Method and platform for carrying out flight management on unmanned aerial vehicle through out-of-step pre-judgment

Abstract

The invention discloses a method and platform for unmanned aerial vehicle flight management through out-of-step pre-judgment, and relates to the technical field of unmanned aerial vehicles, and the method comprises the steps: carrying out the associated feature extraction of an out-of-step data set of an unmanned aerial vehicle, and obtaining an out-of-step feature data set of the unmanned aerial vehicle; constructing an unmanned aerial vehicle out-of-step pre-judgment network model library; the flight state data flow and the operation environment data flow of the unmanned aerial vehicle are acquired; obtaining a target unmanned aerial vehicle out-of-step pre-judgment network model; performing out-of-step pre-judgment identification to obtain out-of-step pre-judgment abnormal parameters; and constructing an unmanned aerial vehicle flight control strategy space, determining unmanned aerial vehicle flight control strategy parameters, and performing out-of-step flight control on the target unmanned aerial vehicle. The technical problems of poor flight stability and safety and low task execution efficiency of the unmanned aerial vehicle due to the fact that flight out-of-step in a complex flight environment and under operation requirements cannot be predicted in advance in existing unmanned aerial vehicle flight management are solved, and the technical effect of improving the flight stability, safety and task execution efficiency of the unmanned aerial vehicle is achieved.

Description

Technical Field

[0001] The present application relates to the technical field related to drones, and specifically to a method and platform for drone flight management based on out-of-step prediction. Background Art

[0002] With the rapid development of drone technology, its application in various fields is becoming more and more extensive, including but not limited to aerial photography, agricultural plant protection, logistics distribution, environmental monitoring, etc. The flight stability and safety of drones have always been the key factors restricting their widespread application. However, traditional drone flights mainly rely on preset flight routes, real-time flight parameter monitoring and fixed flight control strategies, which often fail to achieve ideal control effects. Flight route planning is usually based on map information and obstacle detection. In complex flight environments and changing operational requirements, this planning often ignores the dynamic changes and uncertain factors in the drone flight process, such as wind direction, air pressure changes, mechanical failures, etc., which may cause the drone to lose flight steps, that is, the actual flight status is different from the expected flight status. Deviations in the state may not only affect the successful completion of the mission, but may also cause damage to the drone itself and the surrounding environment. Traditional drone flight control (such as PID control, fuzzy control, etc.) ensures the flight safety of the drone by real-time monitoring of key parameters such as the drone's position, speed, attitude, and auxiliary parameters such as battery power and ambient temperature. It can only detect flight anomalies that have already occurred, but cannot predict and prevent potential loss of step risks in advance. It may show good control effects under specific flight environments and mission requirements, but when faced with complex and changeable flight conditions, its adaptability and robustness are often insufficient, and it is difficult to accurately predict the loss of step in the drone flight, thereby affecting the stability and safety of the drone flight and the efficiency and quality of mission execution.

[0003] Therefore, at the current stage, there is a technical problem in the UAV flight management related technologies that cannot predict in advance the flight desynchronization under complex flight environments and operational requirements, which leads to poor UAV flight stability and safety and low mission execution efficiency. Summary of the invention

[0004] This application solves the technical problem that the existing UAV flight management cannot predict flight desynchronization in complex flight environments and operational requirements in advance, which leads to poor UAV flight stability and safety and low mission execution efficiency, by providing a method and platform for UAV flight management based on desynchronization prediction. This achieves the technical effect of improving UAV flight stability, safety and mission execution efficiency.

[0005] The present application provides a method for unmanned aerial vehicle flight management by predicting out-of-step, the method comprising: mining and obtaining a data set of unmanned aerial vehicle out-of-step, extracting associated features from the unmanned aerial vehicle out-of-step data set, and obtaining a data set of unmanned aerial vehicle out-of-step feature; obtaining unmanned aerial vehicle application attribute information, sampling and predicting training identification of the unmanned aerial vehicle out-of-step feature data set based on the unmanned aerial vehicle application attribute information, and constructing a network model library for predicting out-of-step; carrying a sensor group on a target unmanned aerial vehicle, the sensor group integrating a position sensor, a speed sensor, an attitude sensor, an air pressure sensor, and a temperature and humidity sensor, and collecting and obtaining a data stream of the flight status of the unmanned aerial vehicle through the sensor group and operating environment data stream; based on the application attribute information of the target UAV, the characteristics are matched with the UAV out-of-step prediction network model library to obtain the target UAV out-of-step prediction network model; the out-of-step prediction identification is performed on the UAV flight status data stream and the operating environment data stream through the target UAV out-of-step prediction network model to obtain the out-of-step prediction abnormal parameters; according to the out-of-step prediction abnormal parameters, the UAV flight control strategy space is constructed, the out-of-step balance target is defined to control and optimize the UAV flight control strategy space, determine the UAV flight control strategy parameters, and the target UAV is controlled and managed through the UAV flight control strategy parameters.

[0006] In a possible implementation, the construction of the UAV out-of-step prediction network model library also performs the following processing: extracting elements from the UAV application attribute information to obtain UAV application attribute elements, wherein the UAV application attribute elements include UAV specification elements and application scenario elements; extracting and integrating attribute parameters of the UAV specification elements and application scenario elements based on the UAV out-of-step data set to obtain a UAV application attribute parameter set; performing data mapping sampling on the UAV out-of-step feature data set through the UAV application attribute parameter set to obtain a multi-application attribute parameter out-of-step feature data set; performing out-of-step prediction training on the multi-application attribute parameter out-of-step prediction feature data set respectively to obtain a multi-application attribute parameter out-of-step prediction network model set; using the UAV application attribute parameter set to perform attribute parameter calibration on the multi-application attribute parameter out-of-step prediction network model set to construct the UAV out-of-step prediction network model library.

