Unmanned aerial vehicle nest control method and system

Abstract

The invention discloses an unmanned aerial vehicle nest control method and system, and the method comprises the steps: obtaining the first configuration data of a flying unmanned aerial vehicle, and obtaining the local configuration data of an unmanned aerial vehicle nest matched with the flying unmanned aerial vehicle; determining second configuration data of the flying unmanned aerial vehicle according to the target task request, and performing comparative analysis on the first configuration data and the second configuration data to obtain a to-be-replaced configuration index; screening the local configuration data based on the to-be-replaced configuration index, and controlling the unmanned aerial vehicle nest to execute an internal configuration component transfer signal according to a screening result, so that the unmanned aerial vehicle nest reaches an initial switching state; acquiring first flight pose data of the flying unmanned aerial vehicle; predicting a drop point position of the flying unmanned aerial vehicle based on the first flight pose data to obtain a predicted drop point coordinate; and controlling the unmanned aerial vehicle nest to execute a target switching state signal generated by the initial switching state and the predicted drop point coordinates. According to the method, powerful support is provided for multi-task application of the unmanned aerial vehicle.

Description

Technical Field

[0001] This application relates to the technical field of unmanned aerial vehicles, and particularly to a method and system for controlling an unmanned aerial vehicle nest. Background Art

[0002] With the wide application of unmanned aerial vehicle technology, when unmanned aerial vehicles perform diverse tasks, they need to switch tasks and dock with the nest efficiently and accurately.

[0003] In the traditional mode, the unmanned aerial vehicle only starts to execute the deployment work for the next task after returning to the nest. This mode is applicable to simple scenarios with relatively single task types and low switching frequencies. However, in today's complex and changeable real environment with highly diverse task requirements, especially when task switching is extremely frequent, it may lead to task delays and resource waste.

[0004] Therefore, how to optimize the control method of the unmanned aerial vehicle nest to improve the response speed of the unmanned aerial vehicle nest has become a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] This application provides a method and system for controlling an unmanned aerial vehicle nest to solve the technical problem of how to optimize the control method of the unmanned aerial vehicle nest, so as to achieve the effect of improving the response speed of the unmanned aerial vehicle nest.

[0006] To solve the above technical problem, an embodiment of this application provides a method for controlling an unmanned aerial vehicle nest, which is characterized by including: In response to a target task request, obtain the first configuration data of the flying unmanned aerial vehicle, and obtain the local configuration data of the unmanned aerial vehicle nest that matches the flying unmanned aerial vehicle; According to the target task request, determine the second configuration data of the flying unmanned aerial vehicle, and perform comparison and analysis on the first configuration data and the second configuration data based on natural language technology to obtain the configuration indicators to be replaced of the flying unmanned aerial vehicle; Based on the configuration indicators to be replaced, screen the local configuration data, and control the unmanned aerial vehicle nest to execute the internal configuration component transfer signal according to the screening result, so that the unmanned aerial vehicle nest reaches the initial switching state; When the position information of the flying unmanned aerial vehicle is detected to meet the preset range, obtain the first flight pose data captured by the first sensor provided on the flying unmanned aerial vehicle; Predict the landing position of the flying unmanned aerial vehicle based on the first flight pose data to obtain the predicted landing coordinates; Control the unmanned aerial vehicle nest to execute the target switching state signal generated by the initial switching state and the predicted landing coordinates.

[0007] As one of the preferred solutions, after obtaining the first configuration data of the flying drone, it further includes: Obtain the original data of the priority level, remaining time, and completion progress of the current task of the flying drone, and normalize the original data to obtain the current task feature vector; Match the current task feature vector with a pre-constructed decision rule table to calculate the switching matching degree of the current task; If the switching matching degree is greater than a preset task switching threshold, it is determined that task switching is required.

[0008] As one of the preferred solutions, the determining the second configuration data of the flying drone according to the target task request includes: Input the target task request into a pre-trained large language model for parsing to obtain a parsing result; Match the parsing result with a pre-constructed knowledge graph, and extract the flight data, sensor data, communication data, and mounting data corresponding to the target task request from the knowledge graph according to the matching result; Generate the second configuration data according to the flight data, the sensor data, the communication data, and the mounting data.

[0009] As one of the preferred solutions, the predicting the landing point position of the flying drone based on the first flight pose data to obtain a predicted landing point coordinate includes: Obtain the second flight pose data captured by a second sensor provided in the drone nest; Perform weighted fusion on the first flight pose data and the second flight pose data to calculate the fused pose data of the flying drone; Input the fused pose data into a pre-trained drone attitude model for prediction to obtain the predicted landing point coordinate of the flying drone.

