Unmanned aerial vehicle control method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an unmanned aerial vehicle control method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a preset control strategy in response to an inspection task starting operation of an unmanned aerial vehicle, and controlling the unmanned aerial vehicle to start executing an inspection task through the preset control strategy; then, acquiring surrounding environment data of the unmanned aerial vehicle through a sensor of the unmanned aerial vehicle, and performing feature extraction on the surrounding environment data of the unmanned aerial vehicle through a convolutional neural network to obtain key features; and the unmanned aerial vehicle is controlled and adjusted based on the key features through a preset incremental learning algorithm until the unmanned aerial vehicle completes the inspection task, so that the unmanned aerial vehicle is dynamically adjusted based on the environment data collected in real time. When the environment changes, the incremental learning algorithm can update the control strategy according to the new data, so that the unmanned aerial vehicle can flexibly cope with various complex power station environments, thereby improving the accuracy and efficiency of routing inspection, and powerfully guaranteeing the safe and stable operation of the power station.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and in particular to a method, device, electronic equipment and storage medium for controlling an unmanned aerial vehicle. Background Art

[0002] With the continuous development of the power system, the inspection tasks of power facilities, especially the inspection workload of power station equipment, are increasing. Traditional inspection methods often rely on manual or fixed automated equipment, which are inefficient and have certain limitations. As an emerging inspection tool, drones have been widely used in the field of power inspection due to their flexibility and efficiency.

[0003] Generally speaking, the power station environment is complex and changeable, and drones need to deal with multiple challenges when performing inspection tasks, such as complex power station structure, weather changes, electromagnetic interference, etc.

[0004] However, traditional drone control methods are usually based on fixed control strategies, which results in the drone being unable to efficiently perform inspection tasks when facing complex power plant environments (such as equipment failures, electromagnetic interference, weather changes, etc.), reducing inspection efficiency and accuracy. Summary of the invention

[0005] In view of the above problems, a method, device, electronic device and storage medium for controlling a drone are proposed to overcome the above problems or at least partially solve the above problems, including: A method for controlling an unmanned aerial vehicle, the method comprising: In response to the inspection task start operation of the UAV, a preset control strategy is obtained and the UAV is controlled by the preset control strategy to start executing the inspection task; Acquire the surrounding environment data of the drone through the sensor of the drone, and extract features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; The drone is controlled and adjusted based on the key features through a preset incremental learning algorithm until the drone completes the inspection task.

[0006] Optionally, the feature extraction of the drone surrounding environment data by a convolutional neural network to obtain key features includes: Based on the dimensional information of the drone surrounding environment data, a multi-layer convolution operation is performed on the drone surrounding environment data to obtain the key features.

[0007] Optionally, performing a multi-layer convolution operation on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features includes: Determine one-dimensional data in the drone surrounding environment data, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; Determining two-dimensional data in the surrounding environment data of the drone, and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; Determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; The key feature is obtained according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

[0008] Optionally, determining one-dimensional data in the drone surrounding environment data, and performing a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result includes: One-dimensional data is determined in the surrounding environment data of the drone, and a one-dimensional convolution operation is performed on the one-dimensional data using the following formula to obtain a one-dimensional convolution result:

[0009] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0010] Optionally, determining two-dimensional data in the surrounding environment data of the drone, and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result includes: Determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data using the following formula to obtain a two-dimensional convolution result:

[0011] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0012] Optionally, determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result includes: Determine three-dimensional data in the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data using the following formula to obtain a three-dimensional convolution result:

[0013] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0014] Optionally, the controlling and adjusting the UAV based on the key features by using a preset incremental learning algorithm includes: The flight attitude and / or path and / or inspection operation of the UAV is controlled and adjusted based on the key features through a preset incremental learning algorithm.

[0015] Optionally, the method further includes: Based on the inspection tasks completed by the drone, determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor during the drone inspection, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; Reinforcement learning is performed on the drone based on the reward function to obtain an optimal control strategy, and the preset control strategy is replaced by the optimal control strategy.

[0016] Optionally, the formula of the reward function is:

[0017] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the UAV, R sec This is the safety factor of the drone.

[0018] Optionally, the performing reinforcement learning on the drone based on the reward function to obtain an optimal control strategy includes: Based on the reward function, the UAV state space parameters and the UAV action space parameters, reinforcement learning is performed on the UAV in a parallel system to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

[0019] Optionally, based on the reward function, the state space parameters of the drone and the action space parameters of the drone, reinforcement learning is performed on the drone in a parallel system to obtain the optimal control strategy, including: Based on the reward function, the drone state space parameters and the drone action space parameters, the drone is reinforced learned in a parallel system according to the following mathematical expression to obtain the optimal control strategy:

[0020] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

[0021] Optionally, the inspection task is an electric power facility inspection task, and the sensor of the UAV includes at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

[0022] Optionally, the drone surrounding environment data includes at least one or more of appearance image data of power facilities, temperature distribution data of power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

[0023] A drone control device, comprising: An inspection task starting module is used to respond to the inspection task starting operation of the UAV, obtain a preset control strategy and control the UAV to start executing the inspection task through the preset control strategy; A feature extraction module, used to obtain the surrounding environment data of the drone through the sensor of the drone, and extract features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; A control adjustment module is used to control and adjust the UAV based on the key features by using a preset incremental learning algorithm until the UAV completes the inspection task.

[0024] Optionally, the feature extraction module includes: The feature extraction submodule is used to perform a multi-layer convolution operation on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features.

