UAV tracking method, device, UAV and storage medium

Abstract

The present invention discloses a method, device, drone and storage medium for tracking a drone, the method comprising: acquiring an image of a target, identifying the target from the image using a neural network recognition model, and determining the category and position of the target; tracking the identified target using a target tracking model, and continuously acquiring the azimuth information of the target; based on the azimuth information, calculating the error between the actual distance value between the drone and the target and the preset ideal distance, reducing the error using an anti-disturbance control algorithm, and determining the flight parameters of the drone for tracking the target. The above scheme is used for drone tracking of targets in a satellite-denied environment, has low application cost, reduces tracking errors, improves the estimation capability and anti-disturbance capability of the anti-disturbance control algorithm, and thus improves the accuracy of target tracking.

Description

Technical Field

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

[0002] At present, the position of drones in a denied environment is usually determined by vision, and the position of drones is settled through feature recognition, map matching, etc. Based on this method, it is suitable for conventional take-off, hovering, landing, heading flight and other tasks. However, it is difficult to meet the requirements of tracking.

[0003] ADRC absorbs the technical essence of PID control "eliminating errors based on errors" and the achievements of modern control theory. Its core idea is to use the simple integral series relationship between system output and control input as the standard type, and regard the part of the system dynamics that is different from the standard type as the "total disturbance". According to the control input and system output, the extended state observer reconstructs the system output, the derivative of the output and the total disturbance, and then obtains the control signal to estimate and eliminate the system's "total disturbance" in real time. As the core of ADRC, the limited estimation ability of the extended state observer is the key to limiting the improvement of ADRC performance. Since the fixed observer parameters are difficult to adapt to time-varying signals, the performance of ADRC is limited. Summary of the invention

[0004] In view of the above problems, the present invention is proposed to provide a drone tracking method, device, drone and storage medium that overcome the above problems or at least partially solve the above problems.

[0005] According to one aspect of the present invention, a method for tracking a drone is provided, the method comprising:

[0006] Acquire an image of a target, identify the target from the image using a neural network recognition model, and determine the type and location of the target;

[0007] Tracking the identified target using a target tracking model to continuously obtain the position information of the target;

[0008] Based on the position information, the error between the actual distance value between the UAV and the target and the preset ideal distance is calculated, the error is reduced by using an anti-disturbance control algorithm, and the flight parameters of the UAV for tracking the target are determined.

[0009] In some embodiments, acquiring an image of a target includes:

[0010] Acquiring an image of the target using a camera in an optical pod, wherein the target includes a ground moving target and an air flying target;

[0011] Using an auto-disturbance rejection control algorithm to reduce the error and determine the flight parameters of the drone for tracking the target includes:

[0012] The extended state observer in the active disturbance rejection control algorithm is used to reconstruct and control the flight parameters of the UAV, and the error is estimated in real time and eliminated or reduced.

[0013] In some embodiments, using the extended state observer in the active disturbance rejection control algorithm to reconstruct the flight parameters of the drone to estimate and eliminate or reduce the error in real time includes:

[0014] A P-type iterative learning control is introduced into the extended state observer, a proportional factor is determined according to the estimated error, and the adjustable parameters of the extended state observer are optimized using the proportional factor.

[0015] In some implementations, introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing the adjustable parameters of the extended state observer using the proportional factor includes:

[0016] Constructing a learning law for the adjustable parameter based on P-type iterative learning control;

[0017] Based on the learning law, expressions of the control law and the extended state observer are constructed.

[0018] In some implementations, the formula for constructing the learning law of the adjustable parameter based on P-type iterative learning control is:

[0019]

[0020] Where L is the learning gain, e k (t) is the deviation signal of the kth iteration, and b0 and W0 are the adjustable parameters of the extended state observer.

[0021] In some implementations, constructing the learning law of the adjustable parameter based on P-type iterative learning control further includes:

[0022] A mapping mechanism is introduced into the formula of the learning law, wherein the mapping mechanism includes the product of a saturation function and the formula of the learning law.

[0023] In some embodiments, the method further comprises:

[0024] The dynamic model of the UAV is obtained using the Newton-Euler equation;

[0025] Introducing two virtual control quantities as control inputs of the horizontal position model into the dynamic model respectively, so as to decouple the dynamic model;

[0026] Converting the decoupled dynamic model into a second-order single-input single-output model;

[0027] The convergence of the extended state observer introduced with P-type iterative learning control and the closed-loop stability of the active disturbance rejection control algorithm are solved and verified according to the single-input single-output model.

