An unmanned aerial vehicle-based rain enhancement environment detection path optimization method

Abstract

This invention relates to the field of UAV flight path optimization technology, specifically to a method for optimizing rain enhancement environment detection paths based on UAVs. The method includes: acquiring a three-dimensional grid set of the target airspace and initial atmospheric background field data; collecting real-time data on supercooled water content, vertical airflow velocity, and cloud droplet number concentration using the UAV; configuring Gaussian process regression prior parameters based on the initial background field and updating the joint prediction distribution with real-time data; obtaining the predicted mean and information entropy variance of the rain enhancement catalytic potential by transforming the joint prediction distribution through a nonlinear activation function probability density transformation; constructing a collection function that integrates potential gains and path energy consumption costs, and iteratively detecting the grid corresponding to its maximum value as the next target waypoint until the optimal catalytic zone is locked. This invention achieves high-efficiency, low-energy dynamic path optimization under limited onboard fuel constraints, significantly improving the positioning accuracy and operational success rate of cloud seeding windows.

Description

Technical Field

[0001] This invention relates to the field of drone flight path optimization technology, and more specifically, to a method for optimizing rain enhancement environmental detection paths based on drones. Background Technology

[0002] In artificial rain enhancement operations, using drones carrying cloud physics instruments to enter the target cloud system and find "catalytic potential zones" with abundant supercooled water content and continuous updrafts is a key prerequisite for improving rain enhancement efficiency. However, drones are limited by their onboard fuel capacity, resulting in extremely limited loiter time for detection; at the same time, the spatial distribution of meteorological elements within the cloud (such as liquid water content and vertical airflow velocity) is extremely uneven and highly uncertain.

[0003] Existing drone-based rain enhancement detection path planning methods typically rely on pre-set zigzag or spiral routes, lacking dynamic feedback based on real-time environmental perception. This can easily lead to drones consuming large amounts of onboard fuel in low-potential areas, thus missing the optimal cloud seeding window.

[0004] Therefore, how to dynamically optimize the detection path using real-time data collected by UAVs under limited airborne fuel constraints, and quickly locate the best rain enhancement catalysis area in the cloud with the highest probability and lowest energy consumption, is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] The purpose of this invention is to provide a method for optimizing the detection path of rain enhancement environment based on unmanned aerial vehicles (UAVs) to solve the above-mentioned technical problems.

[0006] To achieve the above objectives, the embodiments of this application provide the following technical solutions: This application provides a method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles (UAVs), the method comprising: A set of three-dimensional grid indexes of the target airspace after rasterization is obtained. After the UAV reaches the edge area of ​​the target airspace, the UAV is controlled and its onboard meteorological sensors are used to randomly collect real-time atmospheric parameter data of at least three grid cells corresponding to the edge area. The real-time atmospheric parameter data includes supercooled water content, vertical airflow velocity and cloud droplet number concentration. Based on the collected real-time atmospheric parameter data, the mean of the predicted rain enhancement catalytic potential and the information entropy variance characterizing the uncertainty of the mean of the predicted rain enhancement catalytic potential are updated through a Gaussian process regression model for each grid cell in the three-dimensional grid index set. Based on the updated predicted mean of rain enhancement catalytic potential and the variance of information entropy, and combined with the path energy cost parameters from the current position of the UAV to each undetected grid cell, a data acquisition function is constructed, and the corresponding data acquisition function response value is calculated for each undetected grid cell using the data acquisition function. The grid cell corresponding to the maximum value in the response value of the acquisition function is taken as the next target waypoint, and a control command is generated to fly from the current position to the next target waypoint; In response to the predicted average value of the rain enhancement catalytic potential reaching the preset catalytic operation threshold, path optimization is stopped and the current grid cell coordinates are output to execute the rain enhancement operation.

[0007] Optionally, each grid cell in the three-dimensional grid index set corresponds to a geospatial coordinate and initial atmospheric background field data; The initial atmospheric background field data is a three-dimensional gridded meteorological element background value generated by interpolation based on numerical weather prediction models and historical sounding data. The initial atmospheric background field data is used to configure the prior mean parameter and covariance kernel function scale parameter of the Gaussian process regression model. Then, in response to the real-time atmospheric parameter data of the grid cells, the real-time atmospheric parameter data is used as the observation sample points and input into the configured Gaussian process regression model to generate a joint predicted distribution for the supercooled water content value and the vertical airflow velocity value.