[0025] Optionally, the feature extraction submodule includes: A one-dimensional convolution unit, used to determine one-dimensional data in the surrounding environment data of the drone, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; A two-dimensional convolution unit, used to determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; A three-dimensional convolution unit, used to determine three-dimensional data in the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; A feature extraction unit is used to obtain the key feature according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

[0026] Optionally, the one-dimensional convolution unit includes: The one-dimensional convolution subunit is used to determine one-dimensional data in the surrounding environment data of the drone, and perform a one-dimensional convolution operation on the one-dimensional data through the following formula to obtain a one-dimensional convolution result:

[0027] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0028] Optionally, the two-dimensional convolution unit includes: The two-dimensional convolution subunit is used to determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data through the following formula to obtain a two-dimensional convolution result:

[0029] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0030] Optionally, the three-dimensional convolution unit includes: The three-dimensional convolution subunit is used to determine the three-dimensional data in the surrounding environment data of the UAV, and perform a three-dimensional convolution operation on the three-dimensional data by the following formula to obtain a three-dimensional convolution result:

[0031] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0032] Optionally, the control adjustment module includes: The control adjustment submodule is used to control and adjust the flight attitude and / or path and / or inspection operation of the UAV based on the key features through a preset incremental learning algorithm.

[0033] Optionally, the device further comprises: A reward function determination module, used to determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor of the drone during the inspection based on the inspection tasks completed by the drone, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; A reinforcement learning module is used to perform reinforcement learning on the drone based on the reward function to obtain an optimal control strategy, and replace the preset control strategy with the optimal control strategy.

[0034] Optionally, the formula of the reward function is:

[0035] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the UAV, R sec This is the safety factor of the drone.

[0036] Optionally, the reinforcement learning module includes: The reinforcement learning submodule is used to perform reinforcement learning on the UAV in a parallel system based on the reward function, the UAV state space parameters and the UAV action space parameters to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

[0037] Optionally, the reinforcement learning submodule includes: A reinforcement learning unit is used to perform reinforcement learning on the drone in a parallel system according to the following mathematical expression based on the reward function, the drone state space parameters and the drone action space parameters to obtain the optimal control strategy:

[0038] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

[0039] Optionally, the inspection task is an electric power facility inspection task, and the sensor of the UAV includes at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

[0040] Optionally, the drone surrounding environment data includes at least one or more of appearance image data of power facilities, temperature distribution data of power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

[0041] An electronic device comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements the drone control method as described above when executed by the processor.

[0042] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the drone control method described above is implemented.

[0043] The embodiments of the present invention have the following advantages: In an embodiment of the present invention, a method for controlling a drone is proposed. By responding to the inspection task start operation of the drone, a preset control strategy is obtained and the drone is controlled to start the inspection task through the preset control strategy; then the drone's surrounding environment data is obtained through the drone's sensor, and the drone's surrounding environment data is extracted through a convolutional neural network to obtain key features; and then the drone is controlled and adjusted based on the key features through a preset incremental learning algorithm until the drone completes the inspection task, thereby realizing dynamic adjustment of the drone based on real-time collected environmental data. When the environment changes, the incremental learning algorithm can update the control strategy according to the new data, so that the drone can flexibly respond to various complex power station environments, thereby improving the accuracy and efficiency of the inspection and effectively ensuring the safe and stable operation of the power station. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the accompanying drawings required for use in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0045] Figure 1 is a flowchart of steps of a drone control method provided by some embodiments of the present invention; Figure 2 is a schematic structural diagram of a drone control device provided in some embodiments of the present invention; Figure 3 is a block diagram of an electronic device provided in some embodiments of the present invention; Figure 4 is a schematic diagram of a computer-readable medium provided in some embodiments of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] With the continuous development of the power system, the inspection tasks of power facilities, especially the inspection workload of power station equipment, are increasing. Traditional inspection methods often rely on manual or fixed automated equipment, which are inefficient and have certain limitations. As an emerging inspection tool, drones have been widely used in the field of power inspection due to their flexibility and efficiency.

[0048] Generally speaking, the power station environment is complex and changeable, and drones need to deal with multiple challenges when performing inspection tasks, such as complex power station structure, weather changes, electromagnetic interference, etc.

[0049] However, traditional drone control methods are usually based on fixed control strategies, which results in the drone being unable to efficiently perform inspection tasks when facing complex power plant environments (such as equipment failures, electromagnetic interference, weather changes, etc.), reducing inspection efficiency and accuracy.

[0050] In the embodiment of the present invention, based on the core technical concept of combining feature extraction with incremental learning algorithm to adjust the inspection control of the UAV, the UAV control method in the related art is improved. The present invention will be described in detail with reference to the accompanying drawings below: Reference Figure 1 , shows a flowchart of a method for controlling a drone provided by some embodiments of the present invention, which may specifically include the following steps: Step 101, in response to the inspection task start operation of the UAV, obtaining a preset control strategy and controlling the UAV to start executing the inspection task through the preset control strategy; In a specific implementation, when the UAV receives an inspection task execution command for the first time, it can first control the UAV to start executing the inspection task according to an initial preset control strategy, and then try to control and adjust the UAV according to the surrounding environment data obtained by the UAV.

