A Situation Awareness and Assessment Method Based on Dynamic Data

Abstract

This application provides a situation awareness and assessment method based on dynamic data, especially a situation awareness and assessment method based on multi-source dynamic data fusion. The method includes: obtaining an original data set, which includes UAV flight attitude data, environmental obstacle data, meteorological data, and target dynamic data; judging whether the data quality of the original data set is lower than a preset threshold. If there is data missing, an interpolation algorithm is used to fill in the missing values. If there is noise interference, a filtering algorithm is used to remove the noise, generating a corrected data set; based on the corrected data set, the multi-source data is aligned to a unified coordinate system through a preset spatio-temporal mapping model, generating a fused data set; according to the dynamic characteristic parameters of the fused data set, the data acquisition frequency is adjusted to generate a sampled optimized data set; the sampled optimized data set is input into the evaluation model R, and a situation assessment result including the environmental risk level is output.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular, to a situation awareness and evaluation method based on dynamic data. Background Art

[0002] When an unmanned aerial vehicle (UAV) executes complex tasks, the real-time acquisition and processing of dynamic data is the core link of situation awareness. With the increasing diversity and uncertainty of the UAV flight environment, the data dimensions collected by sensors are constantly expanding, including flight attitude, environmental obstacles, meteorological information, and target dynamics. However, there are technical contradictions in the real-time fusion of multi-source heterogeneous data: on the one hand, high-frequency data acquisition requires the system to have fast processing capabilities to ensure real-time response to the dynamic environment; on the other hand, the fusion of multi-dimensional data requires complex computational models, which may lead to an increase in system latency and difficulty in meeting real-time requirements.

[0003] In addition, the quality of dynamic data is crucial for the accuracy of situation assessment. In a complex environment, sensors may be affected by noise, interference, or hardware failures, resulting in data distortion or missing. How to ensure the reliability and integrity of data without reducing the data acquisition frequency is another technical contradiction. For example, when the UAV is flying at low altitude, ground reflection signals and building occlusion may cause positioning data drift, and high-altitude airflow changes may affect the accuracy of attitude sensors. These problems require the system to have the ability to correct and complete data while processing data in real time to avoid deviations in situation assessment results. Summary of the Invention

[0004] Aiming at the deficiencies in the prior art, the purpose of the present invention is to propose a situation awareness and evaluation method based on dynamic data, especially a situation awareness and evaluation method based on the fusion of multi-source dynamic data, to solve the problems mentioned in the above background art section.

[0005] The present invention provides a situation awareness and evaluation method based on dynamic data, mainly including:

[0006] Obtain an original data set, where the original data set includes UAV flight attitude data, environmental obstacle data, meteorological data, and target dynamic data;

[0007] Judge whether the data quality of the original data set is lower than a preset threshold. If there is missing data, use an interpolation algorithm to fill in the missing values. If there is noise interference, use a filtering algorithm to remove the noise to generate a corrected data set;

[0008] Based on the corrected data set, align the multi-source data to a unified coordinate system through a preset spatio-temporal mapping model to generate a fused data set;

[0009] Adjust the data acquisition frequency according to the dynamic characteristic parameters of the fusion data set to generate a sampling optimization data set;

[0010] Input the sampling optimization data set into the evaluation model R to output a situation evaluation result including the environmental risk level. The sampling optimization data set includes: a set of flight attitude, environmental obstacles, and meteorological data. The evaluation model R is:

[0011]

[0012] Among them, R a is the attitude risk function, R o is the distance risk function, R m is the meteorological risk function, and R represents the comprehensive risk assessment value; w a , w o , w m respectively represent the weight coefficients of flight attitude, environmental obstacles, and meteorological data, and w a +w o +w m = 1.

[0013] The technical solution provided by the embodiments of the present invention may include the following beneficial effects: By real-time collecting and processing multi-source dynamic data fusion such as UAV flight attitude data, environmental obstacle data, and meteorological data, it is possible to comprehensively and accurately perceive the flight state of the UAV and the surrounding environment. Using the Kalman filter algorithm to smooth the flight attitude data, the collision detection algorithm to judge the relative distance between the UAV and the obstacle, and establishing a meteorological impact model to analyze the impact of meteorological conditions on flight, potential safety hazards such as abnormal attitude, approaching obstacles, and bad weather can be discovered in time, so as to take corresponding measures, effectively reduce the risk of flight accidents, and improve the flight safety of the UAV in complex environments.