[0008] Optionally, after generating the joint predicted distribution for the subcooled water content value and the vertical airflow velocity value, the method further includes: The joint prediction distribution is input into a nonlinear activation function mapping module based on cloud physics threshold features. The nonlinear activation function mapping module performs a probability density transformation on the joint prediction distribution to obtain the mean of the rain enhancement catalytic potential prediction for each grid cell and the information entropy variance characterizing the uncertainty of the mean of the rain enhancement catalytic potential prediction.

[0009] Optionally, the acquisition function is a function model for improving the algorithm, and the response value of the acquisition function is positively correlated with the excess of the predicted mean of the rain enhancement catalytic potential and the information entropy variance, and negatively correlated with the path energy consumption cost parameter.

[0010] Optionally, the process of generating the path energy consumption cost parameters includes: Obtain the estimated remaining range of the UAV, calculate the spatial Euclidean distance between the current position grid cell and the candidate target grid cell, and weight and sum the spatial Euclidean distance with the energy compensation correction term corresponding to the vertical airflow velocity observation value to generate a path energy cost parameter that characterizes the estimated flight energy consumption required to fly from the current position to the candidate target grid cell.

[0011] The beneficial effects of this invention are as follows: This invention utilizes the technical features of "updating the predicted mean and information entropy variance of rain enhancement catalytic potential corresponding to each grid cell through a Gaussian process regression model based on collected real-time atmospheric parameter data" and "randomly collecting at least 3 edge area grid cells as initial observation samples." This allows the UAV to continuously integrate newly added multi-dimensional real-time atmospheric parameter data during the detection process, dynamically update the spatial distribution estimate of rain enhancement catalytic potential across the entire airspace and its uncertainty perception, thereby eliminating the dependence on preset fixed routes and realizing data-driven adaptive iterative optimization of the detection path, significantly reducing the probability of wasting fuel in low-potential areas.

[0012] Secondly, based on the updated predicted mean of rain enhancement catalytic potential and the variance of information entropy, combined with the path energy cost parameters from the UAV's current position to each undetected grid cell, a data acquisition function is constructed. Furthermore, the technical features of "the response value of the data acquisition function being positively correlated with the excess of the predicted mean of rain enhancement catalytic potential and the variance of information entropy, and negatively correlated with the path energy cost parameters" enable the construction of a data acquisition function that integrates detection gains and movement costs. When making decisions about the next target grid, this function simultaneously suppresses the greedy tendency to simply pursue high-potential targets while ignoring long-distance flight energy consumption. This ensures that the UAV, under limited onboard fuel constraints, always rapidly approaches the optimal catalytic region along the optimal direction of "high potential gains - low energy cost," balancing rain enhancement detection efficiency with the need for safe return flight.

[0013] Secondly, by "inputting the joint prediction distribution into a nonlinear activation function mapping module based on cloud physical threshold features and performing probability density transformation to obtain the predicted mean and information entropy variance of rain enhancement catalytic potential," the potential assessment strictly follows the cloud precipitation physical law that "significant catalytic value is only achieved after factors such as supercooled water content and vertical airflow velocity cross physical thresholds," avoiding misjudgments caused by linear weighted assessment. Combined with the positive correlation excitation of information entropy variance in the acquisition function, the UAV is driven to prioritize detection at grid points with high potential uncertainty and high expected returns, maximizing the reduction of uncertainty in the spatial distribution of catalytic potential with the fewest additional detections, enabling the system to quickly converge and lock the true optimal cloud seeding location within a very limited remaining flight window.

[0014] Secondly, by using the technical feature of "using initial atmospheric background field data to configure the prior mean parameters and covariance kernel function scaling parameters of the Gaussian process regression model", the prior knowledge of atmospheric state provided by numerical weather prediction models and historical sounding data is injected into the statistical learning model. This provides the UAV with an initial estimated spatial field structure that conforms to meteorological and physical laws during the cold start stage at the edge of the cloud, significantly reducing the convergence time and fuel consumption required to start the detection from a zero random field.

