Unmanned aerial vehicle state parameter prediction method and system based on operation condition data cluster division
By dividing UAV flight data into operational condition data clusters and adaptively updating the model, the problem of decreased prediction accuracy of UAVs under multiple operational conditions is solved, and efficient and accurate prediction of UAV state parameters is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HARBIN INST OF TECH
- Filing Date
- 2024-11-08
- Publication Date
- 2026-05-08
AI Technical Summary
Under various operating conditions, existing technologies struggle to establish highly adaptable prediction models for drones, leading to decreased prediction accuracy. Furthermore, multi-model methods require significant storage space and training time.
A method based on operational condition data clustering was adopted, and the DTW algorithm was used to cluster the flight data to establish a 1DCNN benchmark prediction model library. The model was then matched and updated using similarity measurement and domain adaptation algorithms to optimize the model training parameters.
It improves the adaptability and accuracy of the prediction model, reduces storage space and training time, enhances the model's generalization ability, and achieves accurate prediction of UAV operating state parameters.
Smart Images
Figure CN121997058A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of aircraft technology, specifically relating to an aircraft condition monitoring technology. Background Technology
[0002] During flight, various systems of a drone can collect and record a large amount of flight data, which contains information about the drone's operational status. Drone status monitoring technology can use this status information to determine whether the drone has deviated from its given operational state, promptly and effectively identify potential safety hazards, and support subsequent decision-making and adjustments to mission plans. It is an important technical measure to protect drone flight safety.
[0003] Prediction-based methods can train predictive models based on expected UAV flight data, monitoring the actual operational status of UAVs through accurate predictions of key parameters. However, the actual operating conditions of UAVs are not static; the diversity and dynamism of these conditions inevitably lead to more complex characteristics in the collected flight data. Changes in the distribution characteristics of flight data and alterations in the correlations between multidimensional data make it difficult to accurately construct predictive models for key operational parameters. If a single model is built for the entire operational process using a global modeling approach, averaging features from multiple operating conditions, important local data features are easily lost. Conversely, if a multi-model approach is used, separate, accurate models need to be built for different operating conditions, which presents a significant challenge for data partitioning and model classification, and requires substantial storage space and training time. Summary of the Invention
[0004] This invention proposes a method and system for predicting UAV state parameters based on the division of operational condition data clusters. Its purpose is to solve the problem of decreased prediction model accuracy caused by the complexity of UAV data characteristics and the emergence of new operational conditions under multiple operational conditions.
[0005] The present invention proposes a method for predicting UAV state parameters based on operational condition data cluster partitioning, which includes:
[0006] S1: Collect the flight data of the UAV, preprocess it, divide the preprocessed flight data, and obtain the flight dataset;
[0007] S2: Based on the flight dataset, establish a benchmark prediction model library;
[0008] S3: Based on the UAV's test data, the optimal prediction model matching method based on similarity metric is used to obtain the optimal similarity result matrix. Based on the optimal similarity matching, the benchmark prediction model is selected.
[0009] S4: Use a domain adaptive algorithm to adjust the training parameters of the baseline prediction model and update the baseline prediction model;
[0010] S5: The updated baseline prediction model is used to predict the state parameters of the UAV.
[0011] Furthermore, a preferred solution is provided: the preprocessing includes: data cleaning, data alignment, data standardization, and data reconstruction of the flight data.
[0012] Furthermore, a preferred solution is provided: the division of the preprocessed flight data is performed by using the DTW algorithm to divide the processed data into data clusters.
[0013] Furthermore, a preferred embodiment is provided: the data cluster partitioning includes:
[0014] Calculate the DTW similarity of the observed variables of the UAV operational status for different sorties;
[0015] Traverse different flights and calculate their DTW similarity to construct a similarity matrix;
[0016] Calculate the minimum similarity value for different flights for each row of the similarity matrix and update the coefficient matrix accordingly;
[0017] Iteratively merge the sorties until all data clusters are divided.
[0018] Furthermore, a preferred solution is provided: the benchmark prediction model library is constructed using 1DCNN.
[0019] Furthermore, a preferred solution is provided: the optimal prediction model matching method includes:
[0020] Calculate the DTW similarity between the target domain dataset and the flight dataset;
[0021] The DTW similarity between each flight dataset and the target domain dataset is calculated sequentially to obtain a similarity result matrix;
[0022] For the target domain dataset, the flight dataset with the highest similarity is selected for matching.