[0014] Further, in the data preprocessing stage of the present invention, by judging the data quality and using the interpolation algorithm to fill in the missing values and the filtering algorithm to denoise, a corrected data set is generated, ensuring the high quality of the input data. Further, using the preset spatio-temporal mapping model to align the multi-source data to a unified coordinate system to generate a fusion data set, eliminating the time and space differences between different data sources, and improving the consistency and comparability of the data. In addition, adjusting the data acquisition frequency according to the dynamic characteristic parameters of the fusion data set to generate a sampling optimization data set makes the data acquisition more reasonable and efficient, avoiding data redundancy and information loss, and providing an accurate and reliable basis for subsequent situation evaluation and flight strategy generation. Brief Description of the Drawings

[0015] Figure 1 is a flowchart of a situation awareness and evaluation method based on dynamic data of the present invention. Detailed implementation manners

[0016] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of this specification.

[0017] As Figure 1 , a situation awareness and assessment method based on dynamic data in this embodiment may specifically include:

[0018] Step S101: Obtain an original data set, where the original data set includes UAV flight attitude data, environmental obstacle data, meteorological data, and target dynamic data.

[0019] Obtain UAV flight attitude data, and use the Kalman filtering algorithm to smooth the flight attitude data. Extract obstacle position information from the environmental obstacle data, and combine the smoothed flight attitude data to use a collision detection algorithm to determine the relative distance between the UAV and the obstacle. Obtain meteorological data, analyze parameters such as wind speed, wind direction, and temperature, and combine the UAV flight attitude data to establish a meteorological impact model.

[0020] Specifically, in one of the embodiments, the following formula can be used for the Kalman filtering smoothing of the attitude data:

[0021]

[0022] Where X k represents the attitude state vector at the current moment, F represents the state transition matrix, K k represents the Kalman gain matrix, Z k represents the observation value vector, and H represents the observation matrix. X k may include parameters respectively representing the actual pitch angle, roll angle, and yaw angle X k， pitch , X k， roll , X k， yaw of the UAV.

[0023] The relative distance D between the UAV and the obstacle can be calculated through the following formula:

[0024]

[0025] Where x d , y d , z dRepresents the coordinates of the drone in three-dimensional space, x o 、y o 、z o Represents the coordinates of the obstacle in three-dimensional space.

[0026] In one of the embodiments, the meteorological influence model can be represented by the following formula:

[0027]

[0028] Where, V w Represents the comprehensive wind speed vector, V r Represents the reference wind speed vector, V a Represents the air flow velocity, θ represents the wind direction angle, α represents the attenuation coefficient, and h represents the height.

[0029] Specifically, the flight attitude data of the drone can be obtained through the onboard inertial measurement unit (IMU). The IMU usually includes an accelerometer, a gyroscope, and a magnetometer, and can measure the three-axis acceleration, angular velocity, and magnetic azimuth of the drone in real time. For example, assume that during the flight of the drone, the IMU measures the X-axis acceleration to be 3 m / s², the Y-axis angular velocity to be 5 rad / s, and the Z-axis magnetic azimuth to be 45°. These data can be processed through the Kalman filter algorithm to eliminate noise and improve data accuracy. The environmental obstacle data can be obtained through a lidar (LiDAR) or a vision sensor. Assume that the lidar scans an obstacle 10 meters ahead, and its coordinates are (10, 5, 2). These data can be clustered and identified through a point cloud processing algorithm to determine the type and location of the obstacle. The meteorological data can be obtained through meteorological sensors. For example, the wind speed sensor measures the current wind speed to be 5 m / s, and the temperature sensor measures the current temperature to be 25°C. These data can be analyzed through a numerical weather prediction model to predict the meteorological changes in the next period of time.

[0030] Step S102, determine whether the data quality of the original data set is lower than a preset threshold. If there is data missing, use an interpolation algorithm to fill in the missing values. If there is noise interference, use a filtering algorithm to denoise and generate a corrected data set.

[0031] Obtain the original data set, judge the data quality in the original data set, and determine whether it is lower than the preset threshold. If the data quality is lower than the preset threshold, further determine whether there is data missing in the data set. If there is data missing, use a linear interpolation algorithm to fill in the missing values. If the data quality is lower than the preset threshold and there is no data missing, determine whether there is noise interference in the data set. If there is noise interference, use a median filtering algorithm to denoise the data. According to the above processing results, generate a corrected data set, which contains the filled missing values and the denoised data.