[0051] Step 102, obtaining the surrounding environment data of the drone through the sensor of the drone, and extracting features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; In a specific implementation, the inspection mission of the drone can be the inspection mission of power facilities, and the drone can be equipped with a variety of sensors, including but not limited to optical cameras, infrared thermal imagers, wind speed and direction sensors, electromagnetic interference sensors, etc. These sensors are used to collect real-time data about the drone's surrounding environment, such as appearance image data of power station equipment, temperature distribution data of equipment, current meteorological conditions data, and electromagnetic environment data.

[0052] On this basis, the convolutional neural network can be used to extract features from the drone's surrounding environment data obtained by the drone's sensors, and to extract key features from the collected data, such as the shape, color, texture, etc. of the power station equipment, as well as characteristic parameters in the meteorological data and electromagnetic data; thereby providing reference data for subsequent drone control adjustments.

[0053] In some embodiments of the present invention, the feature extraction of the drone surrounding environment data by a convolutional neural network to obtain key features includes: Based on the dimensional information of the drone surrounding environment data, a multi-layer convolution operation is performed on the drone surrounding environment data to obtain the key features.

[0054] In a specific implementation, based on the dimensional information of the drone's surrounding environment data, multi-layer convolution operations can be performed on the drone's surrounding environment data to obtain key features; for example, one-dimensional data (such as wind speed data) can be determined in the drone's surrounding environment data, and a one-dimensional convolution operation can be performed on the one-dimensional data to obtain a one-dimensional convolution result, and two-dimensional data (such as image data) can be determined in the drone's surrounding environment data, and a two-dimensional convolution operation can be performed on the two-dimensional data to obtain a two-dimensional convolution result, and three-dimensional data (such as electric field strength data) can be determined in the drone's surrounding environment data and a three-dimensional convolution operation can be performed on the three-dimensional data to obtain a three-dimensional convolution result; thereby completing multi-layer convolution operations on the drone's surrounding environment data, and further performing pooling operations to obtain key features, thereby improving the accuracy of feature extraction and ensuring the comprehensiveness and meticulousness of environmental perception.

[0055] In some embodiments of the present invention, performing a multi-layer convolution operation on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features includes: Determine one-dimensional data in the drone surrounding environment data, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; Determining two-dimensional data in the surrounding environment data of the drone, and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; Determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; The key feature is obtained according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

[0056] In practical applications, one-dimensional data (such as wind speed data) can be determined in the drone's surrounding environment data, and a one-dimensional convolution operation can be performed on the one-dimensional data to obtain a one-dimensional convolution result. Two-dimensional data (such as image data) can be determined in the drone's surrounding environment data, and a two-dimensional convolution operation can be performed on the two-dimensional data to obtain a two-dimensional convolution result. Three-dimensional data (such as electric field strength data) can be determined in the drone's surrounding environment data and a three-dimensional convolution operation can be performed on the three-dimensional data to obtain a three-dimensional convolution result. In this way, multi-layer convolution operations on the drone's surrounding environment data are completed, and pooling operations can be further performed to obtain key features.

[0057] For example, for a one-dimensional data series (such as wind speed data) , the one-dimensional convolution operation can be defined as:

[0058] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight and , b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0059] For a two-dimensional data (such as an image) and a convolution kernel K(i,j), the two-dimensional convolution operation can be defined as:

[0060] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0061] For three-dimensional data (such as electric field strength data) , the three-dimensional convolution operation can be expressed as:

[0062] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0063] Through multi-layer convolution and pooling operations, CNN (Convolutional Neural Networks) can automatically learn high-level features in the collected data, such as the shape, color and texture of power station equipment, meteorological conditions, electromagnetic interference, etc., to provide data support for subsequent processes, ensuring that all types of data can be effectively processed and analyzed during the inspection process.

[0064] In some embodiments of the present invention, determining one-dimensional data in the surrounding environment data of the drone, and performing a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result includes: One-dimensional data is determined in the surrounding environment data of the drone, and a one-dimensional convolution operation is performed on the one-dimensional data using the following formula to obtain a one-dimensional convolution result:

[0065] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0066] In practical applications, as mentioned above, for a one-dimensional data sequence (such as wind speed data) , the one-dimensional convolution operation can be defined as:

[0067] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight and , b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0068] In this embodiment, the one-dimensional convolution operation is concretized by a formula, which helps to improve the operability and accuracy of the convolution calculation and ensure the effective processing of one-dimensional data (such as wind speed, time series, etc.).

[0069] In some embodiments of the present invention, determining two-dimensional data in the surrounding environment data of the drone, and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result includes: Determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data using the following formula to obtain a two-dimensional convolution result:

[0070] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0071] In practical applications, as mentioned above, for a two-dimensional data (such as an image) and a convolution kernel K(i,j), the two-dimensional convolution operation can be defined as:

[0072] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0073] In this embodiment, by clarifying the specific calculation process of the two-dimensional convolution operation and providing a specific design of the convolution kernel, two-dimensional image data (such as the appearance image of a power facility) can be efficiently and accurately extracted, providing accurate data support for subsequent control decisions.

[0074] In some embodiments of the present invention, determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result includes: Determine three-dimensional data from the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data using the following formula to obtain a three-dimensional convolution result:

[0075] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0076] In practical applications, as mentioned above, for three-dimensional data (such as electric field intensity data) , the three-dimensional convolution operation can be expressed as:

[0077] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0078] In this embodiment, by defining a three-dimensional convolution operation, it is possible to process three-dimensional environmental data (such as electric field strength, three-dimensional images, etc.), so that the drone can cope with more complex environmental data types. The introduction of this operation makes the feature extraction process more comprehensive and efficient, and improves the drone's perception of complex electromagnetic environments.