[0023] Furthermore, a preferred embodiment is provided: S4 includes:
[0024] Set the domain adaptive algorithm to adjust the model training parameters;
[0025] Update the prediction model for key parameters of the running state, retain the parameters of the convolutional and pooling layers of the source domain training model, and update the flattening layer and two fully connected layers.
[0026] This invention also proposes a UAV state parameter prediction system based on operational condition data cluster partitioning, comprising:
[0027] Data processing module: used to collect flight data of UAVs, perform preprocessing, divide the preprocessed flight data, and obtain flight datasets;
[0028] Model building module: used to build a benchmark prediction model library based on the flight dataset;
[0029] The optimal matching module is used to obtain the optimal similarity result matrix based on the optimal prediction model matching method based on similarity metric of the UAV test data, and select the benchmark prediction model based on the optimal similarity matching.
[0030] Model update module: Used to adjust the training parameters of the baseline prediction model using a domain adaptive algorithm, and update the baseline prediction model;
[0031] Parameter prediction module: Used to predict the state parameters of the UAV using the updated baseline prediction model.
[0032] The present invention also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor executes a method for predicting UAV state parameters based on the partitioning of operating condition data clusters as described in any of the above-mentioned scheme combinations.
[0033] The present invention also proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the UAV state parameter prediction method based on the partitioning of operating condition data clusters as described in any of the above-described schemes.
[0034] Compared with the prior art, the advantages of the present invention are:
[0035] 1. Adaptability to multiple operating conditions:
[0036] The method described in this invention establishes a predictive model library based on operational condition data clusters, which can adapt to the data characteristics of UAVs under different operational conditions, thus solving the problem that traditional single models are difficult to adapt to changes in multiple operational conditions.
[0037] 2. Data Feature Extraction and Optimization:
[0038] The method described in this invention utilizes the Dynamic Time Warping (DTW) algorithm to divide the original flight data into data clusters, reconstructing similar datasets with easily extracted features, thereby improving the efficiency and accuracy of data feature extraction.
[0039] 3. Model matching and adaptive update:
[0040] This invention achieves optimal model selection for test data through an optimal model matching method based on similarity metric, thereby improving the applicability and accuracy of the prediction model. At the same time, it proposes an adaptive update method for the prediction model based on domain adaptation, which adjusts the model training parameters to enable the model to adapt to the occurrence of new working conditions and maintain prediction accuracy.
[0041] 4. Improved prediction accuracy:
[0042] The method described in this invention can predict the operating status parameters of drones relatively accurately.
[0043] 5. Computational efficiency and resource optimization:
[0044] Compared to multi-model methods, the method described in this invention reduces the requirements for storage space and training time, and improves computational efficiency.
[0045] 6. Model generalization ability:
[0046] The method described in this invention uses a domain adaptation algorithm, which enables the model to not only retain the weights trained from historical data, but also learn complex features under new data conditions, thereby enhancing the model's generalization ability.
[0047] In summary, this invention proposes a method for predicting UAV state parameters based on operational condition data clusters. It establishes a prediction model library based on these operational condition data clusters, and then uses adaptive matching and updating of the models. First, it proposes an operational condition data clustering method based on dynamic time warping, reconstructing the preprocessed raw sortie data into several datasets with similar operational conditions and easily extractable data features, and establishing a benchmark prediction model library based on a one-dimensional convolutional neural network (1DCNN). Then, it constructs an optimal model matching method based on similarity metrics to achieve optimal model selection for the test data. Finally, it proposes an adaptive update method for the prediction model based on domain adaptation, adjusting the model training parameters to update the prediction model and achieve accurate prediction of UAV operational state parameters. This solves the problem of decreased prediction model accuracy due to the complexity of UAV data features and the emergence of new operational conditions under multiple operational conditions.
[0048] This invention is applicable to unmanned aerial vehicle (UAV) status monitoring scenarios. Attached Figure Description
[0049] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0050] Figure 1 The flowchart is a method for predicting UAV state parameters based on the partitioning of operational condition data clusters, as described in a specific embodiment of the present invention.