[0007] In a possible implementation, the method of obtaining a multi-application attribute parameter out-of-step feature data set further performs the following processing: performing feature association matching between the drone application attribute parameter set and the drone out-of-step feature data set respectively, to obtain a multi-attribute parameter primary-order association feature data set; using the multi-attribute parameter primary-order association feature data set as a sampling center, performing neighboring data collection in the drone out-of-step feature data set as a multi-attribute parameter second-order association feature data set; traversing the multi-attribute parameter second-order association feature data set, performing K-order data collection in the drone out-of-step feature data set, to obtain a K-order multi-attribute parameter association feature data set, wherein 2≤K≤8; sampling and mapping integration are performed on the multi-attribute parameter primary-order association feature data set, the multi-attribute parameter second-order association feature data set, and until the K-order multi-attribute parameter association feature data set, to obtain the multi-application attribute parameter out-of-step feature data set.

[0008] In a possible implementation, the construction of the UAV flight control strategy space also performs the following processing: obtaining flight control strategy dimensional information, performing flight control strategy analysis on the out-of-step prediction abnormal parameters according to the flight control strategy dimensional information, and obtaining a flight control strategy dimensional parameter threshold; randomly selecting multiple control strategy dimensional parameters within the flight control strategy dimensional parameter threshold, evaluating and optimizing the multiple control strategy dimensional parameters, and determining an initial control strategy parameter set; cross-combining and mutating the initial control strategy parameter set to obtain a child control strategy parameter set, and performing optimal parameter screening on the child control strategy parameter set to obtain an optimal child strategy parameter set; based on the initial control strategy parameter set and the optimal child strategy parameter set, the UAV flight control strategy space is constructed in combination.

[0009] In a possible implementation, the determination of the UAV flight control strategy parameters also performs the following processing: M parameter intervals are divided in the UAV flight control strategy space, and M interval parameter sets are selected in the M parameter intervals respectively; the out-of-step balance target is defined to perform control effect simulation evaluation on the M interval parameter sets in turn to obtain M interval parameter out-of-step control simulation effects; based on the M interval parameter out-of-step control simulation effects, the M parameter intervals are optimized to determine the flight control strategy search interval; a global parameter comparison and optimization is performed in the flight control strategy search interval to determine the UAV flight control strategy parameters.

[0010] In a possible implementation, the determination of the flight control strategy search interval also performs the following processing: based on the M interval parameter out-of-step control simulation effects, parameter preset proportion screening is performed respectively to obtain M interval control strategy parameters; the effect mean calculation result of the M interval control strategy parameters is used as the M parameter interval control effect; the M parameter intervals are optimized according to the M parameter interval control effects to determine the first strategy parameter interval; the first strategy parameter interval is iteratively divided and optimized until the preset number of interval divisions is reached to determine the flight control strategy search interval.

[0011] In a possible implementation, a step-out balance target is defined to control and optimize the UAV flight control strategy space, determine the UAV flight control strategy parameters, and perform the following processing: use the UAV flight control strategy parameters to perform flight control monitoring on the target UAV to obtain the UAV flight state feedback parameters; perform step-out optimization analysis based on the UAV flight state feedback parameters to determine the strategy parameter optimization variation rules; mutate and update the UAV flight control strategy parameters according to the strategy parameter optimization variation rules to obtain multiple flight control strategy variation parameters, and determine the UAV flight control strategy optimization parameters based on the comparison and optimization of the multiple flight control strategy variation parameters.

[0012] The present application also provides a platform for UAV flight management based on out-of-step prediction, including: an associated feature extraction module, used to mine and obtain a UAV out-of-step data set, extract associated features from the UAV out-of-step data set, and obtain a UAV out-of-step feature data set; a prediction network model library construction module, used to obtain UAV application attribute information, perform data sampling and prediction training identification on the UAV out-of-step feature data set based on the UAV application attribute information, and construct a UAV out-of-step prediction network model library; a UAV data stream acquisition module, used to carry a sensor group on the target UAV, the sensor group integrates a position sensor, a speed sensor, an attitude sensor, an air pressure sensor, and a temperature and humidity sensor, and collects and obtains the UAV flight status data through the sensor group a characteristic matching module, used for matching the characteristics of the target UAV with the UAV out-of-step prediction network model library based on the application attribute information of the target UAV, and obtaining the target UAV out-of-step prediction network model; a out-of-step prediction identification module, used for performing out-of-step prediction identification on the UAV flight status data flow and the operating environment data flow through the target UAV out-of-step prediction network model, and obtaining out-of-step prediction abnormal parameters; a flight control strategy parameter determination module, used for constructing a UAV flight control strategy space according to the out-of-step prediction abnormal parameters, defining an out-of-step balance target to control and optimize the UAV flight control strategy space, determining the UAV flight control strategy parameters, and performing out-of-step flight control and management of the target UAV through the UAV flight control strategy parameters.

[0013] The method and platform for UAV flight management proposed in this application is to extract related features from the UAV out-of-step data set to obtain a UAV out-of-step feature data set; build a UAV out-of-step prediction network model library; collect and obtain the UAV flight status data stream and operating environment data stream; obtain the target UAV out-of-step prediction network model; perform out-of-step prediction identification to obtain out-of-step prediction abnormal parameters; build a UAV flight control strategy space, determine the UAV flight control strategy parameters, and control the target UAV out-of-step flight. It solves the technical problem that the existing UAV flight management cannot predict flight out-of-steps in complex flight environments and operational requirements in advance, which leads to poor UAV flight stability, poor safety, and low mission execution efficiency, and achieves the technical effect of improving UAV flight stability, safety, and mission execution efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure are briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the platform according to the embodiment of the present application. It should be understood that the previous or following operations are not necessarily performed precisely in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0015] Figure 1 A flow chart of a method for predicting out-of-step and performing flight management on a UAV provided in an embodiment of the present application;

[0016] Figure 2 Schematic diagram of the platform structure for unmanned aerial vehicle flight management based on out-of-step prediction provided in an embodiment of the present application.

[0017] Explanation of the reference numerals: associated feature extraction module 10, prediction network model library construction module 20, drone data stream acquisition module 30, characteristic matching module 40, out-of-step prediction and identification module 50, flight control strategy parameter determination module 60. DETAILED DESCRIPTION

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0019] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0020] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, platform, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0021] The present application embodiment provides a method for predicting out-of-step and managing the flight of a drone, such as Figure 1 As shown, the method includes:

[0022] Step S100, mining and acquiring a UAV out-of-step data set, extracting associated features from the UAV out-of-step data set, and obtaining a UAV out-of-step feature data set.