[0028] According to another aspect of the present invention, a drone tracking device is provided, the device comprising:

[0029] A recognition module, adapted to acquire an image of a target, identify the target from the image using a neural network recognition model, and determine the type and location of the target;

[0030] A tracking module, adapted to track the identified target using a target tracking model, and continuously obtain the position information of the target;

[0031] The control module is adapted to calculate the error between the actual distance value between the UAV and the target and the preset ideal distance based on the orientation information, reduce the error by using an anti-disturbance control algorithm, and determine the flight parameters of the UAV for tracking the target.

[0032] According to another aspect of the present invention, a rotary wing drone is provided, comprising: a processor and a memory arranged to store computer executable instructions, wherein when the executable instructions are executed, the processor executes the drone tracking method according to any one of the above embodiments.

[0033] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, they implement any of the above-mentioned drone tracking methods.

[0034] From the above, it can be seen that the drone tracking method disclosed in the present invention can be used for drone tracking of targets in a satellite denial environment. It has low application cost, reduces tracking errors, improves the estimation ability and anti-disturbance ability of the self-anti-disturbance control algorithm, and thus improves the accuracy of target tracking.

[0035] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0037] Figure 1 A schematic diagram of a flow chart of a method for tracking a drone according to an embodiment of the present invention is shown;

[0038] Figure 2 A control structure block diagram of a UAV closed-loop control system according to an embodiment of the present invention is shown;

[0039] Figure 3 It shows that the bottom-level flight control of a quad-rotor drone according to an embodiment of the present invention adopts a classic inner and outer loop control framework;

[0040] Figure 4 A schematic diagram of the structure of a drone tracking device according to an embodiment of the present invention is shown;

[0041] Figure 5 A schematic structural diagram of a rotary-wing UAV according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0042] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0043] Figure 1 The following is a flow chart of a method for tracking a drone according to an embodiment of the present invention. The method can be implemented by a program in a flight control system of the drone. The method includes the following steps:

[0044] Step S110, acquiring an image of a target, identifying the target from the image using a neural network recognition model, and determining the type and position of the target;

[0045] Step S120, tracking the identified target using a target tracking model, and continuously acquiring the position information of the target;

[0046] Step S130, based on the orientation information, calculating the error between the actual distance value between the UAV and the target and the preset ideal distance, reducing the error using an anti-disturbance control algorithm, and determining the flight parameters of the UAV for tracking the target.

[0047] Among them, the ADRC algorithm mainly consists of the following two parts: PD controller: Its main function is to linearly combine the expected value and the system states estimated by the extended state observer through the PD controller to obtain the virtual control quantity. Extended State Observer (ESO): As the core component of ADRC, ESO is responsible for real-time estimation of the system state and the comprehensive disturbance caused by the internal uncertainty (internal disturbance) and external disturbance (external disturbance) of the system. By designing an extended state quantity to track the influence of the unknown part of the model and the external unknown disturbance, ESO can provide feedback linearization for the controller and transform the control object into an integral series type, thereby realizing precise control of the system state. This step combines the input signal of the system with the output of the observer to form a control strategy to achieve effective control of the system.

[0048] The embodiment of the present invention introduces an auto-disturbance rejection control algorithm into the tracking of a target by a UAV, has low application cost, reduces tracking error, improves the estimation capability and anti-disturbance capability of the auto-disturbance rejection control algorithm, and thus improves the accuracy of target tracking.

[0049] Furthermore, the embodiment of the present invention can improve the estimation capability of the observer and reduce the estimation error by proposing an extended state observer design method and an anti-disturbance control system based on iterative learning.

[0050] Thus, in some embodiments, acquiring an image of a target includes:

[0051] Acquiring an image of the target using a camera in an optical pod, wherein the target includes a ground moving target and an air flying target;

[0052] Using an auto-disturbance rejection control algorithm to reduce the error and determine the flight parameters of the drone for tracking the target includes:

[0053] The extended state observer in the active disturbance rejection control algorithm is used to reconstruct and control the flight parameters of the UAV, and the error is estimated in real time and eliminated or reduced.