[0032] Further, in other embodiments, data quality assessment can also be performed on the corrected data set to determine whether it meets the preset threshold requirements. If the quality of the corrected data set is still lower than the preset threshold, the Kalman filter algorithm is used to perform secondary denoising on the data. Finally, a corrected data set that meets the preset threshold requirements is generated, completing the process of improving data quality.

[0033] Specifically, when determining whether the data quality of the original data set is lower than the preset threshold, specific threshold criteria need to be set first. For example, the proportion of missing values does not exceed 5%, and the standard deviation of noise does not exceed 1. By analyzing the data set, it is found that the proportion of missing values in a certain column of data is 8%, exceeding the preset threshold. Therefore, an interpolation algorithm needs to be used to fill in the missing values. Here, the linear interpolation method can be selected. Assuming that the data points before and after the missing value are (1, 10) and (3, 30), then the missing value (2, x) can be calculated through the linear interpolation formula x = 10 + (30 - 10) / (3 - 1) * (2 - 1) = 20. For noise interference, by calculating the data standard deviation to be 15, which exceeds the preset threshold, the median filter algorithm is used for denoising. The filter window size is set to 3, and the data sequence [2, 5, 3, 7, 4] is filtered, taking the median of each window to obtain the corrected sequence [3, 4, 5, 4, 4]. After the above processing, the generated data set meets the preset quality requirements and can be used for subsequent analysis.

[0034] Step S103, based on the corrected data set, align the multi-source data to a unified coordinate system through a preset spatio-temporal mapping model to generate a fused data set.

[0035] Obtain a multi-source data set, perform correction processing on the multi-source data set to obtain a corrected data set. Use a preset spatio-temporal mapping model to perform spatio-temporal relationship mapping on the corrected data set to obtain a mapped data set. According to the mapped data set, determine whether the data is aligned to a unified coordinate system. If not, remap it. Perform fusion processing on the aligned mapped data set to generate a fused data set.

[0036] Further, in other embodiments, machine learning algorithms can also be used to analyze the fused data set to determine the data distribution characteristics. According to the data distribution characteristics, generate an analysis result of the fused data set.

[0037] In one of the embodiments, the spatio-temporal relationship mapping of the corrected data set can be performed through the following spatio-temporal mapping model to obtain a mapped data set:

[0038]

[0039] Among them, M represents the final spatio-temporal mapping result, which can be divided into 1...K data blocks, n represents the number of data sources, and ω i represents the weight coefficient of the i-th data source, T represents the spatio-temporal transformation function, and P i represents the position data of the i-th data source, and R i represents the rotation parameter of the i-th data source.

[0040]

[0041] Among them, F represents the fused data set, K represents the number of data blocks to be fused, and β k represents the fusion weight of the k-th data block, S represents the spatial mapping function, H represents the temporal mapping function, X k and Y k respectively represent the spatial and temporal attributes of the k-th data block.

[0042] Step S104: Adjust the data acquisition frequency according to the dynamic characteristic parameters of the fused data set to generate a sampled optimized data set.

[0043] According to the dynamic characteristic parameters of the fused data set, determine the range of eigenvalue of the data set. For the eigenvalue range, calculate the change amplitude of the dynamicity, and obtain the fluctuation interval of the parameter value. Within the fluctuation interval of the parameter value, judge the adjustment value of the acquisition rate, and determine the optimized range of the sampled value. According to the optimized range of the sampled value, calculate the adjustment amplitude of the frequency value, and generate an optimized scheme for the processed value. Adopt the optimized scheme for the processed value to adjust the data acquisition frequency to obtain a sampled optimized data set. According to the sampled optimized data set, extract the distribution law of the eigenvalue, and judge the adjustment range of the fusion value. Within the adjustment range of the fusion value, generate the final sampled optimized data set to complete the data processing process.