[0079] In some embodiments of the present invention, the inspection task is an inspection task for power facilities, and the sensor of the drone includes at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

[0080] In practical applications, the inspection mission of drones can be the inspection mission of power facilities, and drones can be equipped with a variety of sensors, including but not limited to optical cameras, infrared thermal imagers, wind speed and direction sensors, electromagnetic interference sensors, etc. These sensors are used to collect real-time data about the drone's surrounding environment, such as appearance image data of power station equipment, temperature distribution data of equipment, current meteorological conditions data, and electromagnetic environment data.

[0081] On this basis, the convolutional neural network can be used to extract features from the drone's surrounding environment data obtained by the drone's sensors, and to extract key features from the collected data, such as the shape, color, texture, etc. of the power station equipment, as well as characteristic parameters in the meteorological data and electromagnetic data; thereby providing reference data for subsequent drone control adjustments.

[0082] In some embodiments of the present invention, the drone surrounding environment data includes at least one or more of appearance image data of power facilities, temperature distribution data of power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

[0083] In practical applications, the drone's sensors can be used to collect real-time data about the drone's surrounding environment, such as appearance image data of power station equipment, temperature distribution data of equipment, current meteorological conditions data, and electromagnetic environment data.

[0084] On this basis, the convolutional neural network can be used to extract features from the drone's surrounding environment data obtained by the drone's sensors, and to extract key features from the collected data, such as the shape, color, texture, etc. of the power station equipment, as well as characteristic parameters in the meteorological data and electromagnetic data; thereby providing reference data for subsequent drone control adjustments.

[0085] In this embodiment, by clarifying the types of environmental data (such as the appearance image of power facilities, temperature distribution, meteorological conditions, electromagnetic environment, etc.), it is ensured that the drone can fully collect and process various environmental information, thereby improving the comprehensiveness and accuracy of the inspection. The fusion of different types of data can enhance the system's perception and processing capabilities of complex power station environments.

[0086] Step 103: Control and adjust the UAV based on the key features using a preset incremental learning algorithm until the UAV completes the inspection task.

[0087] In the specific implementation, when the UAV starts the inspection mission, the environmental data collected by the sensor in real time can be transmitted to the incremental learning module. The incremental learning algorithm in the incremental learning module updates the existing control model according to the received environmental data, that is, the control of the UAV is adjusted.

[0088] For example, taking the attitude adjustment of a drone as an example, assuming that the attitude of the drone can be represented by a vector x, and the control input is u, its dynamic equation can be expressed as:

[0089] The incremental learning algorithm can optimize the attitude of the drone by continuously adjusting the control input u. The attitude control algorithm based on incremental learning can use the following optimization objectives:

[0090] in, is the desired attitude, and T is the time step. Similarly, the control module of the UAV can adjust the flight attitude and / or path and / or inspection operation of the UAV in real time according to the updated control strategy output by the incremental learning module. For example, when the incremental learning algorithm finds that the electromagnetic interference in a certain area is strong, the control module can adjust the flight path of the UAV to avoid the area or take anti-interference measures (such as adjusting the flight attitude of the UAV) to ensure the smooth progress of the inspection task.

[0091] In some embodiments of the present invention, the controlling and adjusting the UAV based on the key features by using a preset incremental learning algorithm includes: The flight attitude and / or path and / or inspection operation of the UAV is controlled and adjusted based on the key features through a preset incremental learning algorithm.

[0092] In practical applications, as mentioned above, the control module of the drone can adjust the flight attitude and / or path and / or inspection operation of the drone in real time according to the updated control strategy output by the incremental learning module. For example, when the incremental learning algorithm finds that the electromagnetic interference in a certain area is strong, the control module can adjust the flight path of the drone to avoid the area or take anti-interference measures (such as adjusting the flight attitude of the drone) to ensure the smooth progress of the inspection task.

[0093] In some embodiments of the present invention, the method further comprises: Based on the inspection tasks completed by the drone, determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor during the drone inspection, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; Reinforcement learning is performed on the drone based on the reward function to obtain an optimal control strategy, and the preset control strategy is replaced by the optimal control strategy.

[0094] In the specific implementation, after the inspection task is completed, a parallel system can be built for the drone based on the feature extraction results as an online learning environment; the control performance index can also be built by comprehensively considering factors such as the accuracy, efficiency, stability and safety of the inspection as a reward function. The reward function can comprehensively consider the accuracy of the inspection R acc , efficiency R eff 、Stability of UAV sta and safety R sec For example, when the drone can accurately identify the fault of power plant equipment and efficiently complete the inspection task, a positive reward can be given; conversely, when there is an inspection omission or the drone's flight is unstable, a negative reward can be given. In this way, the inspection behavior of the drone can be optimized by continuously adjusting the reward function.

[0095] Among them, the formula of the reward function can be:

[0096] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the UAV, R sec is the safety factor of the drone; the above weight coefficient should satisfy .

[0097] On this basis, the UAV can be reinforced with learning based on the reward function to obtain the optimal control strategy, and the preset control strategy can be replaced with the optimal control strategy; specifically, the optimal control strategy can be learned through interaction with the environment, for example, the four-tuple can be defined: M(E,S,A,R), where: Environment: A parallel system containing characteristic equations and servo models; State (UAV state space parameters): UAV flight state parameters, UAV control deviation parameters, UAV observation and identification result parameters; Action (UAV action space parameters): UAV controller parameters, UAV allocation law parameters, UAV estimator parameters; Reward: Evaluation function of the drone’s tracking error, stability, remaining control capability, etc. The four-tuple M (E, S, A, R) defines the basic elements of reinforcement learning. When calculating the long-term cumulative reward, the current state can be determined based on the state space parameter S, the action can be selected from the action space parameter A, the action can be executed in the environment E, and the reward feedback can be obtained through the reward function R. By continuously iterating within this framework, the optimal control strategy can be gradually approached.