[0023] Preferably, the desynchronization of a drone usually refers to the situation that the flight state of the drone is inconsistent with the expectation due to some reasons (such as signal interference, mechanical failure, control algorithm problem, etc.) during the flight, and there are phenomena such as deviation from the predetermined trajectory, abnormal speed change, unstable posture, etc. A drone desynchronization dataset under different environmental conditions is obtained, which includes a dataset of desynchronization (i.e., asynchronous or abnormal state) of the drone during flight, for example, it may include the flight parameters, sensor data, image or video data of the drone, etc., and then the associated features of the acquired drone desynchronization dataset are extracted, that is, by analyzing the correlation between the features, the associated features are extracted from the drone desynchronization dataset. Specifically, in the drone desynchronization dataset, The features that can be extracted include the drone's location information (such as longitude, latitude, and altitude), speed information (such as horizontal speed and vertical speed), attitude information (such as pitch angle, yaw angle, and roll angle), and other features related to the drone's loss of synchronization, such as signal strength and response time of control commands. Then, by analyzing the correlation between the features and using association rule mining algorithms (such as the Apriori algorithm and the FP-Growth algorithm), the feature set associated with the drone's loss of synchronization is extracted, and the feature set most related to the drone's loss of synchronization is found. As the drone's loss of synchronization feature data set, it contains all feature information associated with the drone's loss of synchronization, providing a data basis for subsequent analysis, prediction, and model training tasks.

[0024] Step S200, obtaining drone application attribute information, performing data sampling and prediction training identification on the drone out-of-step feature data set based on the drone application attribute information, and building a drone out-of-step prediction network model library.

[0025] Preferably, the drone application attribute information is obtained, which refers to the characteristics and capabilities of the drone in various application scenarios, including but not limited to flight speed, high-altitude vision, easy operation, low cost and high safety, and then the drone out-of-step feature data set is sampled and pre-judgment training marked according to the drone application attribute information. Specifically, representative samples are screened out from a large number of drone out-of-step feature data sets to fully reflect the flight status, fault characteristics and other key information of the drone. For example, random sampling, stratified sampling and other methods are used to ensure the diversity and representativeness of the samples, and then the sampled data are pre-processed and feature extracted, and then each sample is marked with a corresponding pre-judgment training mark. These pre-judgment training marks usually indicate whether the drone is in a state of out-of-step and the severity of the out-of-step. The severity or type of the drone out-of-step error can be determined by predicting the training flag, which can provide accurate supervision information for subsequent model training. Then, based on the results of data sampling and predicting the training flag, a drone out-of-step prediction network model library is constructed. Specifically, according to the characteristics of the drone out-of-step feature and the size of the data set, a suitable machine learning or deep learning algorithm is constructed for model training. For example, support vector machine, random forest, neural network and other algorithms can be used for model construction. The preprocessed drone out-of-step data and predicting the training flag are input into the constructed model for training to obtain multiple drone out-of-step prediction network models, which are finally integrated into a model library. These models can be trained and optimized for different drone types, application scenarios or fault types. By constructing a model library, rapid prediction and identification of drone out-of-step status can be achieved.

[0026] Step S200 further includes step S210, extracting elements from the drone application attribute information to obtain drone application attribute elements, wherein the drone application attribute elements include drone specification elements and application scenario elements; step S220, extracting and integrating attribute parameters of the drone specification elements and application scenario elements based on the drone out-of-step data set to obtain a drone application attribute parameter set; step S230, performing data mapping sampling on the drone out-of-step feature data set through the drone application attribute parameter set to obtain a multi-application attribute parameter out-of-step feature data set; step S240, performing out-of-step prediction training on the multi-application attribute parameter out-of-step feature data set respectively to obtain a multi-application attribute parameter out-of-step prediction network model set; step S250, using the drone application attribute parameter set to perform attribute parameter calibration on the multi-application attribute parameter out-of-step prediction network model set to construct the drone out-of-step prediction network model library.

[0027] Preferably, key elements, namely, drone application attribute elements, are extracted from the drone application attribute information, including drone application attribute elements including drone specification elements and application scenario elements, wherein the drone specification elements include the model, size, weight, flight speed, flight altitude, load capacity and other physical and technical parameters of the drone, and the application scenario elements mainly describe the specific environment or task in which the drone is used, such as agricultural monitoring, topographic mapping, aerial photography, express delivery, etc. Then, based on the drone out-of-step data set, the attribute parameters of the drone specification elements and the application scenario elements are integrated to form a drone application attribute parameter set, that is, parameters related to the drone specifications and application scenarios are extracted from the drone out-of-step data set and integrated. Specifically, data sampling is performed on drone application attributes of different drone specification types and application scenario types, so as to perform model training and parameter identification of the corresponding attribute parameters respectively to form a drone application attribute parameter set.

[0028] Preferably, the drone out-of-step feature data set is sampled by data mapping through the drone application attribute parameter set. Specifically, the drone application attribute parameter set is associated with the drone out-of-step feature data set. According to different combinations of application attribute parameters, the data is sampled to obtain out-of-step feature data under different application attribute parameters, thereby forming a multi-application attribute parameter out-of-step feature data set. Then, the multi-application attribute parameter out-of-step feature data set is trained for out-of-step prediction respectively, that is, a machine learning or deep learning algorithm is used to train the multi-application attribute parameter out-of-step feature data set, and multiple out-of-step prediction network models are constructed, each model corresponding to a set of specific application attribute parameters, so as to obtain an out-of-step prediction network model suitable for different application attribute parameters. Finally, the drone application attribute parameter set is used to calibrate the attribute parameters of the multi-application attribute parameter out-of-step prediction network model set. Specifically, each out-of-step prediction network model is calibrated so that it can accurately identify and predict the drone out-of-step situation under the corresponding application attribute parameters. The calibrated out-of-step prediction network model set is integrated into a drone out-of-step prediction network model library, so that in actual applications, a suitable model can be selected according to the specifications and application scenarios of the drone for out-of-step prediction.