[0054] In some embodiments, using the extended state observer in the ADRC algorithm to reconstruct and control the flight parameters of the drone, and estimating and reducing the error in real time includes:

[0055] A P-type iterative learning control is introduced into the extended state observer, a proportional factor is determined according to the estimated error, and the adjustable parameters of the extended state observer are optimized using the proportional factor.

[0056] Iterative learning control (ILC) can simulate human brain learning and realize self-regulation function. It does not rely on the precise mathematical model of the system. It is a typical intelligent control method. The essence of this design is to learn and optimize the adjustable parameters of the extended state observer by introducing an iterative learning mechanism, and provide a proportional factor according to the changing estimated error. Therefore, the bandwidth of the extended state observer can be learned and adjusted according to the estimated error, and it changes in real time. This is of great significance to actual engineering control.

[0057] The P-type iterative learning control can be written as

[0058]

[0059] Where L is the learning gain, e k (t) is the deviation signal of the kth iteration, and are the control signals for the kth and k+1th iterations respectively.

[0060] In some embodiments, introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing the adjustable parameters of the extended state observer using the proportional factor includes:

[0061] Constructing a learning law for the adjustable parameter based on P-type iterative learning control;

[0062] Based on the learning law, expressions of the control law and the extended state observer are constructed.

[0063] In some optional embodiments, combined with Figure 2 As shown, the formula for constructing the learning law of the adjustable parameters based on P-type iterative learning control is:

[0064]

[0065] Where L is the learning gain, e k (t) is the deviation signal of the kth iteration, and b0 and W0 are the adjustable parameters of the extended state observer.

[0066] By introducing a The learning law of the adjustable parameters of the extended state observer can be described as

[0067]

[0068] On this basis, an iterative learning observer is constructed to build the control law and the formula of the extended state observer.

[0069] In some embodiments, constructing the learning law of the adjustable parameter based on P-type iterative learning control further includes:

[0070] A mapping mechanism is introduced into the formula of the learning law, wherein the mapping mechanism includes the product of a saturation function and the formula of the learning law.

[0071] For the learning law of the above adjustable parameters, as long as the appropriate learning gain L is selected, the parameters and estimation errors of the extended state observer will generally converge within a bound. In order to ensure a strict formula, a mapping mechanism can be introduced into the algorithm:

[0072]

[0073] The bounds of the saturation function sat() can be arbitrarily large, so that the actual generated parameters will not exceed it. Obviously, with the above mapping, it can be determined that the parameters are always bounded. After a limited number of iterations, a smaller estimation error can be obtained.

[0074] In some embodiments, the method further comprises:

[0075] The dynamic model of the UAV is obtained using the Newton-Euler equation;

[0076] Introducing two virtual control quantities as control inputs of the horizontal position model into the dynamic model respectively, so as to decouple the dynamic model;

[0077] Converting the decoupled dynamic model into a second-order single-input single-output model;

[0078] The convergence of the extended state observer that introduces P-type iterative learning control and the closed-loop stability of the active disturbance rejection control algorithm are solved and verified based on the single-input single-output model.

[0079] In the following detailed embodiment, a specific implementation of the above embodiment is introduced, so as to further illustrate the above embodiment of the present invention.

[0080] 1) Attitude-position inner and outer loop controller design model

[0081] In order to facilitate the controller design, the quadcopter UAV model needs to be simplified and processed. The quadcopter UAV dynamics model obtained by the Newton-Euler equation is as follows:

[0082]

[0083] where d i(i=1, 2, …, 6) represents the unknown modeling dynamics, the impact of unknown disturbances on each subsystem and the uncertainty. x , I y , I z is the moment of inertia of the three axes. From the dynamic model (1), we can see that the control input of the quadrotor drone system is only four (total force F and three-axis torque x t ,y t , z t ), but has six outputs (x, y, z and f,q,y,), which is a typical under-driven and strongly coupled system. To this end, two virtual control quantities are introduced, U x U y As the control input of the horizontal position model,

[0084]

[0085] Therefore, the quadrotor drone system is decoupled into the following six independent subsystems:

[0086]