[0015] Other features and advantages of the invention will be set forth in the following description and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of a method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles (UAVs) as described in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0019] Example:

[0020] This embodiment provides a method for optimizing the path of rain enhancement environmental detection based on UAVs, which is applied to rain enhancement detection operations targeting stratified mixed clouds. The UAV is equipped with an airborne microwave radiometer (which receives the microwave radiation brightness temperature emitted by supercooled water droplets in the cloud and outputs the observed value of supercooled water content of the grid cell at the current location through inversion calculation), a cloud particle spectrum probe (which performs single-particle counting and particle size measurement of cloud droplets through forward laser scattering principle and outputs the observed value of cloud droplet number concentration through integration statistics), and an airborne anemometer (which combines the UAV attitude and ground speed data provided by the inertial navigation system to separate the three-dimensional motion velocity of the atmosphere through vector wind speed solution and takes its vertical component as the observed value of vertical airflow velocity).

[0021] like Figure 1As shown, this embodiment provides a method for optimizing the rain enhancement environment detection path based on UAVs. The method includes steps S100, S200, S300, S400 and S500.

[0022] Step S100: Obtain a three-dimensional grid index set of the target airspace after rasterization. Each grid cell in the three-dimensional grid index set corresponds to a geospatial coordinate and initial atmospheric background field data. After the UAV reaches the edge region of the target airspace, control the UAV and, in conjunction with its onboard meteorological sensors, randomly collect real-time atmospheric parameter data of at least three grid cells corresponding to the edge region. The real-time atmospheric parameter data includes supercooled water content, vertical airflow velocity, and cloud droplet number concentration. Specifically, this can be implemented as follows: Before flight, the ground control system divides the target airspace (horizontal range 10km × 10km, vertical range 1500m to 4500m) into several three-dimensional grid cells. Each grid cell is set to 500m × 500m × 200m (length × width × height), generating a three-dimensional grid index set. .

[0023] The ground control system extracts forecast products for the target airspace from numerical weather prediction models (such as the WRF model) and obtains each grid cell through three-dimensional linear interpolation. The corresponding initial atmospheric background field data. In this embodiment, the initial atmospheric background field data includes: an estimated value of background supercooled water content. (unit: Background vertical airflow velocity estimate (unit: ) and background horizontal wind vector (For subsequent energy consumption calculations).

[0024] Step S200: Based on the collected real-time atmospheric parameter data, update the predicted mean of the rain enhancement catalytic potential corresponding to each grid cell in the three-dimensional grid index set and the information entropy variance characterizing the uncertainty of the predicted mean of the rain enhancement catalytic potential using a Gaussian process regression model. Specifically: Step S210: The initial atmospheric background field data is a three-dimensional gridded meteorological element background value generated by interpolation based on numerical weather prediction models and historical radiosonde data. The initial atmospheric background field data is used to configure the prior mean parameter and covariance kernel function scale parameter of the Gaussian process regression model. Then, in response to the real-time atmospheric parameter data of the grid cells, the real-time atmospheric parameter data is used as the observation sample points and input into the configured Gaussian process regression model to generate a joint predicted distribution for the supercooled water content value and the vertical airflow velocity value. The initial atmospheric background field data is used to configure the prior mean parameter and covariance kernel function scaling parameter of the Gaussian process regression model. In this embodiment, the specific implementation can be as follows: [The text abruptly shifts to a different topic] ...specifically targeting the supercooled water content... Two independent Gaussian process regression models were established with respect to the vertical airflow velocity w. Each model consists of a mean function. and covariance kernel function The only certainty.

[0025] For the supercooled water content prediction model, its prior mean function is set as follows: ; in, For grid cells The prior mean of the Gaussian regression model for the subcooled water content. Grid cells obtained from interpolation of numerical weather prediction models Estimated background subcooled water content at the location.

[0026] For the vertical airflow velocity prediction model, its prior mean function is set as follows: ; in, For grid cells The prior mean of the regression model for the vertical airflow velocity Gaussian process. Grid cells obtained from interpolation of numerical weather prediction models The estimated value of the background vertical airflow velocity at that location.

[0027] Both models use the same form of covariance kernel function, namely the squared exponential kernel function: ; Represents any two grid cells and The covariance between them; and They are grid cells and The three-dimensional coordinate vector in space; This represents the Euclidean distance between two grid cells; The signal variance characterizes the overall spatial variability of a physical quantity; This is a spatial scale parameter that characterizes the attenuation distance of the spatial correlation of physical quantities.

[0028] With the above configuration, the initial atmospheric background field data is explicitly embedded into the prior mean of the Gaussian process regression, providing an initial spatial distribution trend that conforms to meteorological laws for subsequent real-time data updates.

[0029] After taking off and entering the edge grid cell of the cloud, the drone hovers and collects data. The onboard sensors analyze and output the following real-time atmospheric parameter data: Supercooled water content observation value Vertical airflow velocity observations and cloud droplet number concentration observations .