[0010] As one of the preferred solutions, the obtaining the second flight pose data captured by the second sensor provided in the drone nest includes: Obtain the image data of the flying drone captured by visual sensors provided in each direction of the drone nest, perform image enhancement and denoising processing on the image data to obtain a preprocessed image; Analyze the preprocessed image based on image processing technology to obtain the six-degree-of-freedom attitude of the flying drone; Construct an image sequence of the flying drone according to the preprocessed image, and analyze the image sequence based on the optical flow method to obtain the motion trajectory of the flying drone; Generate the second flight pose data according to the six - degree - of - freedom pose and the motion trajectory.

[0011] As one of the preferred solutions, the step of performing weighted fusion on the first flight pose data and the second flight pose data to calculate the fused pose data of the flying drone includes: Construct a distance weight formula from the flying drone to the drone nest, expressed as: Where, is a monotonically decreasing function with respect to the distance , is the rate of weight change, is the distance threshold, is the weight of the first flight pose data, is the weight of the second flight pose data; Fuse the first flight pose data and the second flight pose data according to the distance weight formula to obtain the fused pose data of the flying drone.

[0012] As one of the preferred solutions, before inputting the fused pose data into the pre - trained drone pose model for prediction, it further includes: Obtain the historical landing data of the flying drone, where the historical landing data includes historical flight pose data and historical landing position data; Perform weighted processing on the historical flight pose data based on the distance between the flying drone and the drone nest to obtain a first sample data set; Construct an initial drone pose model based on the LSTM network, input the first sample data set into the initial drone pose model, and construct a loss function according to the obtained output result and the historical landing position data; Optimize the initial drone pose model based on the loss function to obtain a trained drone pose model.

[0013] As one of the preferred solutions, the step of controlling the drone nest to execute the target switching state signal generated by the initial switching state and the predicted landing point coordinates includes: Construct an objective function that minimizes the landing deviation of the flying drone according to the predicted landing point coordinates; Solve the objective function based on the dynamic programming solution algorithm, and optimize the solution process through pruning strategies and approximate dynamic programming methods to obtain the optimal decision of the drone nest; Generate the control scheme for the drone nest according to the optimal decision, and control the drone nest to execute the target switching state signal generated by the control scheme.

[0014] As one of the preferred solutions, after generating the control scheme for the drone nest according to the optimal decision, it further includes: Calculate the predicted landing time of the flying drone to the predicted landing point coordinates according to the first flight pose data; Compare the predicted landing time with the adjustment time of the obtained control scheme, and control the speed of the flying drone according to the comparison result to make the predicted landing time match the adjustment time.

[0015] Another embodiment of the present application provides a drone nest control system, including: An acquisition module, configured to acquire the first configuration data of the flying drone in response to a target task request, and acquire the local configuration library of the drone nest matching the flying drone; An analysis module, configured to determine the second configuration data of the flying drone according to the target task request, and perform comparison and analysis on the first configuration data and the second configuration data respectively based on natural language technology to obtain the configuration indicators to be replaced of the flying drone; A mobilization module, configured to screen in the local configuration data based on the configuration indicators to be replaced, and control the drone nest to execute the internal configuration component mobilization signal according to the screening result to make the drone nest reach the initial switching state; A detection module, configured to acquire the first flight pose data captured by the first sensor provided on the flying drone when the position information of the flying drone is detected to meet the preset range; A prediction module, configured to predict the landing point position of the flying drone based on the first flight pose data to obtain the predicted landing point coordinates; A control module, configured to control the drone nest to execute the target switching state signal generated by the initial switching state and the predicted landing point coordinates.

[0016] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following: 1) By predicting the landing point position of the flying drone, the present application generates the target switching state signal in advance, enabling the nest to complete the corresponding adjustments and preparations when the drone is about to land, further shortening the time interval from the drone landing to being able to execute the next task, and overall improving the response speed of the drone system to tasks.

[0017] 2) From data acquisition, analysis and comparison, control of the drone nest component mobilization to prediction of the landing point and generation of the target switching status signal, the entire process of this application requires little manual participation. The drone nest can autonomously complete a series of operations, achieving the automation of drone task switching and drone nest docking, reducing the influence of human factors, and improving the stability and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic flow chart of a drone nest control method in one embodiment of this application; Figure 2 is a schematic structural diagram of a drone nest control system in one embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Next, the technical solutions in the embodiments of this application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. The purpose of providing these embodiments is to make the disclosure of this application more thorough and comprehensive. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the scope of protection of this application.