[0051] Figure 2 The prediction results of the training set model for the target dataset I as described in the third specific embodiment of the present invention;
[0052] Figure 3 This refers to the prediction results of the target dataset I test set model as described in the third specific embodiment of the present invention. Detailed Implementation
[0053] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application can also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.
[0054] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0055] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0057] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.
[0058] Implementation Method 1:
[0059] Reference Figure 1 This implementation method is described below.
[0060] A method for predicting UAV state parameters based on operational condition data cluster partitioning includes:
[0061] S1: Collect UAV flight data and preprocess it. Divide the preprocessed flight data into segments to obtain a flight dataset. The preprocessing includes: data cleaning, data alignment, data standardization, and data reconstruction of the flight data. The division of the preprocessed flight data is performed by using the DTW algorithm to divide the processed data into data clusters.
[0062] Furthermore, the data cluster partitioning includes:
[0063] Calculate the DTW similarity of the observed variables of the UAV operational status for different sorties;
[0064] Traverse different flights and calculate their DTW similarity to construct a similarity matrix;
[0065] Calculate the minimum similarity value for different flights for each row of the similarity matrix and update the coefficient matrix accordingly;
[0066] Iteratively merge flight records until all data clusters are divided;
[0067] S2: Based on the flight dataset, a benchmark prediction model library is established using 1DCNN;
[0068] S3: Based on the UAV's test data, the optimal prediction model matching method based on similarity metric is used to obtain the optimal similarity result matrix. Based on the optimal similarity matching, the benchmark prediction model is selected.
[0069] The optimal prediction model matching method includes:
[0070] Calculate the DTW similarity between the target domain dataset and the flight dataset;
[0071] The DTW similarity between each flight dataset and the target domain dataset is calculated sequentially to obtain a similarity result matrix;
[0072] For the target domain dataset, the flight dataset with the highest similarity is selected for matching;
[0073] S4: The baseline prediction model is updated by adjusting its training parameters using a domain adaptive algorithm, including:
[0074] Set the domain adaptive algorithm to adjust the model training parameters;
[0075] Update the prediction model for key parameters of the running state, retain the parameters of the convolutional and pooling layers of the source domain training model, and update the flattening layer and two fully connected layers.
[0076] S5: The updated baseline prediction model is used to predict the state parameters of the UAV.
[0077] Implementation Method Two:
[0078] This embodiment is a further illustrative example of the UAV state parameter prediction method based on operational condition data cluster partitioning described in Embodiment 1.
[0079] The method for predicting UAV state parameters based on operational condition data cluster partitioning described in this embodiment includes the following detailed steps:
[0080] 1. Construction of a benchmark prediction model library based on the partitioning of operating condition data clusters:
[0081] Flight data from drones is collected and preprocessed. The preprocessed flight data is then divided to obtain a flight dataset. Based on the flight dataset, a benchmark prediction model library is established using 1DCNN.
[0082] This implementation first performs data preprocessing on the UAV flight data, including data cleaning, data alignment, data standardization, and data reconstruction. Then, the DTW algorithm is used to divide the preprocessed multi-operational-condition flight data into data clusters.
[0083] Suppose the data representing the operational status of a single UAV flight to be segmented is X = [x1, x2, ..., x...]. n ], x n ∈R d×1 This is the input to the model;
[0084] Where d is the variable feature length and n is the number of sample flights to be divided.
[0085] Define the maximum similarity as s max Calculate the DTW similarity of the UAV operational status observation variables for sorties a and b:
[0086]
[0087] The larger the value, the lower the similarity of the operating conditions between the two sorties.
[0088] Starting from flight 1, iterate through different flights and calculate their DTW similarity to construct a similarity matrix. Then, calculate the minimum similarity value for each flight in the matrix rows.
[0089]
[0090] In the formula, min() will obtain the minimum distance and the defined maximum similarity s. max Compare and update the coefficient matrix;
[0091] s max The calculation can be automatically determined based on data characteristics:
[0092]
[0093] In the formula, mean() represents the mean value, and thre is the threshold adjustment parameter. The smaller the parameter, the higher the data similarity within the data cluster, and the stricter the restriction on the similarity of working conditions. In actual application, this parameter can be adjusted as needed. In this embodiment, its value is set to 0.8.