[0029] Step S230 further includes step S231, performing feature association matching between the drone application attribute parameter set and the drone out-of-step feature data set respectively, to obtain a multi-attribute parameter primary-order association feature data set; step S232, taking the multi-attribute parameter primary-order association feature data set as the sampling center, performing neighboring data collection in the drone out-of-step feature data set as a multi-attribute parameter second-order association feature data set; step S233, traversing the multi-attribute parameter second-order association feature data set, performing K-order data collection in the drone out-of-step feature data set, to obtain a K-order multi-attribute parameter association feature data set, wherein 2≤K≤8; step S234, sampling and mapping integration of the multi-attribute parameter primary-order association feature data set, the multi-attribute parameter second-order association feature data set, and until the K-order multi-attribute parameter association feature data set, to obtain the multi-application attribute parameter out-of-step feature data set.

[0030] Preferably, the application attribute parameter set of the drone (flight speed, altitude, battery power, wind direction, etc. of the drone) and the out-of-step feature data set of the drone (instability during flight, deviation from the scheduled route, sensor data abnormality, etc.) are used for feature association matching to obtain a primary-order association feature data set, which reflects the preliminary relationship between the application attributes of the drone and the out-of-step features. The primary-order association feature data set is then used as the sampling center to collect neighboring data in the drone out-of-step feature data set, that is, to find data points close to the primary-order association features in the data space, thereby forming a richer multi-attribute parameter second-order association feature data set containing more relevant information; then the second-order association feature data set is traversed, and K-order data is collected in the drone out-of-step feature data set, where K represents the order of collection and is an integer between 2 and 8. Finally, the primary-order association feature data set, the second-order association feature data set, and all K-order association feature data sets are sampled and mapped and integrated, including data alignment, normalization, dimensionality reduction and other processing, to form a unified and complete multi-application attribute parameter out-of-step feature data set.

[0031] Step S300: A sensor group is mounted on the target UAV, wherein the sensor group includes a position sensor, a speed sensor, an attitude sensor, an air pressure sensor, and a temperature and humidity sensor, and the flight status data stream and the operating environment data stream of the UAV are collected and acquired through the sensor group.

[0032] Preferably, a sensor group is mounted on the target UAV, and these sensors are used to collect the flight status data stream and the operating environment data stream of the UAV. Specifically, the sensor group includes a position sensor, a speed sensor, an attitude sensor, an air pressure sensor, a temperature and humidity sensor, etc. The position sensor (such as a GPS unit) is used to obtain the geographic location information of the UAV in real time, the speed sensor (such as an airspeed meter, a ground speed meter, etc.) is used to measure the flight speed of the UAV, the attitude sensor (such as an accelerometer, a gyroscope, a magnetometer, an electronic compass, etc.) is used to monitor the flight attitude of the UAV, including the pitch angle, the yaw angle and the roll angle, and the air pressure sensor (such as a barometer) is used to measure the air pressure value of the environment in which the UAV is located, and then calculate the height information, the temperature and humidity The sensor is used to monitor the temperature and humidity information of the drone's environment; when the drone is in flight, each sensor will collect the drone's flight status data and operating environment data in real time. The flight status data includes position, speed, attitude, etc., reflecting the real-time flight status of the drone. The operating environment data includes air pressure, temperature, humidity, etc., reflecting the real-time status of the drone's flight environment. The collected data will be transmitted through the drone's internal communication module. Usually, this data will be sent to the drone's central processor or ground control station. The ground control station or central processor processes and analyzes the received data, extracts useful information for the drone's flight control and mission planning, and realizes real-time monitoring and early warning of the drone's flight status.

[0033] Step S400: performing feature matching with the UAV out-of-step prediction network model library based on the application attribute information of the target UAV to obtain a target UAV out-of-step prediction network model.

[0034] Preferably, key characteristics and parameters are extracted from the application attribute information of the target UAV for feature matching with the models in the model library. Specifically, based on the extracted characteristics and parameters, the model that best matches the target UAV is screened out in the model library, that is, multiple prediction network models in the UAV out-of-step prediction network model library are compared and evaluated one by one to determine which model is most suitable for the application scenario of the target UAV, and finally the out-of-step prediction network model with the highest matching degree with the target UAV is obtained. As the out-of-step prediction network model of the target UAV, this model will be able to accurately predict the possible out-of-step situations that may occur in the UAV during flight, and provide strong support for the flight control and mission planning of the UAV.

[0035] Step S500: perform out-of-step prediction and identification on the flight status data stream and the operating environment data stream of the target UAV through the out-of-step prediction network model to obtain out-of-step prediction abnormal parameters.

[0036] Preferably, before the UAV flight status data stream and the operating environment data stream are input into the target UAV out-of-step prediction network model, they are preprocessed, such as data cleaning, denoising, format conversion, etc., to ensure the accuracy and consistency of the data, and then the preprocessed data stream is input into the target UAV out-of-step prediction network model to perform out-of-step prediction identification, that is, analyze and identify anomalies or deviations in the data, which may indicate that the UAV is about to lose step. Specifically, abnormal parameters that are inconsistent with the normal flight state are identified, which may include position offset, speed fluctuation, attitude loss of control, etc., as out-of-step prediction abnormal parameters, and the corresponding abnormal parameter types are identified, such as position abnormality, speed abnormality, attitude abnormality, etc., reflecting the possible out-of-step situation of the UAV.

[0037] Step S600, constructing a UAV flight control strategy space according to the out-of-step prediction abnormal parameters, defining an out-of-step balance target to control and optimize the UAV flight control strategy space, determining the UAV flight control strategy parameters, and performing out-of-step flight control on the target UAV through the UAV flight control strategy parameters.