[0030] The drone flight control computer will contain a sample set of the above observations: ; Input two Gaussian process regression models for the subcooled water content and vertical airflow velocity, respectively.

[0031] The current cloud edge grid cell is denoted as the current grid. The supercooled water content model outputs the predicted mean supercooled water content for each undetected grid cell within the target airspace. With prediction variance The vertical airflow velocity model outputs the predicted mean of vertical airflow velocity. With prediction variance The outputs of the two models together constitute the joint predicted distribution of the physical quantities of the grid cell. In this embodiment, it is assumed that the supercooled water content and the vertical airflow velocity are independent of each other under given spatial conditions; therefore, the joint distribution can be represented by the product of their respective marginal distributions.

[0032] Step S220: Input the joint prediction distribution into a nonlinear activation function mapping module based on cloud physics threshold features. Perform probability density transformation on the joint prediction distribution through the nonlinear activation function mapping module to obtain the mean of the rain enhancement catalytic potential prediction for each grid cell and the information entropy variance characterizing the uncertainty of the mean of the rain enhancement catalytic potential prediction. Specifically, this can be implemented as follows: Because the relationship between a simple physical quantity value and the rain enhancement catalytic effect is not linear (for example, supercooled water only has a value greater than...), (Only then can it have significant catalytic significance). This embodiment introduces a nonlinear activation function mapping module based on cloud physical threshold features.

[0033] This module has a built-in S-shaped growth curve function, and the specific calculation formula is as follows: ; in, The catalytic threshold is the amount of supercooled water. The effective threshold for updrafts; The reference cloud droplet number concentration is used; α, β, and k are the sensitivity coefficients contributed by the supercooled water content, the vertical airflow velocity, and the steepness adjustment coefficient of the S-shaped growth curve, respectively. .

[0034] For grid cells The rain enhancement potential value at the location is dimensionless and ranges from 0 to 1; For grid cells Cloud droplet number concentration value at the location; It is the arctangent function; For grid cells Vertical airflow velocity value at location, in units ; For grid cells The value of subcooled water content at the location, in units of .

[0035] Note the input here. and It is a random variable that follows a joint prediction distribution. (This is used to calculate the grid cell.) Rain enhancement catalytic potential value at the location predicted mean of rain enhancement catalytic potential and information entropy variance This embodiment employs the unscented transformation method. Using the aforementioned joint prediction distribution (mean vector and covariance matrix), a set of Sigma sampling points is generated. Each sampling point is then substituted into the aforementioned nonlinear activation function for calculation to obtain the corresponding rain enhancement catalytic potential sampling value. Finally, statistical reconstruction is used to obtain the predicted mean and information entropy variance of the rain enhancement catalytic potential.

[0036] Step S300: Based on the updated rain enhancement catalytic potential prediction mean and information entropy variance, and combined with the path energy consumption cost parameter from the current position of the UAV to each undetected grid cell, a collection function is constructed, and the corresponding collection function response value is calculated for each undetected grid cell. The collection function response value is positively correlated with the excess of the rain enhancement catalytic potential prediction mean and the information entropy variance, and negatively correlated with the path energy consumption cost parameter.

[0037] The process of generating the path energy consumption cost parameters includes: Obtain the estimated remaining range of the UAV, calculate the spatial Euclidean distance between the current location grid cell and the candidate target grid cell, and weight and sum the spatial Euclidean distance with the energy compensation correction term corresponding to the observed vertical airflow velocity to generate a path energy cost parameter characterizing the estimated flight energy consumption required to fly from the current location to the candidate target grid cell. The specific calculation method can be as follows: To ensure the safe return of the drone, it is necessary to calculate the data from the current grid. Head to candidate target grid cell The estimated flight energy consumption.

[0038] The path energy consumption cost parameter The calculation formula is: ; in, The spatial Euclidean distance between the two points; The background horizontal wind vector of the flight direction and the target grid cell The included angle; This refers to the drone's cruising speed. These are the baseline energy consumption weighting coefficient, which is proportional to distance, and the additional energy consumption weighting coefficient, which is proportional to wind resistance. Essentially, this parameter characterizes the estimated distance of equivalent energy consumption required for the UAV to overcome air resistance and wind field influences.