[0044] Specifically, according to the dynamic characteristic parameters of the fusion dataset, first, by analyzing the volatility and trend of the time series data, the optimal frequency of data collection is determined. For example, by using the Fast Fourier Transform (FFT) to perform frequency domain analysis on historical data, it is found that the main frequency of the data is concentrated between 1 Hz and 10 Hz. Therefore, the sampling frequency is set to 2 Hz to meet the Nyquist sampling theorem. Then, an adaptive sampling algorithm, such as a prediction model based on Kalman filtering, is adopted to dynamically adjust the sampling interval. When the data change rate accelerates, the system will automatically increase the sampling frequency to 5 Hz to ensure that key information is captured; when the data tends to be stable, the sampling frequency is reduced to 5 Hz to reduce redundant data. During the data collection process, the sliding window technique is used to process the real-time data. The window size is set to 10 seconds, and it slides 1 second each time to ensure the continuity and timeliness of the data. Through the above methods, the generated sampling optimization dataset not only improves the efficiency of data collection but also significantly reduces the consumption of storage and computing resources. For example, in an aircraft sensor network, by adopting the optimized sampling strategy, the data storage volume is reduced by 30%, and at the same time, the capture rate of key events is increased by 15%. This process fully demonstrates the important role of dynamic characteristic parameters in data collection optimization.

[0045] In step S105, input the sampling optimization dataset into the evaluation model, and output the situation assessment result including the environmental risk level. The sampling optimization dataset includes the set of flight attitude, environmental obstacles, and meteorological data, and a comprehensive evaluation model is established to conduct risk assessment on the flight environment of the UAV. The specific evaluation model is as follows:

[0046] Define the attitude risk function R based on the deviation of the attitude state vector X j : a :

[0047]

[0048] X j， pitch 、X j， roll 、X j， yaw respectively represent the actual pitch angle, roll angle, and yaw angle of the UAV, calculate the deviation of these angles from the expected value (expected), and take the sum of their squares as the risk index. w a1 、w a2 、w a3 are weight coefficients, indicating the contribution of each attitude parameter to the risk.

[0049] Define a distance risk function R based on the relative distance D between the UAV and the obstacle o :

[0050]

[0051] Use the reciprocal of the obstacle distance D as the risk index. The closer the distance, the higher the risk. is a very small constant used to avoid division by zero.

[0052] Based on the combined wind speed vector V w Define a meteorological risk function R m :

[0053]

[0054] Combine the above three risk assessment functions to form a comprehensive risk assessment model R:

[0055]

[0056] where R represents the comprehensive risk assessment value; w a , w o , w m represent the weight coefficients of flight attitude, environmental obstacles, and meteorological data respectively, and w a +w o +w m = 1.

[0057] Exemplarily:

[0058] Suppose at a certain moment, the attitude state vector X of the UAV k is expressed as:

[0059] Pitch angle: X k，pitch = 5°

[0060] Roll angle: X k，roll = 3°

[0061] Yaw angle: X k，yaw = 2°

[0062] The expected attitude angle is: X expected，pitch = 0°; X expected，roll = 0°; X expected，yaw = 0°.

[0063] In addition, the relative distance D between the UAV and the obstacle is 10 meters.

[0064] The combined wind speed vector V w = 5 m / s.

[0065] The weight coefficients are set as: w a = 0.4; w o = 0.3; w m = 0.3;

[0066] Calculate each risk index:

[0067] Attitude risk: R a = 13.9

[0068] Obstacle risk: R o = 10 + 0.0011 ≈ 0.1

[0069] Meteorological risk: R m = ∣5∣ = 5

[0070] Comprehensive risk: R = 0.4×13.9 + 0.3×0.1 + 0.3×5 = 5.56 + 0.03 + 1.5 = 7.09

[0071] Through this comprehensive risk assessment model, the flight environment of the UAV can be comprehensively risk - assessed, thus providing a scientific basis for the formulation of flight strategies.

[0072] Step S106, based on the situation assessment result and the preset decision rule base, use the deep reinforcement learning algorithm to generate flight strategy instructions.

[0073] Obtain the situation assessment result and extract key feature data. Input the extracted feature data into the preset decision rule base to match applicable rules. Use the deep reinforcement learning algorithm to construct a policy generation model. Integrate the matched decision rules in the policy generation model to optimize the policy generation process. According to the optimized policy generation model, calculate the flight strategy instructions. Adjust the parameters of the policy generation model to improve the accuracy of the policy instructions. Output the final flight strategy instructions and apply them to the flight control system.