[0094] Taking the update process of s1 as an example, let's assume... The first data cluster for operational conditions is flight 1; assuming Update x1 to x 1i This involves merging two flights and iteratively updating and calculating their similarity to all remaining flights:
[0095]
[0096] This process is repeated until... At this point, the first operational data cluster is the set of sortie 1, sortie i, and so on up to sortie m. The first operational data cluster is now divided, and a new dataset 1 is obtained. The above process is repeated in the remaining sortie data until all data clusters are divided.
[0097] Subsequently, a prediction model for the UAV's operational state parameters is constructed based on a one-dimensional convolutional neural network. The 1DCNN prediction modeling process is abstracted as f... 1DCNN (g) The model input is the UAV dataset after data clustering.
[0098] For each dataset, define the multivariate X = [x1, x2, ..., x...]. n ], x n ∈R d×1 The model is trained by distinguishing between input and output variables, thereby obtaining key parameters to predict the optimal model.
[0099] The model output variables are the predicted values of the UAV state parameters, and the mathematical representation of the model is as follows:
[0100]
[0101] This results in a baseline prediction model library constructed from multiple data clusters divided according to operating conditions:
[0102] F = [f 1DCNN1 (g),f 1DCNN2(g),f 1DCNN3 (g),L]F (3)
[0103] 2. Optimal prediction model matching based on similarity metrics:
[0104] Based on the UAV's test data, an optimal prediction model matching method based on similarity metrics is used to obtain the optimal similarity result matrix. Based on the optimal similarity matching, a baseline prediction model is selected, specifically including:
[0105] First, calculate the DTW similarity between the target domain dataset and the dataset after data clustering (source domain dataset):
[0106]
[0107] Among them, S i Let T represent the target variable of the i-th source domain dataset, i = 1, 2, ..., M. j Let x represent the target variable of the j-th target domain dataset, where j = 1, 2, ..., N, and x represents the state representation variable.
[0108] Then, the DTW similarity between each source domain dataset and the target domain dataset is calculated sequentially to obtain the similarity result matrix S:
[0109]
[0110] For the j-th target domain dataset, select the optimal source domain dataset, i.e., the one with the highest similarity, for matching:
[0111]
[0112] Therefore, the optimal similarity result matrix is The baseline prediction model is selected based on the optimal similarity match, ultimately achieving optimal prediction model matching based on similarity metrics:
[0113] F best =f se (f 1DCNN1 (g),f 1DCNN2 (g),...,f 1DCNNN (g)) (7)
[0114] In the formula, f se This is a mathematical expression of the above method for selecting the optimal model based on DTW similarity.
[0115] 3. Prediction model update based on domain adaptation method:
[0116] For a dataset that matches the baseline prediction model, setting the domain adaptive algorithm to adjust the model training parameters and update the prediction model with key parameters of the running state can be represented as:
[0117] F new =f da (F best (8)
[0118] In the formula, f da To provide a mathematical representation of the prediction model update process, the update method is the model domain adaptive method.
[0119] The model mainly consists of convolutional layers, pooling layers, flattening layers, and two fully connected layers, and is a multi-input single-output model.
[0120] This implementation retains the parameters of the convolutional and pooling layers of the model trained in the source domain to preserve the model weights trained from historical data. It updates the flattening layer and the two fully connected layers to enable the model to fully learn the complex features under the changing conditions of new data.
[0121] After completing the model update, the new dataset's multidimensional variables X new =[x new1 ,x new2 ,L,x newn ], x new ∈R d×1 As input, the estimated output value of the target variable can be obtained:
[0122]
[0123] This completes the entire prediction model update process, thereby enabling accurate prediction of UAV state parameters.
[0124] Implementation Method 3:
[0125] Reference Figure 2 , Figure 3 This implementation method is described below. Figure 2 , Figure 3 This is the prediction result of the prediction model update method on the target dataset I.
[0126] This embodiment is a further illustrative example of the UAV state parameter prediction method based on operational condition data cluster partitioning described in Embodiment 1 and Embodiment 2.
[0127] This implementation uses real data from a fixed-wing UAV, and the specific parameters are shown in Table 1.