[0039] Construct a data acquisition function based on the desired improvement algorithm, in the following form: ; in, For candidate target mesh cells The calculated response value of the acquisition function; For target mesh cells The predicted average of the rain enhancement catalytic potential at the location; For target mesh cells The standard deviation of information entropy at a given location is the square root of the variance of information entropy. This represents the maximum value of the predicted average of the rain enhancement catalytic potential in the currently detected grid cells; ; It is the standard normal cumulative distribution function; It is the standard normal probability density function; It is a tiny positive number, used to avoid the denominator being zero. To standardize and improve quantities.

[0040] The numerator of the acquisition function encourages the UAV to travel to areas with high prediction potential or high uncertainty, while the denominator penalizes long-distance flights with high energy consumption. The flight control computer traverses all unexplored grids and calculates a series of acquisition function response values.

[0041] Step S400: The grid cell corresponding to the maximum value in the acquisition function response value is taken as the next target waypoint, and a control command is generated to fly from the current position to the next target waypoint. Specifically, the implementation method can be as follows: Find the maximum value in the set of acquisition function response values ​​to obtain the corresponding next target waypoint. Before generating control commands to the next destination waypoint, the system performs safety verification logic: Read the current remaining available fuel from the energy management system. .

[0042] like ( If the safety factor is set to 0.4, then the target waypoint is determined to have a risk of crashing.

[0043] At this point, the maximum value grid is abandoned, and the second largest grid cell that satisfies the energy consumption constraint is selected as the corrected next target waypoint based on the sorting results of the acquisition function response value. The corresponding flight control command is generated and sent to the UAV flight control system for execution.

[0044] Step S500: In response to the predicted average value of the rain enhancement catalytic potential reaching the preset catalytic operation threshold, path optimization is stopped and the current grid cell coordinates are output to execute the rain enhancement operation. Specifically, this can be implemented as follows: After the drone arrives at the new target waypoint, repeat steps S200 to S400.

[0045] During this process, as the number of detection points increases, the predicted mean of the rain enhancement catalytic potential of the target airspace continuously approaches the actual distribution. When the predicted mean of the rain enhancement catalytic potential of any grid cell is greater than or equal to the preset catalytic operation threshold (e.g., 0.85), the system determines that the optimal cloud seeding window has been locked, immediately stops the path optimization loop, outputs the three-dimensional coordinates of the grid cell, and prompts the ground operator to perform the silver iodide flame ignition operation or trigger the UAV automatic seeding device.

[0046] Through the above implementation methods, the present invention ensures that the UAV always flies along the optimal path of "high potential benefit - low detection cost" in the dynamically changing cloud environment, which significantly improves the success rate of rain enhancement operations and the safety of UAV flight.

[0047] This embodiment utilizes the technical features of "updating the predicted mean and information entropy variance of rain enhancement catalytic potential for each grid cell using a Gaussian process regression model based on collected real-time atmospheric parameter data" and "randomly collecting at least three edge area grid cells as initial observation samples." This allows the UAV to continuously integrate newly added multi-dimensional real-time atmospheric parameter data during the detection process, dynamically updating the spatial distribution estimate of rain enhancement catalytic potential across the entire airspace and its uncertainty perception. This frees it from dependence on preset fixed routes, achieving data-driven adaptive iterative optimization of the detection path and significantly reducing the probability of fuel consumption in low-potential areas.

[0048] Secondly, based on the updated predicted mean of rain enhancement catalytic potential and the variance of information entropy, and combined with the path energy cost parameters from the UAV's current position to each undetected grid cell, a data acquisition function is constructed. This function is characterized by the fact that the response value of the data acquisition function is positively correlated with the excess of the predicted mean of rain enhancement catalytic potential and the variance of information entropy, and negatively correlated with the path energy cost parameters. This results in a data acquisition function that integrates detection gains and movement costs. When making decisions about the next target grid, this function simultaneously suppresses the greedy tendency to simply pursue high-potential targets while ignoring long-distance flight energy consumption. It ensures that the UAV, under limited fuel constraints, always rapidly approaches the optimal catalytic region along the optimal direction of "high potential gain - low energy cost," balancing rain enhancement detection efficiency with the need for safe return flight.