[0020] In the description of this application, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", "third", etc. may explicitly or implicitly include one or more of such features. In the description of this application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0021] In the description of this application, it should be noted that unless otherwise clearly defined and limited, the terms "installed", "connected", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, and it can be the internal communication of two components. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are only for the purpose of illustration and do not indicate or imply that the device or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation of this application. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0022] In the description of this application, it should be noted that unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0023] An embodiment of this application provides a method for controlling a drone nest. Specifically, please refer to Figure 1 , Figure 1 which shows a schematic flow diagram of the method for controlling a drone nest in one of the embodiments of this application, and it includes steps S1 - S6: S1: In response to a target task request, obtain the first configuration data of the flying drone, and obtain the local configuration data of the drone nest that matches the flying drone; In this step, the first configuration data includes information such as the current flight parameters, sensor status, communication settings, and mounted devices of the drone, and these data reflect the self - state of the drone before receiving a new task request. The drone nest that matches the flying drone also has its local configuration data, including information such as the location of the nest, charging capacity, maintenance equipment status, and storage capacity. Obtaining these data helps the system comprehensively consider the task execution environment and support conditions of the drone.

[0024] Preferably, in an embodiment of this application, after obtaining the first configuration data of the flying drone, it further includes: Obtain the original data of the priority level, remaining time, and completion progress of the current task of the flying drone, and normalize the original data to obtain the current task feature vector; Match the current task feature vector with a pre - constructed decision rule table, and calculate the switching matching degree of the current task; If the switching matching degree is greater than a preset task switching threshold, it is determined that task switching is required.

[0025] Among them, the priority level indicates the importance of the task, the remaining time reflects the time urgency of the task, and the completion progress reflects the execution degree of the task. Since different types of data (such as priority level, remaining time, and completion progress) may have different dimensions and value ranges, in order to facilitate unified analysis and processing, these original data are normalized. The normalized data is integrated into a current task feature vector, and this vector can comprehensively reflect the state characteristics of the current task.

[0026] Match the obtained current task feature vector with the pre - constructed decision rule table. The decision rule table is pre - constructed based on certain business logics and experiences, which stipulates the relationship between different task feature vectors and the possibility of task switching. Through the matching process, calculate the switching matching degree of the current task. This matching degree quantifies the possibility of the current task being switched. For example, if the decision rule table stipulates that when the task feature vector is within a certain specific range, the task switching matching degree is 0.8, then when the current task feature vector matches it, a switching matching degree of 0.8 will be obtained. The higher the matching degree, the more likely the current task needs to be switched.

[0027] Compare the calculated switching matching degree with the preset task switching threshold. If the switching matching degree is greater than the preset threshold, this indicates that the situation of the current task meets the pre - set conditions for task switching, so it is determined that task switching is required; otherwise, the execution of the current task is maintained. For example, in a specific unmanned aerial vehicle application scenario, the business logic may be that if the current task has a low priority, a long remaining time, and a high completion progress, a new high - priority task is allowed to preempt the current task, that is, task switching is performed.

[0028] S2: According to the target task request, determine the second configuration data of the flying unmanned aerial vehicle, and based on natural language technology, compare and analyze the first configuration data and the second configuration data to obtain the configuration indicators to be replaced for the flying unmanned aerial vehicle; Preferably, in an embodiment of the present application, determining the second configuration data of the flying unmanned aerial vehicle according to the target task request includes: Input the target task request into a pre - trained large - language model for parsing to obtain a parsing result; Match the parsing result with the pre - constructed knowledge graph, and extract the flight data, sensor data, communication data, and mounting data corresponding to the target task request from the knowledge graph according to the matching result; Generate the second configuration data according to the flight data, sensor data, communication data, and mounting data.

[0029] Among them, the large - language model has strong natural language understanding ability, and can perform semantic analysis, keyword extraction, etc. on the text information in the task request to obtain the parsing result. This parsing result initially understands the user's requirements for the target task.

[0030] The parsing results obtained by the large language model are matched with the pre - constructed knowledge graph. The knowledge graph is a structured semantic knowledge base that organizes and stores various knowledge in a graphical structure, containing rich knowledge information about drone flight, sensors, communication, payloads, etc., as well as the relationships between them. According to the matching results, flight data (such as flight altitude, speed, flight route, etc.), sensor data (such as sensor type, parameter settings, etc.), communication data (such as communication frequency, bandwidth, etc.), and payload data (such as payload device type, function, etc.) corresponding to the target task request are extracted from the knowledge graph.

[0031] Based on the extracted flight data, sensor data, communication data, and payload data, second configuration data for the target task is generated. This set of data will be used to guide the drone to adjust its own state to meet the requirements of the new target task.

[0032] After that, the first configuration data and the second configuration data are converted into natural language text. Using semantic understanding techniques in natural language processing, in - depth analysis is carried out on the text - based first configuration data and second configuration data to extract key features and semantic information. For example, descriptions regarding flight parameters, sensor performance, communication settings, etc. in the data are identified and their specific meanings are understood.

[0033] Through techniques such as similarity calculation and entity matching in natural language processing, the first configuration data and the second configuration data are compared. For example, compare whether the numerical values of the flight altitude in the two data are the same, whether the sensor types are consistent, etc. For similar concepts in the text description, the degree of difference can also be judged through semantic similarity calculation.