[0038] Preferably, a UAV flight control strategy space is constructed according to the out-of-step prediction abnormal parameters, and the control strategy space is controlled and optimized by defining the out-of-step balance target to determine the UAV flight control strategy parameters. Specifically, the possible out-of-step situations of the UAV and the severity of the out-of-step situation are determined according to the abnormal parameters output by the out-of-step prediction network model. According to the out-of-step situation, a corresponding flight control strategy space is constructed to deal with the possible out-of-step problem. The UAV flight control strategy space is a set of multiple possible flight control strategies, covering multiple aspects such as attitude control, speed control, and position control of the target UAV. The out-of-step balance target is defined, that is, during the flight of the UAV, by adjusting the flight control strategy, the UAV can quickly restore a stable flight state to avoid the occurrence of out-of-step situations or reduce their impact. The out-of-step balance target is refined into specific control indicators, such as attitude stability, speed fluctuation range, position offset, etc., which are used to evaluate the effectiveness of different flight control strategies. The effect is obtained and used as the basis for control optimization. Then, based on the optimization algorithm (such as genetic algorithm, particle swarm optimization, etc.), control optimization is performed in the flight control strategy space. Specifically, according to the out-of-step equilibrium target and control index, the strategy parameters in the flight control strategy space are adjusted and optimized. Through iterative calculation, the optimal solution is gradually approached to obtain the best flight control strategy parameters, that is, the UAV flight control strategy parameters are determined. Finally, the optimized flight control strategy parameters are applied to the target UAV to realize the flight control of the UAV. During the flight, the flight status data stream and the operating environment data stream are monitored in real time. When the out-of-step prediction abnormal parameters are detected, the corresponding flight control strategy is immediately started to control the out-of-step flight of the UAV, such as adjusting the attitude, speed, position and other parameters of the UAV, so that the UAV can quickly restore the stable flight state, avoid the occurrence of out-of-step or reduce its impact, thereby realizing precise control and optimization of the UAV flight state and improving the flight safety and stability of the UAV.

[0039] Step S600 further includes step S610, obtaining flight control strategy dimensional information, performing flight control strategy analysis on the out-of-step prediction abnormal parameters according to the flight control strategy dimensional information, and obtaining a flight control strategy dimensional parameter threshold; step S620, randomly selecting multiple control strategy dimensional parameters within the flight control strategy dimensional parameter threshold, evaluating and optimizing the multiple control strategy dimensional parameters, and determining an initial control strategy parameter set; step S630, cross-combining and mutating the initial control strategy parameter set to obtain a child control strategy parameter set, and performing optimal parameter screening on the child control strategy parameter set to obtain an optimal child strategy parameter set; step S640, based on the initial control strategy parameter set and the optimal child strategy parameter set, combining and constructing the UAV flight control strategy space.

[0040] Preferably, obtaining the flight control strategy dimensional information refers to collecting various dimensional parameters that can be controlled during the flight of the UAV, such as flight attitude (pitch, yaw, roll, etc.), flight trajectory (route, altitude, speed, etc.), speed (airspeed, ground speed, etc.), flight mode (manual, automatic, fixed altitude, fixed point, etc.), etc. According to the collected flight control strategy dimensional information, the out-of-step prediction abnormal parameters (i.e., parameters that may cause flight instability or loss of control) are analyzed, and then the reasonable range or threshold of each control dimensional parameter is determined to ensure that the UAV can fly stably under the control strategy, and then multiple control strategy dimensional parameter values ​​are randomly selected within each flight control strategy dimensional parameter threshold to form multiple different control strategy dimensional combinations, and multiple control strategy dimensional parameters are evaluated and optimized to select the ones with better performance. Several groups are used as the initial control strategies, that is, the initial control strategy parameter set is determined; the initial control strategy parameter set is then cross-combined and mutated to generate more control strategy combinations, that is, the offspring control strategy parameter set. These combinations are usually called offspring in genetic algorithms. The offspring control strategy parameter set is evaluated to screen out parameter combinations with better performance, that is, through iterative optimization, the optimal control strategy is gradually approached to form a superior offspring strategy parameter set. Finally, based on the initial control strategy parameter set and the superior offspring strategy parameter set, the UAV flight control strategy space is combined to construct a variety of possible control strategy combinations. The UAV can select the most appropriate control strategy from it according to the actual situation (such as flight environment, mission requirements, etc.) to ensure that the UAV can fly stably and efficiently under various conditions.

[0041] Step S600 further includes step S650, dividing M parameter intervals in the UAV flight control strategy space, and selecting M interval parameter sets in the M parameter intervals respectively; step S660, defining the out-of-step balance target to perform control effect simulation evaluation on the M interval parameter sets in turn, and obtaining the out-of-step control simulation effects of M interval parameters; step S670, performing interval optimization on the M parameter intervals based on the out-of-step control simulation effects of the M interval parameters, and determining the flight control strategy search interval; step S680, performing global parameter comparison and optimization in the flight control strategy search interval to determine the UAV flight control strategy parameters.

[0042] Preferably, in the UAV flight control strategy space, according to the value range of each control parameter, it is divided into M parameter intervals, in each divided parameter interval, multiple representative parameter values ​​are selected to form M interval parameter sets, representing the possible control strategies in each interval, and then the out-of-step equilibrium target is defined, that is, the stable flight state or performance index expected to be achieved through the control strategy. Specifically, the simulation model is used to perform simulation evaluation of the flight control effect on the M interval parameter sets in turn, and the simulation effect of the out-of-step control of the M interval parameters is obtained, such as measuring the effect through performance indicators (such as stability, accuracy, energy consumption, etc.), and then based on the M intervals, the simulation model is used to evaluate the flight control effect of the M interval parameter sets in turn. The simulation effect of parameter out-of-step control is obtained by optimizing the M parameter intervals to find out those intervals with better control effects, that is, to determine the flight control strategy search interval as the candidate area for subsequent search of flight control strategy parameters. Finally, in the flight control strategy search interval determined by optimization, a global comparison and optimization of parameters is performed. For example, a more sophisticated search algorithm (such as grid search, random search, genetic algorithm, etc.) is used to find the optimal or approximately optimal control strategy parameter combination in the search interval. Finally, the flight control strategy parameters of the UAV are determined for the actual flight control of the UAV to achieve flight performance and stability under various flight conditions.