[0128] Table 1 Specific Flight Status Parameters of the UAV
[0129] parameter parameter parameter parameter Eastward speed normal acceleration Left outer aileron control surface report Roll rate Northbound speed longitudinal acceleration Left inner aileron control surface return Inertial navigation yaw rate Skyward speed lateral acceleration Inertial navigation longitude Inertial navigation track angle longitude upper right V-shaped tail control surface report Inertial navigation latitude Inertial navigation normal acceleration latitude Lower right V-shaped tail control surface report Inertial navigation eastward velocity Inertial navigation longitudinal acceleration track angle Left upper V tail control surface report Inertial navigation northbound velocity Inertial navigation lateral acceleration Geographical altitude Lower left V-shaped tail control surface report Inertial navigation upward velocity Inertial navigation axial acceleration barometric altitude Right outer aileron control surface report Inertial navigation angle Inertial navigation pitch angle relative height Right inner aileron control surface return Inertial navigation pitch rate Inertial navigation roll angle
[0130] Taking roll angle as an example, it is used as a parameter to characterize the UAV's state. A total of 10 flight data sets of varying lengths were used, with the first six sets serving as source domain data and the last four sets as target domain data. During the experiment, the ratio of data used for training to data used for testing was 8:2.
[0131] The above-mentioned UAV state parameter prediction method was used to automatically divide the data from 10 sorties. The resulting data cluster classifications are shown in Table 2.
[0132] Table 2 Data Cluster Partition Results
[0133] Dataset numbering after source domain partitioning Source domain raw data flight number Dataset numbering after target partitioning Target raw data flight number Ⅰ 1，4，6 Ⅰ 7，9 Ⅱ 2，3，5 Ⅱ 8，10
[0134] A 1DCNN was used to construct a prediction model for key parameters reflecting the operational status of the UAVs from the partitioned data clusters. The root mean square error (RMSE) and mean absolute error (MAE) were used as performance indicators to evaluate the prediction accuracy of the model. The evaluation results of the prediction model for the partitioned data clusters and the prediction model for the original sortie data are shown in Tables 3 and 4, respectively.
[0135] Table 3. Prediction results of the data cluster model after partitioning.
[0136]
[0137] Table 4. Prediction Results of the Original Flight Data Model
[0138]
[0139] It can be seen that the prediction model constructed after data clustering has higher prediction accuracy than the original flight prediction model.
[0140] Using the optimal prediction model matching method based on similarity metric proposed in this invention, the DTW similarity metric was performed between the new target dataset and the new source dataset in the partitioned data cluster. The results are shown in Table 5.
[0141] Table 5. Similarity Measurement Results
[0142] Dataset New dataset from source domain I New Dataset II from Source Domain Target New Dataset I 3136747 2905934 Target New Dataset II 3016476 2764839
[0143] As can be seen from Table 5, the DTW of the two target new datasets and the source domain new dataset II is smaller, that is, the model matching result is the source domain new dataset II.
[0144] Subsequently, the prediction model update method based on domain adaptation described in Implementation Method 1 is used. Taking target dataset I as an example, the prediction results using the aforementioned prediction model update method based on domain adaptation are as follows: Figure 2As shown.
[0145] Table 6 shows the prediction results comparing the proposed optimal model matching with non-optimal model matching, and the proposed prediction model update method with the retraining model method.
[0146] Table 6 Comparison of Multi-Data Domain Adaptive Methods
[0147]
[0148]
[0149] from Figure 2 It can be seen that the prediction model update method described above has good fitting results, and the overall range and fluctuation of the residuals are relatively small.
[0150] As can be seen from Table 6, the proposed UAV state parameter prediction method based on the division of operational condition data clusters effectively improves the prediction accuracy of the model.
[0151] In summary, the results show that the UAV state parameter prediction method based on operational condition data cluster partitioning described in the above embodiments can predict the UAV's operational state parameters relatively accurately, providing strong support for the application of this invention.
[0152] Implementation Method Four:
[0153] A UAV state parameter prediction system based on operational condition data cluster partitioning, the system being implemented based on the UAV state parameter prediction method based on operational condition data cluster partitioning described in Embodiments 1 and 2, the system comprising:
[0154] Data processing module: used to collect flight data of UAVs, perform preprocessing, divide the preprocessed flight data, and obtain flight datasets;
[0155] Model building module: used to build a benchmark prediction model library based on the flight dataset;
[0156] The optimal matching module is used to obtain the optimal similarity result matrix based on the optimal prediction model matching method based on similarity metric of the UAV test data, and select the benchmark prediction model based on the optimal similarity matching.