[0049] Secondly, by "inputting the joint prediction distribution into a nonlinear activation function mapping module based on cloud physical threshold features and performing probability density transformation to obtain the predicted mean and information entropy variance of rain enhancement catalytic potential," the potential assessment strictly follows the cloud precipitation physical law that "significant catalytic value is only achieved after factors such as supercooled water content and vertical airflow velocity cross physical thresholds," avoiding misjudgments caused by linear weighted assessment. Combined with the positive correlation excitation of information entropy variance in the acquisition function, the UAV is driven to prioritize detection at grid points with high potential uncertainty and high expected returns, maximizing the reduction of uncertainty in the spatial distribution of catalytic potential with the fewest additional detections, enabling the system to quickly converge and lock the true optimal cloud seeding location within a very limited remaining flight window.

[0050] Secondly, by using the technical feature of "using initial atmospheric background field data to configure the prior mean parameters and covariance kernel function scaling parameters of the Gaussian process regression model", the prior knowledge of atmospheric state provided by numerical weather prediction models and historical sounding data is injected into the statistical learning model. This provides the UAV with an initial estimated spatial field structure that conforms to meteorological and physical laws during the cold start stage at the edge of the cloud, significantly reducing the convergence time and fuel consumption required to start the detection from a zero random field.

[0051] Similarly, the solution in this embodiment can also be adapted to purely electric-powered drones, and is not limited thereto. The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. 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.

Claims

1. A method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles (UAVs), characterized in that, The method includes: A set of three-dimensional grid indexes of the target airspace after rasterization is obtained. After the UAV reaches the edge area of ​​the target airspace, the UAV is controlled and its onboard meteorological sensors are used to randomly collect real-time atmospheric parameter data of at least three grid cells corresponding to the edge area. The real-time atmospheric parameter data includes supercooled water content, vertical airflow velocity and cloud droplet number concentration. Based on the collected real-time atmospheric parameter data, the mean of the predicted rain enhancement catalytic potential and the information entropy variance characterizing the uncertainty of the mean of the predicted rain enhancement catalytic potential are updated through a Gaussian process regression model for each grid cell in the three-dimensional grid index set. Based on the updated predicted mean of rain enhancement catalytic potential and the variance of information entropy, and combined with the path energy cost parameters from the current position of the UAV to each undetected grid cell, a data acquisition function is constructed, and the corresponding data acquisition function response value is calculated for each undetected grid cell using the data acquisition function. The grid cell corresponding to the maximum value in the response value of the acquisition function is taken as the next target waypoint, and a control command is generated to fly from the current position to the next target waypoint; In response to the predicted average value of the rain enhancement catalytic potential reaching the preset catalytic operation threshold, path optimization is stopped and the current grid cell coordinates are output to execute the rain enhancement operation.

2. The method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Each grid cell in the three-dimensional grid index set corresponds to a geospatial coordinate and initial atmospheric background field data; The initial atmospheric background field data is a three-dimensional gridded meteorological element background value generated by interpolation based on numerical weather prediction models and historical sounding data. The initial atmospheric background field data is used to configure the prior mean parameter and covariance kernel function scale parameter of the Gaussian process regression model. Then, in response to the real-time atmospheric parameter data of the grid cells, the real-time atmospheric parameter data is used as the observation sample points and input into the configured Gaussian process regression model to generate a joint predicted distribution for the supercooled water content value and the vertical airflow velocity value.

3. The method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles (UAVs) according to claim 2, characterized in that, After generating the joint predicted distribution for the supercooled water content value and the vertical airflow velocity value, the method further includes: The joint prediction distribution is input into a nonlinear activation function mapping module based on cloud physics threshold features. The nonlinear activation function mapping module performs a probability density transformation on the joint prediction distribution to obtain the mean of the rain enhancement catalytic potential prediction for each grid cell and the information entropy variance characterizing the uncertainty of the mean of the rain enhancement catalytic potential prediction.

4. The method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles according to claim 3, characterized in that, The acquisition function is a function model for the algorithm to be improved. The response value of the acquisition function is positively correlated with the excess of the predicted mean of the rain enhancement catalytic potential and the information entropy variance, and negatively correlated with the path energy consumption cost parameter.

5. The method for optimizing the rain enhancement environment detection path based on unmanned aerial vehicles (UAVs) according to claim 4, characterized in that, The process of generating the path energy consumption cost parameters includes: Obtain the estimated remaining range of the UAV, calculate the spatial Euclidean distance between the current position grid cell and the candidate target grid cell, and weight and sum the spatial Euclidean distance with the energy compensation correction term corresponding to the vertical airflow velocity observation value to generate a path energy cost parameter that characterizes the estimated flight energy consumption required to fly from the current position to the candidate target grid cell.

Images