[0034] Based on the comparison and analysis, the differences between the first configuration data and the second configuration data are identified. These differences are the configuration indicators that need to be adjusted or replaced. For example, if the sensor type in the first configuration data is an ordinary camera, while the second configuration data requires a lidar, then the sensor type is a configuration indicator to be replaced.

[0035] The identified differences are further sorted out and summarized to determine specific configuration indicators to be replaced. These indicators can be presented in the form of a list or structured data, clearly indicating which configurations need to be replaced or adjusted, as well as the direction and target values of the replacement or adjustment. For example, the configuration indicators to be replaced can be expressed as: "Sensor type: Replace from ordinary camera with lidar; Flight altitude: Adjust from 100 meters to 150 meters".

[0036] S3: Based on the configuration indicators to be replaced, screen the local configuration data, and according to the screening results, control the drone nest to execute the internal configuration component mobilization signal, so that the drone nest reaches the initial switching state; Specifically, the configuration metrics to be replaced are compared and screened one by one with the local configuration data. For each configuration metric to be replaced, the system looks for matching configuration component information in the local configuration data. For example, if the configuration metric to be replaced requires replacing the drone's sensor from a normal camera to a lidar, the system will check in the local configuration data for available lidar sensors, as well as their quantity and status (such as whether they are available, whether maintenance is required, etc.).

[0037] Based on the screening results, if there are configuration components in the local configuration data that meet the configuration metrics to be replaced, the system will generate corresponding internal configuration component mobilization signals. These signals contain clear operation instructions, such as which storage location in the drone nest to retrieve a specific sensor or mounting device from, and how to install it on the drone. The generated mobilization signals are transmitted through the communication link between the drone nest and the control system to the control unit of the drone nest. The communication link can be a wired network (such as Ethernet) or a wireless network (such as Wi-Fi, Bluetooth, etc.), ensuring that the signals can be accurately and timely transmitted to the drone nest.

[0038] After receiving the mobilization signal, the control unit of the drone nest will control the mechanical devices and electrical systems inside the nest to perform corresponding operations according to the instructions in the signal. For example, it controls the robotic arm to retrieve the required sensor from the storage rack and move it to the installation position of the drone; or adjusts the parameters of the charging device to meet the new power requirements of the drone.

[0039] S4: When it is detected that the position information of the flying drone meets the preset range, obtain the first flight pose data captured by the first sensor provided on the flying drone; Among them, the preset range is a geographical area or location interval preset according to the specific application scenario. The first sensor provided on the flying drone is a sensor for measuring the flight pose of the drone, and commonly includes an inertial measurement unit (IMU), a magnetometer, a barometer, etc. The IMU can measure the acceleration and angular velocity of the drone, and through integral operations, the attitude (such as pitch angle, roll angle, yaw angle) and speed information of the drone can be obtained; the magnetometer can measure the direction of the earth's magnetic field to determine the heading of the drone; the barometer can measure the atmospheric pressure, and calculate the altitude of the drone through the relationship between pressure and height.

[0040] When the system detects that the position information of the flying drone meets the preset range, it will trigger a data acquisition operation. Through the internal communication bus of the drone (such as I2C, SPI, etc.), the system can read the captured first flight pose data from the first sensor. These data are usually stored in the registers of the sensor in binary or text format, and the system needs to read and parse the data according to the communication protocol of the sensor, and convert it into actual physical quantities (such as angles, speeds, heights, etc.).

[0041] S5: Predict the landing position of the flying drone based on the first flight pose data to obtain the predicted landing coordinates; Preferably, in an embodiment of the present application, predicting the landing position of the flying drone based on the first flight pose data to obtain the predicted landing coordinates includes: Obtain the second flight pose data captured by the second sensor provided in the drone nest; Perform weighted fusion on the first flight pose data and the second flight pose data, and calculate to obtain the fused pose data of the flying drone; Input the fused pose data into a pre-trained drone attitude model for prediction to obtain the predicted landing coordinates of the flying drone.

[0042] Specifically, to improve the prediction accuracy, not only the first flight pose data captured by the first sensor on the flying drone is utilized, but also the second flight pose data captured by the second sensor on the drone nest is combined. By performing weighted fusion on these two sets of data and then inputting them into the pre-trained drone attitude model, more reliable predicted landing coordinates can be obtained.

[0043] Preferably, in an embodiment of the present application, obtaining the second flight pose data captured by the second sensor provided in the drone nest includes: Obtain the image data of the flying drone captured by the visual sensors arranged in all directions of the drone nest, perform image enhancement and denoising processing on the image data to obtain the preprocessed image; Analyze the preprocessed image based on image processing technology to obtain the six-degree-of-freedom attitude of the flying drone; Construct an image sequence of the flying drone based on the preprocessed image, and analyze the image sequence based on the optical flow method to obtain the motion trajectory of the flying drone; Generate the second flight pose data according to the six-degree-of-freedom attitude and the motion trajectory.