[0157] Model update module: Used to adjust the training parameters of the baseline prediction model using a domain adaptive algorithm, and update the baseline prediction model;
[0158] Parameter prediction module: Used to predict the state parameters of the UAV using the updated baseline prediction model.
[0159] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. This is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
[0160] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.
[0161] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.
[0162] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware implementation, a completely software implementation, or an implementation combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0163] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0164] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0165] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit its protection scope. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading this disclosure, they can still make various changes, modifications or equivalent substitutions to the specific embodiments of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the claims to be published and pending approval.
Claims
1. A method for predicting UAV state parameters based on operational condition data cluster partitioning, characterized in that, The method includes: S1: Collect the flight data of the UAV, preprocess it, divide the preprocessed flight data, and obtain the flight dataset; S2: Based on the flight dataset, establish a benchmark prediction model library; S3: Based on the UAV's test data, the optimal prediction model matching method based on similarity metric is used to obtain the optimal similarity result matrix. Based on the optimal similarity matching, the benchmark prediction model is selected. S4: Use a domain adaptive algorithm to adjust the training parameters of the baseline prediction model and update the baseline prediction model; S5: The updated baseline prediction model is used to predict the state parameters of the UAV.
2. The method for predicting UAV state parameters based on operational condition data cluster partitioning according to claim 1, characterized in that, The preprocessing includes: data cleaning, data alignment, data standardization, and data reconstruction of the flight data.
3. The method for predicting UAV state parameters based on operational condition data cluster partitioning according to claim 1, characterized in that, The process of dividing the preprocessed flight data involves using the DTW algorithm to divide the processed data into data clusters.
4. The method for predicting UAV state parameters based on operational condition data cluster partitioning according to claim 3, characterized in that, The data cluster partitioning includes: Calculate the DTW similarity of the observed variables of the UAV operational status for different sorties; Traverse different flights and calculate their DTW similarity to construct a similarity matrix; Calculate the minimum similarity value for different flights for each row of the similarity matrix and update the coefficient matrix accordingly; Iteratively merge the sorties until all data clusters are divided.
5. The method for predicting UAV state parameters based on operational condition data cluster partitioning according to claim 1, characterized in that, The baseline prediction model library is built using 1DCNN.
6. The method for predicting UAV state parameters based on operational condition data cluster partitioning according to claim 1, characterized in that, The optimal prediction model matching method includes: Calculate the DTW similarity between the target domain dataset and the flight dataset; The DTW similarity between each flight dataset and the target domain dataset is calculated sequentially to obtain a similarity result matrix; For the target domain dataset, the flight dataset with the highest similarity is selected for matching.
7. The method for predicting UAV state parameters based on operational condition data cluster partitioning according to claim 1, characterized in that, S4 includes: Set the domain adaptive algorithm to adjust the model training parameters; Update the prediction model for key parameters of the running state, retain the parameters of the convolutional and pooling layers of the source domain training model, and update the flattening layer and two fully connected layers.
8. A UAV state parameter prediction system based on operational condition data cluster partitioning, characterized in that, The system includes: Data processing module: used to collect flight data of UAVs, perform preprocessing, divide the preprocessed flight data, and obtain flight datasets; Model building module: used to build a benchmark prediction model library based on the flight dataset; The optimal matching module is used to obtain the optimal similarity result matrix based on the optimal prediction model matching method based on similarity metric of the UAV test data, and select the benchmark prediction model based on the optimal similarity matching. Model update module: Used to adjust the training parameters of the baseline prediction model using a domain adaptive algorithm, and update the baseline prediction model; Parameter prediction module: Used to predict the state parameters of the UAV using the updated baseline prediction model.
9. A computer device, including a memory and a processor, characterized in that, The memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the UAV state parameter prediction method based on the partitioning of operating condition data clusters as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the UAV state parameter prediction method based on the partitioning of operational condition data clusters as described in any one of claims 1-7.