[0043] Step S670 further includes step S671, performing parameter preset proportion screening based on the M interval parameter out-of-step control simulation effects to obtain M interval control strategy parameters; step S672, calculating the mean effect result of the M interval control strategy parameters as the M parameter interval control effect; step S673, performing interval optimization on the M parameter intervals according to the M parameter interval control effects to determine the first strategy parameter interval; step S674, iteratively dividing and optimizing the first strategy parameter interval until the preset number of interval divisions is reached to determine the flight control strategy search interval.

[0044] Preferably, based on the simulation effects of out-of-step control of M interval parameters, the control effects in each interval are sorted, and the parameters with better performance are screened out according to a preset ratio (such as the first 50%) to obtain M interval control strategy parameters, and then the control effect mean value of the M interval control strategy parameters is calculated to obtain representative control strategy parameters and their effect mean values ​​for each interval as the corresponding M parameter interval control effects, and then the M parameter intervals are sorted according to the control effects, and the interval with the best control effect is selected as the first strategy parameter interval, and then the first strategy parameter interval is iteratively divided and optimized, that is, the interval is further subdivided into smaller sub-intervals, and the above-mentioned parameter preset ratio screening, mean value calculation and other steps are repeated in each sub-interval to further narrow the selection range of the strategy parameters. Through multiple iterations, the optimal control strategy parameter combination is gradually approached, and a more accurate control strategy search interval is formed. When the number of iterations reaches the preset number of interval divisions, the iteration is stopped. The interval obtained at this time is the flight control strategy search interval, and the parameter combination in this interval has a higher control effect.

[0045] Step S600 further includes step S690, using the UAV flight control strategy parameters to perform flight control monitoring on the target UAV to obtain the UAV flight status feedback parameters; step S6100, performing out-of-step optimization analysis based on the UAV flight status feedback parameters to determine the strategy parameter optimization variation rules; step S6110, mutating and updating the UAV flight control strategy parameters according to the strategy parameter optimization variation rules to obtain multiple flight control strategy variation parameters, and determining the UAV flight control strategy optimization parameters based on the comparison and optimization of the multiple flight control strategy variation parameters.

[0046] Preferably, after the UAV flight control strategy parameters are determined, the UAV flight control strategy parameters are further optimized through actual flight monitoring, feedback parameter analysis, strategy parameter variation and optimization comparison. Specifically, the UAV flight control strategy parameters are used to perform actual flight control monitoring of the target UAV. During the flight, the UAV flight state feedback parameters are obtained in real time, including flight speed, altitude, attitude, position, etc., and then based on the obtained UAV flight state feedback parameters, out-of-step optimization analysis is performed, that is, the stability, accuracy, response speed and other aspects of the flight state are evaluated, and compared with the expected flight state. According to the analysis results, the strategy parameter optimization variation rules are determined. The strategy parameter optimization variation rules It includes which parameters to adjust, the direction of adjustment (increase or decrease), the magnitude of adjustment, etc., and the UAV flight control strategy parameters are mutated and updated according to the determined strategy parameter optimization mutation rules. That is, on the basis of the original parameters, a certain degree of adjustment is made according to the rules to generate multiple flight control strategy variation parameters, and the generated multiple flight control strategy variation parameters are simulated and evaluated or actually tested in flight to obtain their corresponding flight state feedback parameters. Finally, based on these flight state feedback parameters, a comparison is performed to find the best, and the optimal or approximately optimal UAV flight control strategy optimization parameters are determined for use in subsequent UAV flight control, so that the UAV can achieve better flight performance and stability under various flight conditions.

[0047] In the above, refer to Figure 1 The method for managing the flight of a UAV by predicting out-of-step according to an embodiment of the present invention is described in detail. Figure 2 A platform for unmanned aerial vehicle flight management based on out-of-sync prediction according to an embodiment of the present invention is described.

[0048] The platform for UAV flight management by predicting out-of-step according to the embodiment of the present invention is used to solve the technical problem that the existing UAV flight management cannot predict the flight out-of-step under complex flight environment and operation requirements in advance, which leads to poor flight stability and safety of the UAV and low efficiency of task execution, and achieves the technical effect of improving the flight stability, safety and task execution efficiency of the UAV. The platform for UAV flight management by predicting out-of-step includes: a correlation feature extraction module 10, a prediction network model library construction module 20, a UAV data stream acquisition module 30, a feature matching module 40, a prediction out-of-step identification module 50, and a flight control strategy parameter determination module 60.

[0049] The associated feature extraction module 10 is used to mine and obtain a UAV out-of-step data set, extract associated features from the UAV out-of-step data set, and obtain a UAV out-of-step feature data set; the prediction network model library construction module 20 is used to obtain UAV application attribute information, perform data sampling and prediction training identification on the UAV out-of-step feature data set based on the UAV application attribute information, and construct a UAV out-of-step prediction network model library; the UAV data stream acquisition module 30 is used to carry a sensor group on the target UAV, and the sensor group includes a position sensor, a speed sensor, a posture sensor, an air pressure sensor, and a temperature and humidity sensor, and the UAV flight status data stream and the operating environment data stream are collected and acquired through the sensor group; characteristic matching Module 40 is used to perform feature matching with the UAV out-of-step prediction network model library based on the application attribute information of the target UAV to obtain the out-of-step prediction network model of the target UAV; the out-of-step prediction identification module 50 is used to perform out-of-step prediction identification on the flight status data stream and the operating environment data stream of the UAV through the out-of-step prediction network model of the target UAV to obtain out-of-step prediction abnormal parameters; the flight control strategy parameter determination module 60 is used to construct the UAV flight control strategy space according to the out-of-step prediction abnormal parameters, define the out-of-step balance target to control and optimize the UAV flight control strategy space, determine the UAV flight control strategy parameters, and perform out-of-step flight control and management of the target UAV through the UAV flight control strategy parameters.