[0044] Among them, the visual sensors arranged in all directions of the drone nest capture images of the flying drone to obtain its image data. These visual sensors capture images of the drone from different angles and can provide more comprehensive information. Since there may be problems such as uneven illumination and noise interference in the actual shooting environment, in order to improve the image quality, it is necessary to perform image enhancement and denoising processing on the obtained image data. Image enhancement can adopt methods such as histogram equalization and contrast adjustment to highlight the key information in the image; denoising processing can use algorithms such as Gaussian filtering and median filtering to remove the noise points in the image to obtain the preprocessed image.

[0045] Analyze the preprocessed image based on image processing technology. For example, use feature extraction algorithms (such as SIFT, SURF, etc.) to extract the feature points of the drone in the image. By analyzing the positions and relationships of these feature points and combining the calibration parameters of the camera, calculate the six-degree-of-freedom pose of the flying drone, namely three translational degrees of freedom (translation along the X, Y, and Z axes) and three rotational degrees of freedom (rotation around the X, Y, and Z axes).

[0046] Construct an image sequence of the flying drone based on the preprocessed image. These images are arranged in chronological order and record the states of the drone at different times. Analyze the image sequence based on the optical flow method. The optical flow method calculates the motion speed and direction of an object by analyzing the motion of pixel points in adjacent images. By analyzing the optical flow information of the drone in the image sequence, the motion trajectory of the flying drone can be obtained, including its position changes at each moment.

[0047] Integrate the six-degree-of-freedom pose and motion trajectory information obtained above to generate the second flight pose data. This data contains key information such as the pose and motion state of the drone relative to the drone nest, providing an important basis for subsequent data fusion and landing point prediction.

[0048] Construct a distance weight formula for the flying drone to the drone nest. This formula is a monotonically decreasing function of distance. The core idea is that when the drone is close to the nest, the data obtained by the second sensor on the nest is relatively more accurate, so a higher weight is given to the second flight pose data; when the drone is far from the nest, the data obtained by the first sensor on the flying drone is relatively more reliable, so a higher weight is given to the first flight pose data.

[0049] Preferably, in an embodiment of the present application, the first flight pose data and the second flight pose data are weighted and fused to calculate the fused pose data of the flying drone, including: Construct the distance weight formula for the flying drone to the drone nest, expressed as: Wherein, is a monotonically decreasing function of distance , is the rate of weight change, is the distance threshold, is the weight of the first flight pose data, is the weight of the second flight pose data; Fuse the first flight pose data and the second flight pose data according to the distance weight formula to obtain the fused pose data of the flying drone.

[0050] Among them, the rate of weight change controls how fast the weight changes with distance; the distance threshold is a key parameter used to define the range of the influence of distance on the weight.

[0051] According to the above distance weight formula, the first flight pose data and the second flight pose data are weighted and fused. For each pose parameter (such as position, speed, attitude, etc.), the corresponding weight is calculated according to the weight formula respectively, and the pose parameters in the first and second pose data are weighted and summed to obtain the fused pose data of the flying drone. This fusion method fully considers the reliability differences of different sensor data at different distances, and improves the accuracy and reliability of the fused data.

[0052] Preferably, in an embodiment of the present application, before inputting the fused pose data into a pre-trained drone attitude model for prediction, it further includes: Obtain the historical landing data of the flying drone, where the historical landing data includes historical flight pose data and historical landing position data; Perform weighted processing on the historical flight pose data based on the distance of the flying drone from the drone nest to obtain a first sample data set; Construct an initial drone attitude model based on the LSTM network, input the first sample data set into the initial drone attitude model, and construct a loss function according to the obtained output result and the historical landing position data; Optimize the initial drone attitude model based on the loss function to obtain a trained drone attitude model.

[0053] Specifically, before inputting the fused pose data into a pre-trained drone attitude model for prediction, the model needs to be trained and optimized to ensure that it can accurately predict the landing position of the drone according to the flight pose data.

[0054] In different flight stages, the distance of the drone from the nest will affect the reliability of the data. Generally, when the drone is close to the nest, the sensors around the nest can obtain information about the drone more accurately, and the credibility of the data is relatively high; when the drone is far from the nest, it may be affected by more external factors during the flight, and the reliability of the data is relatively low. Therefore, it is necessary to perform weighted processing on the historical flight pose data according to the distance of the drone from the nest.

[0055] Assign a distance-related weight to each historical flight pose data sample. The closer the sample is to the nest, the higher the weight; the farther the sample is from the nest, the lower the weight. In this way, the influence of data with higher credibility on model training is highlighted, and the accuracy of the model is improved. The weighted historical flight pose data forms a first sample data set.