[0098] On this basis, an optimal control strategy can be found through reinforcement learning , which maximizes the long-term cumulative reward, and its mathematical expression can be:

[0099] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the drone action space parameter, s is the drone state space parameter, and t is time. In this formula, argmax means finding the parameter that can maximize the following formula. "E" stands for expectation, and "Σ" is the summation symbol, which sums the time step t. γ is the discount factor, which usually ranges from 0 to 1 and is used to balance immediate rewards and long-term rewards. The discount factor requires the drone to consider both the immediate rewards brought by the current action and the possible rewards in the future when making decisions. The purpose of this mathematical expression is to find an optimal control strategy in the process of reinforcement learning. (determined by the relevant parameters in the quaternary group), so that when the optimal control strategy is executed in the environment, the long-term cumulative reward is maximized; it enables the UAV to make optimal decisions according to the environment and mission requirements, and improve the accuracy and efficiency of the inspection mission and its adaptability to complex environments.

[0100] In this embodiment, by adopting the "offline + online" reinforcement learning mechanism to simultaneously learn the controller, the allocation rate, and estimate its parameter scheduling law, offline learning "features" and online learning "parameters", the efficiency of online reinforcement learning can be improved.

[0101] In some embodiments of the present invention, the formula of the reward function is:

[0102] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the UAV, R sec This is the safety factor of the drone.

[0103] In practical applications, as mentioned above, the formula of the reward function can be:

[0104] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the UAV, R sec is the safety factor of the drone; the above weight coefficient should satisfy .

[0105] On this basis, the UAV can be reinforced with learning based on the reward function to obtain the optimal control strategy, and then the preset control strategy can be replaced with the optimal control strategy.

[0106] In some embodiments of the present invention, the performing reinforcement learning on the drone based on the reward function to obtain an optimal control strategy includes: Based on the reward function, the UAV state space parameters and the UAV action space parameters, reinforcement learning is performed on the UAV in a parallel system to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

[0107] In practical applications, based on the reward function, the state space parameters of the drone and the action space parameters of the drone, the drone can be reinforced in a parallel system to obtain the optimal control strategy. That is, the optimal control strategy can be learned through interaction with the environment. For example, the four-tuple can be defined: M(E,S,A,R), where: Environment: A parallel system containing characteristic equations and servo models; State (UAV state space parameters): UAV flight state parameters, UAV control deviation parameters, UAV observation and identification result parameters; Action (UAV action space parameters): UAV controller parameters, UAV allocation law parameters, UAV estimator parameters; Reward: Evaluation function of the drone’s tracking error, stability, remaining control capability, etc. The four-tuple M (E, S, A, R) defines the basic elements of reinforcement learning. When calculating the long-term cumulative reward, the current state can be determined based on the state space parameter S, the action can be selected from the action space parameter A, the action can be executed in the environment E, and the reward feedback can be obtained through the reward function R. By continuously iterating within this framework, the optimal control strategy can be gradually approached.

[0108] In some embodiments of the present invention, the method of performing reinforcement learning on the drone in a parallel system based on the reward function, the drone state space parameters and the drone action space parameters to obtain the optimal control strategy includes: Based on the reward function, the drone state space parameters and the drone action space parameters, the drone is reinforced learned in a parallel system according to the following mathematical expression to obtain the optimal control strategy:

[0109] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

[0110] In practical applications, as mentioned above, an optimal control strategy can be found through reinforcement learning , which maximizes the long-term cumulative reward, and its mathematical expression can be:

[0111] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the drone action space parameter, s is the drone state space parameter, and t is time. In this formula, argmax means finding the parameter that can maximize the following formula. "E" stands for expectation, and "Σ" is the summation symbol, which sums the time step t. γ is the discount factor, which usually ranges from 0 to 1 and is used to balance immediate rewards and long-term rewards. The discount factor requires the drone to consider both the immediate rewards brought by the current action and the possible rewards in the future when making decisions. The purpose of this mathematical expression is to find an optimal control strategy in the process of reinforcement learning. (determined by the relevant parameters in the quaternary group), so that when the optimal control strategy is executed in the environment, the long-term cumulative reward reaches the maximum; it enables the UAV to make the best decision based on the optimal control strategy, according to the environment and task requirements, and improve the accuracy and efficiency of the inspection task and its adaptability to complex environments.

[0112] In an embodiment of the present invention, a method for controlling a drone is proposed. By responding to the inspection task start operation of the drone, a preset control strategy is obtained and the drone is controlled to start the inspection task through the preset control strategy; then the drone's surrounding environment data is obtained through the drone's sensor, and the drone's surrounding environment data is extracted through a convolutional neural network to obtain key features; and then the drone is controlled and adjusted based on the key features through a preset incremental learning algorithm until the drone completes the inspection task, thereby realizing dynamic adjustment of the drone based on real-time collected environmental data. When the environment changes, the incremental learning algorithm can update the control strategy according to the new data, so that the drone can flexibly respond to various complex power station environments, thereby improving the accuracy and efficiency of the inspection and effectively ensuring the safe and stable operation of the power station.

