM and Transformer, and the task reliability score is output through the fully connected layer.

[0112] Finally, the overall reliability trend index is calculated based on the reliability score, high reliability and low reliability thresholds are set based on historical data, and a task reliability assessment report containing analysis, prediction results and recommendations is generated.

[0113] The details are shown in Table 1 below:

[0114] Table 1 Task reliability evaluation report

[0115]

[0116]

[0117] Through the analysis of the data in the above table, it can be clearly seen that the clarity of the terrain image and the stability of the weather have a significant impact on the mission reliability score. When the clarity of the terrain image is high and the weather is stable (such as in Experiment 1), the mission reliability score is high, and the overall reliability is on an upward trend; when the weather is unstable and the density of enemy activities increases (such as in Experiment 2), the mission reliability score drops significantly, and the overall reliability is on a downward trend; even if the enemy activity density is high, but the clarity of the terrain image is low and the weather is stable (such as in Experiment 3), the mission reliability score remains at a medium level, and the overall reliability tends to be stable.

[0118] The present invention can effectively predict the reliability score of the task by integrating multimodal data and using a deep time series prediction model combining LSTM and Transformer. For example, in Experiment 1, the terrain image clarity was 85%, the weather stability was 78%, the enemy activity density was 22%, and the task reliability score reached 0.83, showing a high task reliability. In Experiment 2, the terrain image clarity was 63%, the weather stability was 42%, the enemy activity density was 58%, and the task reliability score was only 0.62, reflecting a low task reliability.

[0119] The above results show that the method of the present invention has obvious advantages in dealing with complex environments and dynamic changes, and can provide more reliable support for decision makers.

[0120] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net, characterized in that: include, Collect multimodal data and preprocess the multimodal data; Based on the preprocessed multimodal data, multimodal features are extracted and fused using a multimodal fusion network to generate a high-dimensional environmental feature vector. Predict the reliability score of the task based on the high-dimensional environment feature vector; Based on the mission reliability score, the overall reliability trend of the mission is evaluated and a mission reliability assessment report is generated.

2. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 1, characterized in that: The multimodal data includes terrain image data, weather data and enemy dynamic data.

3. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 2, characterized in that: The multimodal data is preprocessed, and the specific steps are as follows: Gaussian filtering is used to remove noise from terrain images; Use outlier detection to remove missing values ​​and abnormal data in weather data, and use time alignment algorithm to unify weather data to the timeline of task execution; The normalization method is used to standardize the enemy's dynamic data.

4. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 3, characterized in that: The multimodal features are extracted based on the preprocessed multimodal data, and the specific steps are as follows: Use convolutional neural network to perform convolution processing on the preprocessed terrain image to extract terrain features; Extract weather characteristics from preprocessed weather data through statistical analysis; Graph neural network is used to extract spatial distribution features from preprocessed enemy dynamic data.

5. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 4, characterized in that: The multimodal features are fused using a multimodal fusion network to generate a high-dimensional environment feature vector. The specific steps are as follows: The zero-padding method is used to align the dimensions of multimodal features, and the self-attention mechanism is used to dynamically assign the importance weight of each modality feature. The expression is: Among them, α m is the attention weight of the mth modality feature, v is the attention score vector, U m is the attention weight matrix of the mth modality feature, V m is the feature vector after the alignment of the mth modal features, b m is the bias vector of the mth mode, j represents the index variable of all modal features in the normalization process, U j is the attention weight matrix of the j-th modality feature in the normalization process, V j is the feature vector of the jth modal feature after alignment in the normalization process, b j is the bias vector of the jth modal feature in the normalization process; Based on the attention weights assigned to each modal feature, the attention mechanism is used to perform weighted fusion on multimodal features to generate a high-dimensional environment feature vector, which is expressed as: Among them, X e is a high-dimensional environment feature vector, A is a linear transformation matrix, X is a set of multimodal feature vectors, mean(X) represents the mean value of all elements in the set of multimodal feature vectors, and std(X) represents the standard deviation of all elements in the set of multimodal feature vectors.

6. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 5, characterized in that: The reliability score of the prediction task is predicted based on the high-dimensional environment feature vector. The specific steps are as follows: A deep time series prediction model is built based on LSTM and Transformer. The high-dimensional environment feature vector is input into the deep time series prediction model and combined with the activation function to capture the dynamic change characteristics in the time dimension and predict the reliability score of the task. The expression is: R(t)=σ(W o ·ReLU(W ff ·Transformer(LSTM(X e (t-k:t))+d ff )+d o ); Among them, R(t) is the task reliability score at the current time t, X e (tk:t) represents the high-dimensional environment features in the time window from tk to t, σ is the Sigmoid activation function, W o is the weight matrix of the fully connected layer, W ff is the weight matrix of the feedforward neural network, d ff is the bias vector of the feedforward neural network, d o is the bias scalar of the fully connected layer, and k is the time window length.

7. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 6, characterized in that: The specific steps of building a deep time series prediction model based on LSTM and Transformer are as follows: Based on LSTM, it captures the dynamic change characteristics of task reliability in the time dimension, and combines the Transformer structure to extract global features in long time series; Set up multiple layers of LSTM units, each unit is responsible for capturing the local time dependency within a time window, and the encoded time series is used as the output of the LSTM layer; The Transformer layer uses a multi-head self-attention mechanism based on the output of the LSTM layer and extracts contextual information and long-term dependencies through a feed-forward neural network; The outputs of the LSTM layer and the Transformer layer are fused through the fully connected layer to build a deep time series prediction model.

8. The method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as claimed in claim 6, characterized in that: According to the task reliability score, the overall reliability trend of the task is evaluated, and a task reliability evaluation report is generated. The specific steps are as follows: According to the task reliability score R(t), the average score and extreme value of the task reliability score are obtained and combined with the angular frequency and phase offset to predict the overall reliability trend of the task. The expression is: Among them, F(t:t+k) is the overall reliability trend index of the task in the future time period t to t+k, σ R is the standard deviation of the reliability score, R max is the maximum reliability score of the task, R min is the minimum reliability score of the task, is the average reliability score of the task, γ is the weight coefficient of the trend change rate, δ is the weight coefficient of the periodic fluctuation, is the trend change between adjacent moments, ω is the angular frequency of the periodic fluctuation, φ is the phase offset of the periodic fluctuation, and t represents the current moment; Based on historical task data, define a high reliability threshold H and a low reliability threshold L; When F ≥ H, it is considered that the overall reliability of the task will show an upward trend in the future time period t to t + k; When F < L, it is considered that the overall reliability of the task will show a downward trend in the future time period t to t + k; When L≤F<H, it is considered that the overall reliability of the task tends to be stable in the future time period t to t+k; Generate a mission reliability assessment report based on the assessment results of the overall reliability trend of the mission.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as described in any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program, when executed by a processor, implements the steps of the method for evaluating mission reliability of an unmanned equipment system based on a reconfigurable kill net as described in any one of claims 1 to 8.

Images