[0050] The specific configuration of the prediction network model library construction module 20 will be described in detail below. The prediction network model library construction module 20 may further include: extracting elements from the drone application attribute information to obtain drone application attribute elements, wherein the drone application attribute elements include drone specification elements and application scenario elements; extracting and integrating attribute parameters of the drone specification elements and application scenario elements based on the drone out-of-step data set to obtain a drone application attribute parameter set; performing data mapping sampling on the drone out-of-step feature data set through the drone application attribute parameter set to obtain a multi-application attribute parameter out-of-step feature data set; performing out-of-step prediction training on the multi-application attribute parameter out-of-step feature data set respectively to obtain a multi-application attribute parameter out-of-step prediction network model set; using the drone application attribute parameter set to perform attribute parameter calibration on the multi-application attribute parameter out-of-step prediction network model set to construct the drone out-of-step prediction network model library.

[0051] The specific configuration of the prediction network model library construction module 20 will be described in detail below. The prediction network model library construction module 20 may further include: performing feature association matching with the drone application attribute parameter set and the drone out-of-step feature data set to obtain a multi-attribute parameter primary-order association feature data set; using the multi-attribute parameter primary-order association feature data set as a sampling center, performing neighboring data collection in the drone out-of-step feature data set as a multi-attribute parameter second-order association feature data set; traversing the multi-attribute parameter second-order association feature data set, performing K-order data collection in the drone out-of-step feature data set, and obtaining a K-order multi-attribute parameter association feature data set, where 2≤K≤8; sampling and mapping integration of the multi-attribute parameter primary-order association feature data set, the multi-attribute parameter second-order association feature data set, and the K-order multi-attribute parameter association feature data set to obtain the multi-application attribute parameter out-of-step feature data set.

[0052] The specific configuration of the flight control strategy parameter determination module 60 will be described in detail below. The flight control strategy parameter determination module 60 may further include: obtaining flight control strategy dimension information, performing flight control strategy analysis on the out-of-step prediction abnormal parameters according to the flight control strategy dimension information, and obtaining a flight control strategy dimension parameter threshold; randomly selecting multiple control strategy dimension parameters within the flight control strategy dimension parameter threshold, evaluating and optimizing the multiple control strategy dimension parameters, and determining the initial control strategy parameter set; performing cross-combination and mutation expansion on the initial control strategy parameter set to obtain a child control strategy parameter set, and performing optimal parameter screening on the child control strategy parameter set to obtain an optimal child strategy parameter set; based on the initial control strategy parameter set and the optimal child strategy parameter set, the UAV flight control strategy space is constructed in combination.

[0053] The specific configuration of the flight control strategy parameter determination module 60 will be described in detail below. The flight control strategy parameter determination module 60 may further include: dividing M parameter intervals in the UAV flight control strategy space, and selecting M interval parameter sets in the M parameter intervals respectively; defining the out-of-step balance target to perform control effect simulation evaluation on the M interval parameter sets in turn to obtain the out-of-step control simulation effects of M interval parameters; performing interval optimization on the M parameter intervals based on the out-of-step control simulation effects of the M interval parameters to determine the flight control strategy search interval; performing global comparison and optimization of parameters in the flight control strategy search interval to determine the UAV flight control strategy parameters.

[0054] The specific configuration of the flight control strategy parameter determination module 60 will be described in detail below. The flight control strategy parameter determination module 60 may further include: performing parameter preset ratio screening based on the M interval parameter out-of-step control simulation effects to obtain M interval control strategy parameters; calculating the effect mean of the M interval control strategy parameters as the M parameter interval control effects; performing interval optimization on the M parameter intervals according to the M parameter interval control effects to determine the first strategy parameter interval; iteratively dividing and optimizing the first strategy parameter interval until the preset interval division times are reached to determine the flight control strategy search interval.

[0055] The specific configuration of the flight control strategy parameter determination module 60 will be described in detail below. The flight control strategy parameter determination module 60 may further include: using the UAV flight control strategy parameters to perform flight control monitoring on the target UAV to obtain the UAV flight state feedback parameters; performing out-of-step optimization analysis based on the UAV flight state feedback parameters to determine the strategy parameter optimization variation rules; performing mutation updates on the UAV flight control strategy parameters according to the strategy parameter optimization variation rules to obtain multiple flight control strategy variation parameters, and determining the UAV flight control strategy optimization parameters based on the multiple flight control strategy variation parameters.

[0056] The platform for unmanned aerial vehicle flight management by predicting out-of-step provided in the embodiment of the present invention can execute the method for unmanned aerial vehicle flight management by predicting out-of-step provided in any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0057] Although the present application makes various references to certain modules in the platform according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be implemented; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0058] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for managing UAV flight by predicting out-of-step, characterized in that: The method comprises: Mining and acquiring a UAV out-of-step data set, performing correlation feature extraction on the UAV out-of-step data set, and obtaining a UAV out-of-step feature data set; Acquire the drone application attribute information, perform data sampling and prediction training identification on the drone out-of-step feature data set based on the drone application attribute information, and build a drone out-of-step prediction network model library; A sensor group is mounted on the target UAV, wherein the sensor group includes a position sensor, a speed sensor, an attitude sensor, an air pressure sensor, and a temperature and humidity sensor, and the UAV flight status data stream and the operating environment data stream are collected and acquired through the sensor group; Based on the application attribute information of the target UAV, the characteristics are matched with the UAV out-of-step prediction network model library to obtain the target UAV out-of-step prediction network model; The out-of-step prediction network model of the target UAV is used to predict the out-of-step of the flight status data stream and the operating environment data stream of the UAV, and obtain the out-of-step prediction abnormal parameters; According to the out-of-step prediction abnormal parameters, a UAV flight control strategy space is constructed, an out-of-step balance target is defined to control and optimize the UAV flight control strategy space, determine the UAV flight control strategy parameters, and use the UAV flight control strategy parameters to perform out-of-step flight control on the target UAV.