[0074] Specifically, in the process of generating the flight strategy, first, based on the situation assessment result, collect data such as flight altitude, speed, and angle through sensors, and use the Kalman filtering algorithm for data fusion to eliminate noise interference and obtain accurate flight state information. For example, the flight altitude is 8000 meters, the speed is 250 meters per second, and the angle deviation is 5 degrees. Then, combined with the preset decision rule base, use the Deep Q - Network (DQN) in the deep reinforcement learning algorithm for policy optimization. The neural network structure of DQN contains 3 hidden layers, with the number of nodes in each layer being 256, 128, and 64 respectively. The activation function uses ReLU, and the learning rate is set to 0.001. Through interacting with the environment, the agent continuously tries and makes mistakes in the simulated flight environment, updates the Q - value table, and finally generates the optimal flight strategy. For example, when detecting an obstacle ahead, the agent selects a climbing instruction according to the Q - value table, with a climbing angle of 15 degrees and a duration of 5 seconds. Throughout the process, the system monitors the flight state in real - time to ensure the accuracy and safety of policy execution, and continuously optimizes the decision rule base through the feedback mechanism to improve the intelligence level of the flight strategy.

[0075] Step S107, if it is detected that the deviation between the execution data and the historical data exceeds the preset threshold, trigger the weighted fusion processing of the historical data and the current data to generate a completed data set.

[0076] The deviation threshold of the preset historical data is obtained, and the deviation value between the execution data and the historical data is obtained. If the deviation value exceeds the preset threshold, trigger the weighted fusion processing of the historical data and the current data. The weight values of the historical data and the current data are calculated by the weighted method to determine the fusion ratio. The historical data and the current data are fused according to the weight values by the fusion method to generate the completed data.

[0077] In other embodiments, it may further include: merging the completed data with the original data set to form a new data set. The new data set is detected for deviation by the detection method to determine whether the preset conditions are met. If the deviation value does not exceed the preset threshold, the new data set is used as the final completed data set.

[0078] Specifically, during the data detection process, the system will calculate the deviation value between the current execution data and the historical data in real time. For example, assuming that the mean of the historical data is 100, the standard deviation is 5, and the current data is 120, then the deviation value is (120 - 100) / 5 = 4. The preset deviation threshold of the system is 3. Since 4 is greater than 3, the weighted fusion processing is triggered. During the fusion process, the exponentially weighted moving average algorithm (EWMA) is used, where the weight of the historical data is 7 and the weight of the current data is 3. When calculating the completed data, the formula is completed data = historical data * 7 + current data * 3, that is, completed data = 100 * 7 + 120 * 3 = 106. To ensure the accuracy of the data, the system will also perform smoothing processing on the completed data. The moving average algorithm with a window size of 3 is used, and the average value of the completed data and the data before and after it is taken as the final result. For example, if the completed data sequence is [105, 106, 107], then the final completed data is (105 + 106 + 107) / 3 = 106. Through this weighted fusion and smoothing processing, the system can effectively reduce the data deviation and improve the data quality.

[0079] Step S108, re - execute the situation assessment and flight strategy generation based on the completed data set, and update the flight strategy instruction.

[0080] Obtain the completed data set, perform a situation assessment on the data set to obtain the current situation assessment result. According to the situation assessment result, use the preset flight strategy generation algorithm to generate an initial flight strategy. If the difference between the initial flight strategy and the historical strategy in the completed data set exceeds the preset threshold, adjust the strategy parameters and regenerate the flight strategy. According to the adjusted flight strategy, update the content of the flight instruction. Use the machine learning algorithm to optimize the updated instruction to obtain the optimized flight instruction.

[0081] Specifically, during the process of re - executing the situation assessment and flight strategy generation based on the completed dataset, first, the data pre - processing module cleans and normalizes the original data. For example, the flight altitude data is uniformly converted to the metric unit to ensure data consistency. Subsequently, the Kalman filtering algorithm is used to smooth the flight trajectory and reduce noise interference. The specific parameter settings are that the state transition matrix is 95 and the observation noise covariance is 1. In the situation assessment stage, a convolutional neural network model based on deep learning is used to analyze the flight environment in real - time. The input of the model is the processed flight data, and the output is the environmental risk assessment value. The threshold is set to 8, and if the value exceeds this, it is determined as a high - risk area. Based on the assessment results, the flight strategy generation module uses the reinforcement learning algorithm and optimizes the flight path through the Q - learning method. The learning rate is set to 0.1, the discount factor is 0.9, and after 1000 iterations of training, the optimal flight strategy is generated. Finally, the updated flight strategy instructions are transmitted to the flight control system in real - time through the data communication module to ensure that the aircraft can execute tasks according to the latest strategy. For example, the flight speed is adjusted to 250 meters per second and the heading angle is 45 degrees to avoid high - risk areas and complete tasks efficiently.