[0087] According to system (3), the control objectives are designed separately, U x , U y , F and x t ,y t , z t , so that the system tracks the desired position x r ,y r , z r and attitude angle f r q r y r . Where x r y r z r and r It can be set manually, but the desired roll angle and pitch angle need to be set by the virtual control quantity U x U y According to formula (2), we can get the expected f r q r , specifically expressed as

[0088]

[0089] Observe system (3), the decoupled subsystems of the quadrotor drone can be transformed into the following second-order single-input single-output system

[0090]

[0091] Where u is the system input, such as U in formula (3)x , U y , F and x t ,y t , z t As shown in the figure; y is the system output, namely, 3 position information and 3 attitude angle information; b is the control input coefficient related to the system structure coefficient, which needs to be known before designing the controller; for example, for the roll channel b= 1 / I x , the horizontal position channel can be approximated as b=-g (because the total tension F = mg when hovering); f represents the "unknown total disturbance" or "unknown model function" of each subsystem, which includes known modeling dynamics, unknown modeling dynamics, unknown disturbances and uncertainties, such as the roll channel .

[0092] Set the reference command to y r The control goal is to design the control input u so that the system output y tracks the reference instruction y r , Right now , and at the same time can suppress the influence of unknown disturbances or unknown model functions f.

[0093] 2) Iterative Learning Observer

[0094] Iterative learning control (ILC) can simulate human brain learning and realize self-regulation function. It does not rely on the precise mathematical model of the system. It is a typical intelligent control method. The essence of this design is to learn and optimize the adjustable parameters of the extended state observer by introducing an iterative learning mechanism, and provide a proportional factor according to the changing estimated error. Therefore, the bandwidth of the extended state observer can be learned and adjusted according to the estimated error, and it changes in real time. This is of great significance to actual engineering control.

[0095] P-type iterative learning control is

[0096]

[0097] Where L is the learning gain, e k (t) is the deviation signal of the kth iteration, η k (t) and η k+1 (t) are the control signals for the kth and k+1th iterations respectively.

[0098] If the parameters of the extended state observer can be adjusted by the P-type iterative learning method, then a disturbance rejection control based on P-type iterative learning can be designed. The learning law of the adjustable parameters of the extended state observer can be described as

[0099]

[0100] On this basis, an iterative learning observer is constructed. Its control law and extended state observer are (8) and (9) respectively.

[0101] The control law u is designed as

[0102]

[0103] where k p k b and b0 are the adjustable parameters of the controller, z1z2 and z3 are the outputs of the extended state observer, which are used to estimate the system output, the derivative of the output and the total disturbance of the system respectively.

[0104] The extended state observer is designed as

[0105]

[0106] Where y is the actual output value of the system, k p k b and b0 are the adjustable parameters of the controller, and b1 b2 b3 are the adjustable parameters of the observer.

[0107] The parameters b0 and w0 of the extended state observer are determined by equation (7). Therefore, b0 and w0 vary with the estimation error. The basic structure of the learning anti-disturbance control closed-loop system is as follows: Figure 2 shown.

[0108] For the learning law (7) of the ESO parameters, as long as the appropriate learning gain L is selected, the parameters and estimation errors of the extended state observer will generally converge within a bound. In order to ensure a strict formula, a mapping mechanism can be introduced into the algorithm

[0109]

[0110] The bound of the saturation function sat(×) can be arbitrarily large, so that the actual generated parameters will not exceed it. Obviously, with this mapping, it can be determined that the parameters are always bounded. After a finite number of iterations, a smaller estimation error can be obtained.

[0111] 3) Calculation and stability analysis of iterative learning observer

[0112] The second-order controlled object is

[0113]

[0114] Where w represents the external disturbance, b0 is a non-zero constant, u is the control input, and y is the system output. Represents unknown nonlinear time-varying dynamics that can be estimated and compensated.

[0115] For convenience, , and assuming . Introducing an "expanded state" 3 x , then let 3 x = f, then we have

[0116]

[0117] in, .

[0118] In order to analyze the convergence on the time axis, the convergence of ESO and the closed-loop stability of ADRC are analyzed separately.

[0119] i) Convergence of the Extended State Observer

[0120] When the dynamic model of the controlled object is known, let the observer The estimated values ​​are Then, the ESO of formula (11) can be designed as

[0121]

[0122] in, is the ESO observation vector, is the constant gain of ESO, .