[0113] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0114] Reference Figure 2 , shows a schematic diagram of the structure of a drone control device provided by some embodiments of the present invention, which may specifically include the following modules: The inspection task starting module 201 is used to respond to the inspection task starting operation of the UAV, obtain a preset control strategy and control the UAV to start executing the inspection task through the preset control strategy; A feature extraction module 202 is used to obtain the surrounding environment data of the drone through the sensor of the drone, and extract features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; The control adjustment module 203 is used to control and adjust the UAV based on the key features by using a preset incremental learning algorithm until the UAV completes the inspection task.

[0115] In some embodiments of the present invention, the feature extraction module 202 includes: The feature extraction submodule is used to perform a multi-layer convolution operation on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features.

[0116] In some embodiments of the present invention, the feature extraction submodule includes: A one-dimensional convolution unit, used to determine one-dimensional data in the surrounding environment data of the drone, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; A two-dimensional convolution unit, used to determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; A three-dimensional convolution unit, used to determine three-dimensional data in the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; A feature extraction unit is used to obtain the key feature according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

[0117] In some embodiments of the present invention, the one-dimensional convolution unit includes: The one-dimensional convolution subunit is used to determine one-dimensional data in the surrounding environment data of the drone, and perform a one-dimensional convolution operation on the one-dimensional data through the following formula to obtain a one-dimensional convolution result:

[0118] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0119] In some embodiments of the present invention, the two-dimensional convolution unit includes: The two-dimensional convolution subunit is used to determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data through the following formula to obtain a two-dimensional convolution result:

[0120] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0121] In some embodiments of the present invention, the three-dimensional convolution unit includes: The three-dimensional convolution subunit is used to determine the three-dimensional data in the surrounding environment data of the UAV, and perform a three-dimensional convolution operation on the three-dimensional data by the following formula to obtain a three-dimensional convolution result:

[0122] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0123] In some embodiments of the present invention, the control adjustment module 203 includes: The control adjustment submodule is used to control and adjust the flight attitude and / or path and / or inspection operation of the UAV based on the key features through a preset incremental learning algorithm.

[0124] In some embodiments of the present invention, the device further comprises: A reward function determination module, used to determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor of the drone during the inspection based on the inspection tasks completed by the drone, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; A reinforcement learning module is used to perform reinforcement learning on the drone based on the reward function to obtain an optimal control strategy, and replace the preset control strategy with the optimal control strategy.

[0125] In some embodiments of the present invention, the formula of the reward function is:

[0126] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w secis the safety weight coefficient of the UAV, R sec This is the safety factor of the drone.

[0127] In some embodiments of the present invention, the reinforcement learning module includes: The reinforcement learning submodule is used to perform reinforcement learning on the UAV in a parallel system based on the reward function, the UAV state space parameters and the UAV action space parameters to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

[0128] In some embodiments of the present invention, the reinforcement learning submodule includes: A reinforcement learning unit is used to perform reinforcement learning on the drone in a parallel system according to the following mathematical expression based on the reward function, the drone state space parameters and the drone action space parameters to obtain the optimal control strategy:

[0129] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

[0130] In some embodiments of the present invention, the inspection task is an inspection task for power facilities, and the sensor of the drone includes at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

[0131] In some embodiments of the present invention, the drone surrounding environment data includes at least one or more of appearance image data of power facilities, temperature distribution data of power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

[0132] Some embodiments of the present invention further provide a computer program product, including a computer program, which implements the above drone control method when executed by a processor.

[0133] In addition, an embodiment of the present invention further provides an electronic device, such as Figure 3 As shown, it includes a processor 301, a communication interface 302, a memory 303 and a communication bus 304, wherein the processor 301, the communication interface 302, and the memory 303 communicate with each other through the communication bus 304. Memory 303, used for storing computer programs; The processor 301 is used to execute the program stored in the memory 303 to implement the following steps: In response to the inspection task start operation of the UAV, a preset control strategy is obtained and the UAV is controlled by the preset control strategy to start executing the inspection task; Acquire the surrounding environment data of the drone through the sensor of the drone, and extract features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; The drone is controlled and adjusted based on the key features through a preset incremental learning algorithm until the drone completes the inspection task.

[0134] In an optional embodiment of the present invention, the feature extraction of the drone surrounding environment data by a convolutional neural network to obtain key features includes: Based on the dimensional information of the drone surrounding environment data, a multi-layer convolution operation is performed on the drone surrounding environment data to obtain the key features.

[0135] In an optional embodiment of the present invention, the multi-layer convolution operation is performed on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features, including: Determine one-dimensional data in the drone surrounding environment data, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; Determining two-dimensional data in the surrounding environment data of the drone, and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; Determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; The key feature is obtained according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

[0136] In an optional embodiment of the present invention, determining one-dimensional data in the drone surrounding environment data, and performing a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result includes: One-dimensional data is determined in the surrounding environment data of the drone, and a one-dimensional convolution operation is performed on the one-dimensional data using the following formula to obtain a one-dimensional convolution result:

[0137] Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

[0138] In an optional embodiment of the present invention, determining two-dimensional data in the surrounding environment data of the drone, and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result includes: Determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data using the following formula to obtain a two-dimensional convolution result:

[0139] Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

[0140] In an optional embodiment of the present invention, determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result includes: Determine three-dimensional data from the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data using the following formula to obtain a three-dimensional convolution result:

[0141] Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

[0142] In an optional embodiment of the present invention, the controlling and adjusting the UAV based on the key features by using a preset incremental learning algorithm includes: The flight attitude and / or path and / or inspection operation of the UAV is controlled and adjusted based on the key features through a preset incremental learning algorithm.