[0082] The above content is only an example and explanation of the structure of the present invention. Those skilled in the art of this technology can make various modifications, supplements, or use similar methods to replace the described specific embodiments, as long as they do not deviate from the structure of the invention or exceed the scope defined by this claim book, they should fall within the protection scope of the present invention.

Claims

1. A situation awareness and assessment method based on dynamic data, characterized in that The method includes: Obtain an original data set, which includes UAV flight attitude data, environmental obstacle data, meteorological data, and target dynamic data; Judge whether the data quality of the original data set is lower than a preset threshold. If there is data missing, use an interpolation algorithm to fill in the missing values. If there is noise interference, use a filtering algorithm to denoise and generate a corrected data set; Based on the corrected data set, align multi-source data to a unified coordinate system through a preset spatio-temporal mapping model to generate a fused data set. Among them, the following spatio-temporal mapping model is used to map the spatio-temporal relationship of the corrected data set to obtain a mapped data set: Among them, M represents the final spatio-temporal mapping result, which is divided into 1...K data blocks, n represents the number of data sources, and ω i represents the weight coefficient of the i-th data source, T represents the spatio-temporal transformation function, and P i represents the position data of the i-th data source, and R i represents the rotation parameter of the i-th data source; Among them, F represents the fused dataset, K represents the number of data blocks to be fused, and β k represents the fusion weight of the k-th data block, S represents the spatial mapping function, H represents the temporal mapping function, X k and Y k respectively represent the spatial and temporal attributes of the k-th data block; Adjust the data collection frequency according to the dynamic characteristic parameters of the fused data set to generate a sampled optimized data set; Input the sampled optimized data set into an evaluation model to output a situation evaluation result including the environmental risk level. The sampled optimized data set includes the set of flight attitude, environmental obstacles, and meteorological data. The evaluation model is: Among them, R a is the attitude risk function, R o is the distance risk function, R m is the meteorological risk function, and R represents the comprehensive risk assessment value; w a , w o , w m respectively represent the weight coefficients of flight attitude, environmental obstacles, and meteorological data, and w a + w o + w m = 1; among them, Attitude risk function R a Satisfy: X j， pitch 、X j， roll 、X j， yaw respectively represent the actual pitch angle, roll angle and yaw angle of the UAV, calculate the deviations of these angles from the expected values, and take the sum of their squares as the risk index; w a1 、w a2 、w a3 are weight coefficients, indicating the contributions of each attitude parameter to the risk; Distance risk function R o Satisfy: ； Use the reciprocal of the relative distance D between the drone and the obstacle as the risk indicator. The closer the distance, the higher the risk. is a constant used to avoid division by zero; Meteorological risk function R m : , V w is the wind speed vector.

2. The method according to claim 1, characterized in that, It also includes: Based on the situation evaluation result and a preset decision rule library, use a deep reinforcement learning algorithm to generate a flight strategy instruction; If it is detected that the deviation between the execution data and the historical data exceeds the preset threshold, trigger the weighted fusion processing of the historical data and the current data to generate a completed data set; Based on the completed data set, re-execute the situation evaluation and flight strategy generation to update the flight strategy instruction.

3. The method according to claim 2, wherein The obtaining of the original data set, which includes UAV flight attitude data, environmental obstacle data, meteorological data, and target dynamic data, includes: Obtain UAV flight attitude data and smooth the flight attitude data using a Kalman filtering algorithm; Extract obstacle position information from the environmental obstacle data, combine the smoothed flight attitude data, and use a collision detection algorithm to judge the relative distance between the UAV and the obstacle; Obtain meteorological data, analyze wind speed, wind direction, and temperature parameters, and combine the UAV flight attitude data to establish a meteorological influence model.