[0056] The Long Short-Term Memory network (LSTM) is a special type of Recurrent Neural Network (RNN) that can effectively process sequential data and solve the problem of vanishing or exploding gradients that traditional RNNs encounter when dealing with long sequences. In the task of UAV attitude prediction, the flight pose data is sequential data arranged in chronological order. The LSTM network can capture the temporal dependencies in the data and is very suitable for handling such problems. According to the task requirements and data characteristics, an initial UAV attitude model based on the LSTM network is constructed. The model usually includes an input layer, an LSTM layer, and an output layer. The input layer receives the historical flight pose data in the first sample dataset; the LSTM layer processes and memorizes the input sequential data through internal memory units and gating mechanisms; the output layer then outputs the predicted landing position data.

[0057] The loss function is used to measure the difference between the model's prediction results and the actual historical landing position data. By minimizing the loss function, the model can continuously adjust its own parameters to improve the prediction accuracy. Input the first sample dataset into the initial UAV attitude model, and the model will output the predicted landing position data. Then, compare these prediction results with the actual historical landing position data, and select an appropriate loss function (such as the mean squared error loss function, cross-entropy loss function, etc.) to calculate the difference between the two. The smaller the value of the loss function, the closer the model's prediction results are to the actual situation.

[0058] Select an appropriate optimization algorithm (such as Stochastic Gradient Descent, Adam optimizer, etc.) to minimize the loss function. The optimization algorithm will continuously adjust the model's parameters according to the gradient information of the loss function, so that the model's prediction results gradually approach the actual historical landing position data. During the training process, input the first sample dataset into the model multiple times for iterative training. In each iteration, the optimization algorithm will update the model's parameters according to the gradient of the loss function until the value of the loss function converges to a small range or reaches the preset number of training epochs. After training and optimization, a trained UAV attitude model is obtained, which can be used to predict the fused pose data to obtain the predicted landing coordinates of the flying UAV.

[0059] S6: Control the UAV nest to execute the target switching state signal generated from the initial switching state and the predicted landing coordinates.

[0060] After obtaining the initial switching state and the predicted landing coordinates of the flying UAV, it is necessary to generate a target switching state signal to control the UAV nest so that the nest is ready to receive the landing of the UAV.

[0061] Preferably, in an embodiment of the present application, controlling the UAV nest to execute the target switching state signal generated from the initial switching state and the predicted landing coordinates includes: Construct an objective function that minimizes the landing deviation of the flying drone based on the predicted landing point coordinates; Solve the objective function based on the dynamic programming solution algorithm, and optimize the solution process through pruning strategies and approximate dynamic programming methods to obtain the optimal decision for the drone nest; Generate a control scheme for the drone nest according to the optimal decision, and control the drone nest to execute the target switching state signal generated by the control scheme.

[0062] Among them, the construction of the objective function aims to minimize the landing deviation of the flying drone. The landing deviation can be understood as the difference between the actual landing position of the drone and the predicted landing point coordinates. By considering factors such as the predicted landing point coordinates, the flight characteristics of the drone, and the position of the nest, a mathematical function is constructed to describe this deviation. For example, the objective function can be defined as the sum of the squares of the Euclidean distances between the drone's landing point and the predicted landing point in the plane coordinates, or a more complex function can be constructed by combining multi-dimensional factors such as height deviation. The role of this objective function is to provide a quantitative evaluation criterion for the subsequent solution of the optimal decision, so that the solution process can proceed in the direction of minimizing the landing deviation.

[0063] Dynamic programming is a method for solving the optimization problem of multi-stage decision-making processes. In this scenario, the decision-making process of the drone nest is regarded as a multi-stage process, with different decision options at each stage, and the goal is to find a series of decisions that minimize the final objective function value. Through the dynamic programming algorithm, the complex global optimization problem can be decomposed into multiple sub-problems, and these sub-problems are solved sequentially to finally obtain the global optimal solution.

[0064] During the solution process, pruning strategies are used to reduce unnecessary calculations. By analyzing the characteristics and constraints of the problem, some decision branches that obviously will not lead to the optimal solution are excluded in advance, thus greatly reducing the search space and improving the solution efficiency. For some complex problems, the exact solution may involve excessive computational effort or take too long. The approximate dynamic programming method further improves the solution speed by approximating the problem while ensuring a certain solution accuracy. For example, methods such as value function approximation and policy approximation can be used to simplify the solution process of the problem.

[0065] After the above solution and optimization process, the optimal decision for the drone nest is obtained. This optimal decision includes the specific operations that the nest should take at each stage, such as adjusting the position of the internal equipment of the nest, preparing the charging facilities, opening the hatch, etc. Generate a detailed control scheme according to the optimal decision, and this scheme clarifies the action instructions and time sequence of each actuator of the nest. Finally, control the drone nest to execute the target switching state signal generated by this control scheme, so that the nest performs state switching according to the predetermined plan to prepare for the landing of the drone.