[0143] In an optional embodiment of the present invention, the method further includes: Based on the inspection tasks completed by the drone, determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor during the drone inspection, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; Reinforcement learning is performed on the drone based on the reward function to obtain an optimal control strategy, and the preset control strategy is replaced by the optimal control strategy.

[0144] In an optional embodiment of the present invention, the formula of the reward function is:

[0145] Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the UAV, R sec This is the safety factor of the drone.

[0146] In an optional embodiment of the present invention, the performing reinforcement learning on the drone based on the reward function to obtain an optimal control strategy includes: Based on the reward function, the UAV state space parameters and the UAV action space parameters, reinforcement learning is performed on the UAV in a parallel system to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

[0147] In an optional embodiment of the present invention, the method of performing reinforcement learning on the drone in a parallel system based on the reward function, the drone state space parameters and the drone action space parameters to obtain the optimal control strategy includes: Based on the reward function, the drone state space parameters and the drone action space parameters, the drone is reinforced learned in a parallel system according to the following mathematical expression to obtain the optimal control strategy:

[0148] Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

[0149] In an optional embodiment of the present invention, the inspection task is an electric power facility inspection task, and the sensor of the drone includes at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

[0150] In an optional embodiment of the present invention, the drone surrounding environment data includes at least one or more of appearance image data of the power facilities, temperature distribution data of the power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

[0151] The communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0152] The communication interface is used for communication between the above terminal and other devices.

[0153] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0154] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0155] like Figure 4 As shown, in another embodiment provided by the present invention, a computer-readable storage medium 401 is also provided, in which instructions are stored. When the computer-readable storage medium is run on a computer, the computer executes the drone control method described in the above embodiment.

[0156] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0157] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0158] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0159] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0162] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0163] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the above elements.

[0164] The above is a detailed introduction to a drone control method, device, electronic device and storage medium provided. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technicians in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A drone control method, characterized in that: The method comprises: In response to the inspection task start operation of the UAV, a preset control strategy is obtained and the UAV is controlled by the preset control strategy to start executing the inspection task; Acquire the surrounding environment data of the drone through the sensor of the drone, and extract features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; The drone is controlled and adjusted based on the key features through a preset incremental learning algorithm until the drone completes the inspection task.

2. The method according to claim 1, characterized in that The feature extraction of the surrounding environment data of the drone is performed by the convolutional neural network to obtain key features, including: Based on the dimensional information of the drone surrounding environment data, a multi-layer convolution operation is performed on the drone surrounding environment data to obtain the key features.

3. The method according to claim 2, characterized in that The step of performing a multi-layer convolution operation on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features includes: Determine one-dimensional data in the drone surrounding environment data, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; Determine two-dimensional data in the surrounding environment data of the UAV, and perform a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; Determining three-dimensional data from the surrounding environment data of the drone, and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; The key feature is obtained according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

4. The method according to claim 3, characterized in that The determining one-dimensional data in the surrounding environment data of the drone, and performing a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result includes: One-dimensional data is determined in the surrounding environment data of the drone, and a one-dimensional convolution operation is performed on the one-dimensional data using the following formula to obtain a one-dimensional convolution result: Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

5. The method according to claim 3, characterized in that: The determining of two-dimensional data in the surrounding environment data of the drone and performing a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result includes: Determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data using the following formula to obtain a two-dimensional convolution result: Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

6. The method according to claim 3, characterized in that The determining of three-dimensional data in the surrounding environment data of the drone and performing a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result includes: Determine three-dimensional data from the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data using the following formula to obtain a three-dimensional convolution result: Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

7. The method according to claim 1, characterized in that The controlling and adjusting the UAV based on the key features by using a preset incremental learning algorithm includes: The flight attitude and / or path and / or inspection operation of the UAV is controlled and adjusted based on the key features through a preset incremental learning algorithm.

8. The method according to claim 1, characterized in that The method further comprises: Based on the inspection tasks completed by the drone, determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor during the drone inspection, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; Reinforcement learning is performed on the drone based on the reward function to obtain an optimal control strategy, and the preset control strategy is replaced by the optimal control strategy.

9. The method according to claim 8, characterized in that The formula of the reward function is: Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the drone, R sec This is the safety factor of the drone.

10. The method according to claim 8, characterized in that The performing reinforcement learning on the UAV based on the reward function to obtain an optimal control strategy includes: Based on the reward function, the UAV state space parameters and the UAV action space parameters, reinforcement learning is performed on the UAV in a parallel system to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

11. The method according to claim 10, characterized in that The method of performing reinforcement learning on the drone in a parallel system based on the reward function, the drone state space parameters and the drone action space parameters to obtain the optimal control strategy includes: Based on the reward function, the drone state space parameters and the drone action space parameters, the drone is reinforced learned in a parallel system according to the following mathematical expression to obtain the optimal control strategy: Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

12. The method according to claim 1, characterized in that The inspection task is an inspection task for power facilities, and the sensors of the UAV include at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

13. The method according to claim 1, characterized in that The drone surrounding environment data includes at least one or more of appearance image data of power facilities, temperature distribution data of power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

14. A drone control device, characterized in that: The device comprises: An inspection task starting module is used to respond to the inspection task starting operation of the UAV, obtain a preset control strategy and control the UAV to start executing the inspection task through the preset control strategy; A feature extraction module, used to obtain the surrounding environment data of the drone through the sensor of the drone, and extract features of the surrounding environment data of the drone through a convolutional neural network to obtain key features; A control adjustment module is used to control and adjust the UAV based on the key features by using a preset incremental learning algorithm until the UAV completes the inspection task.