[0123] Subtracting equation (12) from equation (11), we get

[0124]

[0125] Then there is

[0126]

[0127] In order to make the estimation error approach zero, the matrix (A−L) must be Hurwitz. Let

[0128]

[0129] Where w0 is the observer bandwidth.

[0130]

[0131] Substitute equation (16) into equation (13), and let

[0132] Then there is

[0133]

[0134] in .

[0135] In fact, f is unknown and cannot be obtained. Therefore, it cannot be used in ESO. In equation (12), let =0, then equation (13) becomes

[0136]

[0137] make , then (19) becomes

[0138]

[0139] According to Lemma 1: Assume is bounded, that is, there exists a positive number M1 such that , then the estimation error of ESO is bounded, that is, there exists a positive number M2 such that Established.

[0140] ii) Closed-loop stability of ADRC

[0141] The control law of active disturbance rejection control can be written as

[0142]

[0143] Among them, k p and k d are all greater than 0 and are adjustable parameters of the controller. is x i The estimated value of y r reference input, b0 is the adjustable control gain.

[0144] Substituting equation (21) into equation (10), we have

[0145]

[0146] make So,

[0147]

[0148] in is the estimation error, is the following error.

[0149] make , then formula (23) is written as:

[0150]

[0151] Lemma 2: Assume h(X,w) is bounded, that is, there exists a positive number M1 such that , then there exists a linear extended state observer and controller parameters , so that the tracking error of the closed-loop system (4-19) is bounded. Thus, for bounded input y r , the output of the closed-loop system (24) is bounded, that is, the closed-loop system (24) is BIBO stable.

[0152] According to Lemma 1 and Lemma 2, if the ESO parameters are selected appropriately, the estimation error of the ESO is bounded, and thus the tracking error of the closed-loop system is bounded. Therefore, after several iterations, if the estimation error is bounded, the closed-loop system is BIBO stable.

[0153] Figure 3 It is shown that the bottom-level flight control of the quad-rotor drone according to an embodiment of the present invention adopts the classic inner and outer loop control framework; it mainly includes three parts: attitude-position inner and outer loop control, the conversion of force and torque to speed and then to throttle command, and control distribution. The input of the bottom-level flight controller is the given reference command and the posture information obtained by the sensor measurement module, and the output is the throttle command expected by the four motors. The sensor measurement module mainly includes an inertial measurement unit (IMU), a laser rangefinder, an ultrasonic rangefinder, an optical flow, a global positioning system (GPS) and an optical positioning system, etc., and different selection combinations can be made according to the mission requirements and application environment.

[0154] The above embodiments can achieve the following technical effects:

[0155] Iterative learning control (ILC) generates a control signal through the deviation between the desired trajectory and the system output to correct the system trajectory and reduce the following deviation, so that the system output value follows the expected value.

[0156] Combination Figure 4 As shown, the present invention also protects a drone tracking device 400, and the drone tracking device 400 includes:

[0157] The recognition module 410 is adapted to obtain an image of a target, identify the target from the image using a neural network recognition model, and determine the type and location of the target;

[0158] A tracking module 420, adapted to track the identified target using a target tracking model, and continuously obtain the position information of the target;

[0159] The control module 430 is adapted to calculate the error between the actual distance value between the UAV and the target and the preset ideal distance based on the orientation information, reduce the error by using an anti-disturbance control algorithm, and determine the flight parameters of the UAV for tracking the target.

[0160] According to this embodiment, the self-disturbance rejection control algorithm is introduced into the tracking of the target by the UAV, with low application cost, reduced tracking error, improved estimation ability and anti-disturbance ability of the self-disturbance rejection control algorithm, thereby improving the accuracy of target tracking.

[0161] In some embodiments, acquiring an image of a target in the recognition module 410 includes:

[0162] Acquiring an image of the target using a camera in an optical pod, wherein the target includes a ground moving target and an air flying target;

[0163] The control module 430 uses an anti-disturbance control algorithm to reduce the error and determine the flight parameters used by the drone to track the target, including:

[0164] The extended state observer in the active disturbance rejection control algorithm is used to reconstruct and control the flight parameters of the UAV, and the error is estimated and reduced in real time.