2. The method for managing the flight of a UAV by predicting out-of-step according to claim 1, characterized in that: The construction of the UAV out-of-step prediction network model library includes: Extracting elements from the drone application attribute information to obtain drone application attribute elements, where the drone application attribute elements include drone specification elements and application scenario elements; Based on the UAV out-of-sync data set, attribute parameters of the UAV specification elements and application scenario elements are extracted and integrated to obtain a UAV application attribute parameter set; Performing data mapping sampling on the UAV out-of-step feature data set through the UAV application attribute parameter set to obtain a multi-application attribute parameter out-of-step feature data set; Perform out-of-step prediction training on the multi-application attribute parameter out-of-step feature data sets respectively to obtain a multi-application attribute parameter out-of-step prediction network model set; The drone application attribute parameter set is used to calibrate the attribute parameters of the multi-application attribute parameter out-of-step prediction network model set, and the drone out-of-step prediction network model library is constructed.

3. The method for managing unmanned aerial vehicle flight by predicting out-of-step as claimed in claim 2, characterized in that: The step of obtaining a multi-application attribute parameter out-of-sync feature data set includes: By performing feature association matching between the drone application attribute parameter set and the drone out-of-step feature data set, a multi-attribute parameter preliminary association feature data set is obtained; The multi-attribute parameter primary-order correlation feature data set is used as a sampling center, and neighboring data is collected within the UAV out-of-step feature data set as a multi-attribute parameter second-order correlation feature data set; Traversing the second-order correlation feature data set of the multi-attribute parameters, performing K-order data collection in the UAV out-of-step feature data set, and obtaining a K-order multi-attribute parameter correlation feature data set, wherein 2≤K≤8; The multi-attribute parameter primary-order correlation feature data set, the multi-attribute parameter second-order correlation feature data set, and even the K-order multi-attribute parameter correlation feature data set are sampled and mapped and integrated to obtain the multi-application attribute parameter out-of-step feature data set.

4. The method for managing unmanned aerial vehicle flight by predicting out-of-step as claimed in claim 1, characterized in that: The construction of the UAV flight control strategy space includes: Acquire flight control strategy dimension information, perform flight control strategy analysis on the out-of-step prediction abnormality parameter according to the flight control strategy dimension information, and obtain a flight control strategy dimension parameter threshold; Randomly selecting a plurality of control strategy dimension parameters within the flight control strategy dimension parameter threshold, evaluating and optimizing the plurality of control strategy dimension parameters, and determining an initial control strategy parameter set; Cross-combining and mutating the primary control strategy parameter set to obtain a descendant control strategy parameter set, and screening the descendant control strategy parameter set for optimal parameters to obtain an optimal descendant strategy parameter set; Based on the initial control strategy parameter set and the superior descendant strategy parameter set, the UAV flight control strategy space is constructed in combination.

5. The method for managing unmanned aerial vehicle flight by predicting out-of-step as claimed in claim 1, characterized in that: Determining the flight control strategy parameters of the UAV includes: Dividing M parameter intervals in the UAV flight control strategy space, and selecting M interval parameter sets in the M parameter intervals respectively; Define the out-of-step balancing target to perform control effect simulation evaluation on the M interval parameter sets in turn to obtain the out-of-step control simulation effect of the M interval parameters; Based on the simulation effects of the out-of-step control of the M interval parameters, the M parameter intervals are optimized to determine the flight control strategy search interval; A global comparison and optimization of parameters is performed within the flight control strategy search interval to determine the flight control strategy parameters of the UAV.

6. The method for managing the flight of a UAV by predicting out-of-step as claimed in claim 5, characterized in that: Determining the flight control strategy search interval includes: Based on the simulation effects of the M interval parameter out-of-step control, the preset proportions of the parameters are screened to obtain the M interval control strategy parameters; The mean value calculation result of the effect of the M interval control strategy parameters is used as the M parameter interval control effect; Performing interval optimization on the M parameter intervals according to the control effects of the M parameter intervals to determine a first strategy parameter interval; The first strategy parameter interval is iteratively divided and optimized until a preset interval division number is reached, and the flight control strategy search interval is determined.

7. The method for managing unmanned aerial vehicle flight by predicting out-of-step as claimed in claim 1, characterized in that: The method comprises: Using the UAV flight control strategy parameters to perform flight control monitoring on the target UAV to obtain UAV flight state feedback parameters; Perform out-of-step optimization analysis based on the UAV flight state feedback parameters to determine the strategy parameter optimization variation rules; The UAV flight control strategy parameters are mutated and updated according to the strategy parameter optimization and mutation rules to obtain multiple flight control strategy variation parameters, and the UAV flight control strategy optimization parameters are determined based on the comparison and optimization of the multiple flight control strategy variation parameters.

8. A platform for UAV flight management based on out-of-step prediction, characterized in that: The platform is used to implement the method for predicting out-of-step and managing UAV flight as described in any one of claims 1 to 7, and the platform includes: A correlation feature extraction module is used to mine and obtain a UAV out-of-step data set, extract correlation features from the UAV out-of-step data set, and obtain a UAV out-of-step feature data set; A prediction network model library construction module is used to obtain drone application attribute information, perform data sampling and prediction training identification on the drone out-of-step feature data set based on the drone application attribute information, and build a drone out-of-step prediction network model library; The UAV data stream acquisition module is used to carry a sensor group on the target UAV, wherein the sensor group includes a position sensor, a speed sensor, an attitude sensor, an air pressure sensor, and a temperature and humidity sensor, and acquires the flight status data stream and the operating environment data stream of the UAV through the sensor group; A feature matching module is used to perform feature matching with the UAV out-of-step prediction network model library based on the application attribute information of the target UAV to obtain the target UAV out-of-step prediction network model; A step-out prediction identification module is used to predict and identify the step-out of the UAV flight status data stream and the operating environment data stream through the target UAV step-out prediction network model to obtain the step-out prediction abnormality parameters; The flight control strategy parameter determination module is used to construct a UAV flight control strategy space according to the out-of-step prediction abnormal parameters, define the out-of-step balance target to control and optimize the UAV flight control strategy space, determine the UAV flight control strategy parameters, and perform out-of-step flight control on the target UAV through the UAV flight control strategy parameters.

Images