15. The device according to claim 14, characterized in that The feature extraction module comprises: The feature extraction submodule is used to perform a multi-layer convolution operation on the drone surrounding environment data based on the dimensional information of the drone surrounding environment data to obtain the key features.

16. The device according to claim 15, characterized in that The feature extraction submodule comprises: A one-dimensional convolution unit, used to determine one-dimensional data in the surrounding environment data of the drone, and perform a one-dimensional convolution operation on the one-dimensional data to obtain a one-dimensional convolution result; A two-dimensional convolution unit, used to determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data to obtain a two-dimensional convolution result; A three-dimensional convolution unit, used to determine three-dimensional data in the surrounding environment data of the drone, and perform a three-dimensional convolution operation on the three-dimensional data to obtain a three-dimensional convolution result; A feature extraction unit is used to obtain the key feature according to the one-dimensional convolution result, the two-dimensional convolution result and the three-dimensional convolution result.

17. The device according to claim 16, characterized in that The one-dimensional convolution unit comprises: The one-dimensional convolution subunit is used to determine one-dimensional data in the surrounding environment data of the drone, and perform a one-dimensional convolution operation on the one-dimensional data through the following formula to obtain a one-dimensional convolution result: Among them, y is the one-dimensional convolution result, w is the convolution kernel weight, b is the bias, k is the convolution kernel size, and v is the one-dimensional data.

18. The device according to claim 16, characterized in that The two-dimensional convolution unit comprises: The two-dimensional convolution subunit is used to determine two-dimensional data in the surrounding environment data of the drone, and perform a two-dimensional convolution operation on the two-dimensional data through the following formula to obtain a two-dimensional convolution result: Among them, S(x,y) is the two-dimensional convolution result, I is the two-dimensional data, K(i,j) is the two-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, and b is the radius of the convolution kernel in the y direction.

19. The device according to claim 16, characterized in that The three-dimensional convolution unit comprises: The three-dimensional convolution subunit is used to determine the three-dimensional data in the surrounding environment data of the UAV, and perform a three-dimensional convolution operation on the three-dimensional data by the following formula to obtain a three-dimensional convolution result: Among them, S(x,y,z,t) is the three-dimensional convolution result, E is the three-dimensional data, K(i,j,k) is the three-dimensional convolution kernel, a is the radius of the convolution kernel in the x direction, b is the radius of the convolution kernel in the y direction, c is the radius of the convolution kernel in the z direction, and t is time.

20. The device according to claim 14, characterized in that The control adjustment module comprises: The control adjustment submodule is used to control and adjust the flight attitude and / or path and / or inspection operation of the UAV based on the key features through a preset incremental learning algorithm.

21. The device according to claim 14, characterized in that The device also includes: A reward function determination module, used to determine the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor of the drone during the inspection based on the inspection tasks completed by the drone, and determine the reward function based on the inspection accuracy factor, inspection efficiency factor, drone stability factor and drone safety factor; A reinforcement learning module is used to perform reinforcement learning on the drone based on the reward function to obtain an optimal control strategy, and replace the preset control strategy with the optimal control strategy.

22. The device according to claim 21, characterized in that The formula of the reward function is: Among them, R is the reward function, w acc is the inspection accuracy weight coefficient, R acc is the inspection accuracy factor, w eff is the inspection efficiency weight coefficient, R eff is the inspection efficiency factor, w sta is the UAV stability weight coefficient, R sta is the stability factor of the UAV, w sec is the safety weight coefficient of the drone, R sec This is the safety factor of the drone.

23. The device according to claim 21, characterized in that The reinforcement learning module comprises: The reinforcement learning submodule is used to perform reinforcement learning on the UAV in a parallel system based on the reward function, the UAV state space parameters and the UAV action space parameters to obtain the optimal control strategy; wherein the UAV state space parameters at least include UAV flight state parameters, UAV control deviation parameters and UAV observation identification result parameters, and the UAV action space parameters at least include UAV controller parameters, UAV allocation law parameters and UAV estimator parameters.

24. The device according to claim 23, characterized in that The reinforcement learning submodule comprises: A reinforcement learning unit is used to perform reinforcement learning on the drone in a parallel system according to the following mathematical expression based on the reward function, the drone state space parameters and the drone action space parameters to obtain the optimal control strategy: Among them, π * (s) is the optimal control strategy, R is the reward function, is the discount factor, a is the UAV action space parameter, s is the UAV state space parameter, and t is the time.

25. The device according to claim 14, characterized in that The inspection task is an inspection task for power facilities, and the sensors of the UAV include at least one or more of an optical camera, an infrared thermal imager, a wind speed and direction sensor, and an electromagnetic interference sensor.

26. The device according to claim 14, characterized in that The drone surrounding environment data includes at least one or more of appearance image data of power facilities, temperature distribution data of power facilities, meteorological condition data of the drone surrounding environment, and electromagnetic environment data of the drone surrounding environment.

27. An electronic device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the drone control method according to any one of claims 1 to 13 is implemented.

28. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the drone control method according to any one of claims 1 to 13 is implemented.

29. A computer program product, comprising a computer program, which, when executed by a processor, implements the drone control method according to any one of claims 1 to 13.

Images