[0165] In some embodiments, using the extended state observer in the ADRC algorithm to reconstruct and control the flight parameters of the drone, and estimating and reducing the error in real time includes:

[0166] A P-type iterative learning control is introduced into the extended state observer, a proportional factor is determined according to the estimated error, and the adjustable parameters of the extended state observer are optimized using the proportional factor.

[0167] In some embodiments, introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing the adjustable parameters of the extended state observer using the proportional factor includes:

[0168] Constructing a learning law for the adjustable parameter based on P-type iterative learning control;

[0169] Based on the learning law, expressions of the control law and the extended state observer are constructed.

[0170] In some embodiments, the formula for constructing the learning law of the adjustable parameter based on P-type iterative learning control is:

[0171]

[0172] Where L is the learning gain, e k(t) is the deviation signal of the kth iteration, and b0 and W0 are the adjustable parameters of the extended state observer.

[0173] In some embodiments, constructing the learning law of the adjustable parameter based on P-type iterative learning control further includes:

[0174] A mapping mechanism is introduced into the formula of the learning law, wherein the mapping mechanism includes the product of a saturation function and the formula of the learning law.

[0175] In some embodiments, the device is further adapted to:

[0176] The dynamic model of the UAV is obtained using the Newton-Euler equation;

[0177] Introducing two virtual control quantities as control inputs of the horizontal position model into the dynamic model respectively, so as to decouple the dynamic model;

[0178] Converting the decoupled dynamic model into a second-order single-input single-output model;

[0179] The convergence of the extended state observer that introduces P-type iterative learning control and the closed-loop stability of the active disturbance rejection control algorithm are solved and verified based on the single-input single-output model.

[0180] It should be noted that the specific implementation of the above-mentioned device embodiments can be carried out with reference to the specific implementation of the above-mentioned corresponding method embodiments, which will not be repeated here.

[0181] It should be noted that:

[0182] The algorithm and display provided herein are not inherently related to any particular computer, virtual device or other equipment. Various general-purpose devices can also be used together with the teachings based on this. According to the above description, it is obvious to construct the structure required for this type of device. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific language is for the purpose of disclosing the best mode of the present invention.

[0183] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0184] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this method of disclosure should not be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim.

[0185] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and further may be divided into a plurality of submodules or subunits or subcomponents. All features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device so disclosed may be combined in any combination, except that at least some of such features and / or processes or units are mutually exclusive. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature that provides the same, equivalent or similar purpose.

[0186] Furthermore, those skilled in the art will appreciate that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is meant to be within the scope of the present invention and to form different embodiments.

[0187] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components of the drone tracking device according to an embodiment of the present invention. The present invention may also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program for implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0188] An embodiment of the present invention provides a non-volatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the computer executable instruction can execute the above-mentioned drone tracking method in any of the above-mentioned method embodiments.

[0189] Figure 5A schematic structural diagram of an embodiment of a rotary-wing UAV of the present invention is shown. The specific embodiment of the present invention does not limit the specific structure of the rotary-wing UAV.

[0190] like Figure 5 As shown, the rotary wing UAV (including the flight control system) may include: a processor (processor) 502, a communications interface (Communications Interface) 504, a memory (memory) 506, and a communication bus 508.

[0191] The processor 502, the communication interface 504, and the memory 506 communicate with each other via the communication bus 508. The communication interface 504 is used to communicate with other devices such as a client or other server network elements. The processor 502 is used to execute the program 510, which can specifically execute the relevant steps in the above-mentioned drone tracking method embodiment for the rotary wing drone.

[0192] Specifically, the program 510 may include program codes, which include computer operation instructions.

[0193] The processor 502 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the rotary-wing drone may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0194] The memory 506 is used to store the program 510. The memory 506 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0195] Program 510 can be specifically used to enable processor 502 to perform operations corresponding to the above-mentioned drone tracking method embodiment.

[0196] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc., does not indicate any order. These words may be interpreted as names.