[0066] Preferably, in one embodiment of the present application, after generating the control scheme of the drone nest according to the optimal decision, it further includes: Calculating the predicted landing time for the flying drone to reach the predicted landing point coordinates according to the first flight pose data; Comparing the predicted landing time with the adjustment time of the obtained control scheme, and controlling the speed of the flying drone according to the comparison result so that the predicted landing time matches the adjustment time.

[0067] According to the first flight pose data, including information such as the current position, speed, and acceleration of the drone, combined with the predicted landing point coordinates, the predicted landing time required for the flying drone to reach the predicted landing point coordinates can be calculated through kinematic formulas or more complex flight models. For example, without considering complex factors such as air resistance, simple calculations can be performed according to the formulas of uniform linear motion or uniformly accelerated linear motion; for more accurate calculations, various factors such as the flight trajectory of the drone, wind direction, and air density need to be considered.

[0068] Obtain the adjustment time of the control scheme, which represents the time required for the drone nest to adjust from the current state to a state capable of welcoming the landing of the drone. Compare the predicted landing time with the adjustment time to determine whether they match. If the predicted landing time is less than the adjustment time, it means that the drone may arrive at the nest too early and the nest is not yet ready; if the predicted landing time is greater than the adjustment time, it means that the nest may complete the adjustment in advance and wait for the drone for too long.

[0069] According to the comparison result, control the speed of the flying drone so that the predicted landing time matches the adjustment time. If the predicted landing time is less than the adjustment time, the flying speed of the drone can be appropriately reduced to extend the time for it to reach the predicted landing point; if the predicted landing time is greater than the adjustment time, the flying speed of the drone can be appropriately increased to shorten the arrival time. In this way, it is ensured that the drone lands accurately after the nest completes the state switch, improving the operating efficiency and reliability of the entire system.

[0070] One embodiment of the present application provides a drone nest control system. Specifically, please refer to Figure 2 , Figure 2 which shows a schematic diagram of the drone nest control system in one of the embodiments of the present application, and it includes: An acquisition module 11, configured to obtain the first configuration data of the flying drone in response to a target task request, and obtain the local configuration library of the drone nest that matches the flying drone; An analysis module 12, configured to determine second configuration data of the flying drone according to a target task request, and perform comparison and analysis on the first configuration data and the second configuration data respectively based on natural language technology to obtain the configuration metrics to be replaced for the flying drone; A mobilization module 13, configured to screen the local configuration data based on the configuration metrics to be replaced, and control the drone nest to execute an internal configuration component mobilization signal according to the screening result, so that the drone nest reaches an initial switching state; A detection module 14, configured to obtain first flight pose data captured by a first sensor disposed on the flying drone when it is detected that the position information of the flying drone meets a preset range; A prediction module 15, configured to predict the landing position of the flying drone based on the first flight pose data to obtain predicted landing coordinates; A control module 16, configured to control the drone nest to execute a target switching state signal generated from the initial switching state and the predicted landing coordinates.

[0071] Compared with the prior art, the beneficial effects of the embodiments of the present application are at least one of the following: 1) By predicting the landing position of the flying drone, the present application generates a target switching state signal in advance, enabling the drone nest to complete corresponding adjustments and preparations when the drone is about to land, further shortening the time interval from the drone landing to being able to execute the next task, and overall improving the response speed of the drone system to tasks.

[0072] 2) From obtaining data, analyzing and comparing, controlling the mobilization of drone nest components to predicting the landing point and generating a target switching state signal, the whole process of the present application requires little human participation. The drone nest can autonomously complete a series of operations, realizing the automation of drone task switching and drone nest docking, reducing the influence of human factors, and improving the stability and reliability of the system.

[0073] The above-described embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the appended claims.

Claims

1. A method for controlling a drone nest, characterized in that, Including: In response to a target task request, obtain the first configuration data of the flying drone and obtain the local configuration data of the drone nest that matches the flying drone; According to the target task request, determine the second configuration data of the flying drone, and perform a comparison and analysis on the first configuration data and the second configuration data based on natural language technology to obtain the configuration metrics to be replaced for the flying drone; Based on the configuration metrics to be replaced, filter in the local configuration data, and control the drone nest to execute an internal configuration component mobilization signal according to the filtering result, so that the drone nest reaches an initial switching state; When it is detected that the position information of the flying drone meets a preset range, obtain the first flight pose data captured by the first sensor provided on the flying drone; Predict the landing position of the flying drone based on the first flight pose data to obtain a predicted landing coordinate; Control the drone nest to execute the target switching state signal generated by the initial switching state and the predicted landing coordinate.