4. The method according to claim 3, wherein The judging whether the data quality of the original data set is lower than a preset threshold. If there is data missing, use an interpolation algorithm to fill in the missing values. If there is noise interference, use a filtering algorithm to denoise and generate a corrected data set, includes: Obtain the original data set and judge the data quality in the original data set to determine whether it is lower than the preset threshold; If the data quality is lower than the preset threshold, further judge whether there is data missing in the data set; If there is data missing, use a linear interpolation algorithm to fill in the missing values; If the data quality is lower than the preset threshold and there is no data missing, judge whether there is noise interference in the data set; If there is noise interference, use a median filtering algorithm to denoise the data; According to the above processing results, generate a corrected data set, which contains the filled missing values and the denoised data.

5. The method according to claim 4, characterized in that, The based on the corrected data set, align multi-source data to a unified coordinate system through a preset spatio-temporal mapping model to generate a fused data set, includes: Obtain a multi-source data set, perform correction processing on the multi-source data set to obtain a corrected data set; Using a preset spatio-temporal mapping model, map the spatio-temporal relationships of the corrected data set to obtain a mapped data set; Based on the mapped data set, determine whether the data is aligned to a unified coordinate system. If not, remap it; Perform a fusion process on the aligned mapped data set to generate a fused data set.

6. The method according to claim 5, wherein The adjusting the data acquisition frequency according to the dynamic characteristic parameters of the fused data set to generate a sampled optimized data set includes: Based on the dynamic characteristic parameters of the fused data set, determine the range of characteristic values of the data set; For the range of characteristic values, calculate the change amplitude of the dynamicity to obtain the fluctuation range of the parameter values; Within the fluctuation range of the parameter values, determine the adjustment value of the acquisition rate to determine the optimized range of the sampled values; Based on the optimized range of the sampled values, calculate the adjustment amplitude of the frequency values to generate an optimized solution for the processed values; Adopt the optimized solution of the processed values to adjust the data acquisition frequency to obtain a sampled optimized data set; Based on the sampled optimized data set, extract the distribution law of the characteristic values and determine the adjustment range of the fused values; Within the adjustment range of the fused values, generate the final sampled optimized data set to complete the data processing flow.

7. The method according to claim 6, characterized in that, The inputting the sampled optimized data set into an evaluation model and outputting a situation evaluation result including the environmental risk level includes: For the sampled optimized data set, use the evaluation model for processing to obtain the situation evaluation result of the environmental risk level; If there is noise in the sampled data, remove the noise through a data preprocessing method to obtain a pure data set; Based on the pure data set, use the evaluation model for training to determine the model parameters; Obtain the trained model, predict new environmental data, and judge the risk level.

8. The method according to claim 7, wherein The generating a flight strategy instruction using a deep reinforcement learning algorithm based on the situation evaluation result and a preset decision rule library includes: Obtain the situation evaluation result and extract key feature data; Input the extracted feature data into the preset decision rule library to match applicable rules; Use a deep reinforcement learning algorithm to construct a strategy generation model; Fuse the matched decision rules in the strategy generation model to optimize the strategy generation process; Based on the optimized strategy generation model, calculate the flight strategy instruction; Adjust the parameters of the strategy generation model to improve the accuracy of the strategy instruction; Output the final flight strategy instruction for application to the flight control system.

9. The method according to claim 8, characterized in that, The triggering a weighted fusion process of historical data and current data to generate a complemented data set if it is detected that the deviation between the execution data and the historical data exceeds a preset threshold includes: Preset the deviation threshold of the historical data and obtain the deviation value between the execution data and the historical data; If the deviation value exceeds the preset threshold, trigger a weighted fusion process of historical data and current data; Use the weighting method to calculate the weight values of the historical data and the current data to determine the fusion ratio; Fuse the historical data and the current data according to the weight values through the fusion method to generate the complemented data.

10. The method according to claim 9, characterized in that The re-performing situation evaluation and flight strategy generation based on the complemented data set and updating the flight strategy instruction includes: Obtain the complemented data set, perform situation evaluation on the data set to obtain the current situation evaluation result; Based on the situation evaluation result, use a preset flight strategy generation algorithm to generate an initial flight strategy; If the difference between the initial flight strategy and the historical strategy in the supplemented dataset exceeds the preset threshold, adjust the strategy parameters and regenerate the flight strategy; Update the content of the flight instruction according to the adjusted flight strategy; Optimize the updated instruction using a machine learning algorithm to obtain an optimized flight instruction.

Images