Claims

1. A drone tracking method, characterized in that: The method comprises: Acquire an image of a target, identify the target from the image using a neural network recognition model, and determine the type and location of the target; Tracking the identified target using a target tracking model to continuously obtain the position information of the target; Based on the position information, calculating the error between the actual distance between the UAV and the target and the preset ideal distance, reducing the error by using an auto-disturbance rejection control algorithm, and determining the flight parameters of the UAV for tracking the target; The images of the target are acquired by: Acquiring an image of the target using a camera in an optical pod, wherein the target includes a ground moving target and an air flying target; Using an auto-disturbance rejection control algorithm to reduce the error and determine the flight parameters of the drone for tracking the target includes: Reconstruct and control the flight parameters of the UAV using the extended state observer in the active disturbance rejection control algorithm to estimate and reduce the error in real time; Wherein, using the extended state observer in the active disturbance rejection control algorithm to reconstruct and control the flight parameters of the UAV, and estimating and reducing the error in real time includes: Introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing an adjustable parameter of the extended state observer using the proportional factor; In which, introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing the adjustable parameters of the extended state observer using the proportional factor include: Constructing a learning law for the adjustable parameter based on P-type iterative learning control; Based on the learning law, constructing the control law and the expression of the extended state observer; The formula for constructing the learning law of the adjustable parameters based on P-type iterative learning control is: ， Where L is the learning gain, e k (t) is the deviation signal of the kth iteration, b0 and W0 are the adjustable parameters of the extended state observer, b 0k (t) and b 0（k+1） (t) are the values ​​of b0 at the kth iteration and the k+1th iteration, respectively, and w 0k (t) and w 0（k+1） (t) are the values ​​of w0 at the kth iteration and the k+1th iteration respectively, t is the time variable, and k is a positive integer.

2. The method according to claim 1, characterized in that The learning law for constructing the adjustable parameters based on P-type iterative learning control also includes: A mapping mechanism is introduced into the formula of the learning law, and the mapping mechanism is the product of a saturation function and the formula of the learning law.

3. The method according to claim 2, characterized in that The method further comprises: The dynamic model of the UAV is obtained using the Newton-Euler equation; Introducing two virtual control quantities as control inputs of the horizontal position model into the dynamic model respectively, so as to decouple the dynamic model; Converting the decoupled dynamic model into a second-order single-input single-output model; The convergence of the extended state observer introduced with P-type iterative learning control and the closed-loop stability of the active disturbance rejection control algorithm are solved and verified according to the single-input single-output model.

4. A drone tracking device, characterized in that: The device comprises: A recognition module, adapted to acquire an image of a target, identify the target from the image using a neural network recognition model, and determine the type and location of the target; A tracking module, adapted to track the identified target using a target tracking model, and continuously obtain the position information of the target; A control module, adapted to calculate an error between an actual distance value between the UAV and the target and a preset ideal distance based on the position information, reduce the error by using an auto-disturbance rejection control algorithm, and determine a flight parameter for the UAV to track the target; The images of the target acquired in the recognition module include: Acquiring an image of the target using a camera in an optical pod, wherein the target includes a ground moving target and an air flying target; The control module uses an anti-disturbance control algorithm to reduce the error and determine the flight parameters of the drone for tracking the target, including: Reconstruct and control the flight parameters of the UAV using the extended state observer in the active disturbance rejection control algorithm to estimate and reduce the error in real time; Wherein, using the extended state observer in the active disturbance rejection control algorithm to reconstruct and control the flight parameters of the UAV, and estimating and reducing the error in real time specifically includes: Introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing an adjustable parameter of the extended state observer using the proportional factor; In which, introducing P-type iterative learning control into the extended state observer, determining a proportional factor according to the estimated error, and optimizing the adjustable parameters of the extended state observer using the proportional factor include: Constructing a learning law for the adjustable parameter based on P-type iterative learning control; Based on the learning law, constructing the control law and the expression of the extended state observer; The formula for constructing the learning law of the adjustable parameters based on P-type iterative learning control is: ， Where L is the learning gain, e k (t) is the deviation signal of the kth iteration, b0 and w0 are the adjustable parameters of the extended state observer, b 0k (t) and b 0（k+1） (t) are the values ​​of b0 at the kth iteration and the k+1th iteration, respectively, and w 0k (t) and w 0（k+1） (t) are the values ​​of w0 at the kth iteration and the k+1th iteration respectively, t is the time variable, and k is a positive integer.

5. A rotary wing UAV, characterized in that: The invention comprises a processor and a memory arranged to store computer executable instructions, wherein when the executable instructions are executed, the processor performs the drone tracking method according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores one or more programs, which, when executed by a processor, implement the drone tracking method according to any one of claims 1 to 3.

Images