2. The drone nest control method according to claim 1, wherein, After obtaining the first configuration data of the flying drone, it further includes: Obtain the original data of the priority level, remaining time, and completion progress of the current task of the flying drone, and normalize the original data to obtain a current task feature vector; Match the current task feature vector with a pre-constructed decision rule table to calculate the switching matching degree of the current task; If the switching matching degree is greater than a preset task switching threshold, it is determined that task switching is required.

3. The drone nest control method according to claim 2, wherein, The determining the second configuration data of the flying drone according to the target task request includes: Input the target task request into a pre-trained large language model for parsing to obtain a parsing result; Match the parsing result with a pre-constructed knowledge graph, and extract flight data, sensor data, communication data, and mounting data corresponding to the target task request from the knowledge graph according to the matching result; Generate the second configuration data according to the flight data, the sensor data, the communication data, and the mounting data.

4. The drone nest control method according to claim 1, wherein, The predicting the landing position of the flying drone based on the first flight pose data to obtain a predicted landing coordinate includes: Obtain the second flight pose data captured by the second sensor provided on the drone nest; Perform weighted fusion on the first flight pose data and the second flight pose data, and calculate to obtain the fused pose data of the flying drone; Input the fused pose data into a pre-trained drone attitude model for prediction to obtain the predicted landing coordinate of the flying drone.

5. The drone nest control method according to claim 4, wherein, The obtaining the second flight pose data captured by the second sensor provided on the drone nest includes: Obtain the image data of the flying drone captured by the visual sensors provided in each direction of the drone nest, perform image enhancement and denoising processing on the image data to obtain a preprocessed image; Analyze the preprocessed image based on image processing technology to obtain the six-degree-of-freedom attitude of the flying drone. Construct an image sequence of the flying drone based on the preprocessed image, and analyze the image sequence based on the optical flow method to obtain the motion trajectory of the flying drone; Generate the second flying pose data according to the six-degree-of-freedom pose and the motion trajectory.

6. The drone nest control method according to claim 4, wherein, The weighted fusion of the first flying pose data and the second flying pose data to calculate the fused pose data of the flying drone includes: Construct a distance weight formula from the flying drone to the drone nest, expressed as: Among them, is a monotonically decreasing function with respect to the distance , is the rate of change of the weight, is the distance threshold, is the weight of the first flight attitude data, is the weight of the second flight attitude data; Fuse the first flying pose data and the second flying pose data according to the distance weight formula to obtain the fused pose data of the flying drone.

7. The method for controlling a drone nest according to claim 4, wherein Before inputting the fused pose data into a pre-trained drone pose model for prediction, it further includes: Obtain the historical landing data of the flying drone, where the historical landing data includes historical flying pose data and historical landing position data; Perform weighted processing on the historical flying pose data based on the distance between the flying drone and the drone nest to obtain a first sample data set; Construct an initial drone pose model based on the LSTM network, input the first sample data set into the initial drone pose model, and construct a loss function according to the obtained output result and the historical landing position data; Optimize the initial drone pose model based on the loss function to obtain a trained drone pose model.

8. The drone nest control method according to claim 1, characterized in that The control of the drone nest to execute the target switching state signal generated by the initial switching state and the predicted landing point coordinates includes: Construct an objective function that minimizes the landing deviation of the flying drone according to the predicted landing point coordinates; Solve the objective function based on the dynamic programming solution algorithm, and optimize the solution process through pruning strategies and approximate dynamic programming methods to obtain the optimal decision of the drone nest; Generate a control scheme for the drone nest according to the optimal decision, and control the drone nest to execute the target switching state signal generated by the control scheme.

9. The drone nest control method according to claim 7, characterized in that After generating the control scheme for the drone nest according to the optimal decision, it further includes: Calculate the predicted landing time of the flying drone from the first flying pose data to the predicted landing point coordinates; Compare the predicted landing time with the adjustment time of the obtained control scheme, and control the speed of the flying drone according to the comparison result to make the predicted landing time match the adjustment time.

10. A drone nest control system, characterized in that, It includes: An acquisition module for obtaining the first configuration data of the flying drone in response to a target task request and obtaining the local configuration library of the drone nest matching the flying drone; An analysis module for determining the second configuration data of the flying drone according to the target task request, and performing comparison and analysis on the first configuration data and the second configuration data based on natural language technology to obtain the replaceable configuration indicators of the flying drone; A mobilization module, configured to screen the local configuration data based on the to-be-replaced configuration metrics, and control the drone nest to execute an internal configuration component mobilization signal according to the screening result, so that the drone nest reaches an initial switching state; A detection module, configured to obtain first flight attitude data captured by a first sensor provided on the flying drone when the position information of the flying drone is detected to meet a preset range; A prediction module, configured to predict the landing position of the flying drone based on the first flight attitude data to obtain a predicted landing coordinate; A control module, configured to control the drone nest to execute a target switching state signal generated from the initial switching state and the predicted landing coordinate.

Images