An unmanned aerial vehicle adaptive rigid force response control method and device

Abstract

This application provides an adaptive rigid force response control method and apparatus for a launch vehicle. The method involves: acquiring the real-time state of the launch vehicle in the current calculation cycle; generating basic control commands corresponding to the desired attitude information based on real-time attitude information and a basic controller; adaptively generating feedforward compensation control commands corresponding to the deterministic disturbance component in the rigid force based on deterministic disturbance force, deterministic disturbance torque, and a deterministic disturbance feedforward controller; adaptively predicting feedback compensation control commands corresponding to the uncertain disturbance component in the rigid force in the current calculation cycle based on the flight state of the launch vehicle after receiving integrated control commands in the previous calculation cycle and the feedforward compensation control commands in the current calculation cycle; and fusing the basic control commands, feedforward compensation control commands, and feedback compensation control commands to generate an adaptive integrated control command for the launch vehicle. This adaptive response to the rigid force of the launch vehicle improves flight stability and anti-interference capability.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) control technology, and in particular to an adaptive rigid force response control method and device for a transport UAV. Background Technology

[0002] With the development of drone control technology, transport drones have become one of the most popular drone types. When a transport drone performs a mission, the reaction force exerted on the drone by the cargo it carries during contact and flight is non-fixed. Examples include swaying liquids and suspended cargo. In other words, the cargo can be considered a non-fixed load on the transport drone, or a rigid contact with the environment. Due to the uncertainty of the cargo's position and mass, the drone's fuselage is typically subjected to complex external forces and torques. These forces are dynamic, nonlinear, and uncertain, severely affecting the drone's flight stability and control accuracy, and representing a key technological bottleneck restricting its application in complex scenarios.

[0003] Existing technologies mainly address this problem through two approaches: one is a model-based control method, which has strict requirements on load type and poor versatility; the other is a disturbance observer (DOB)-based method, which treats all external forces as a single unknown disturbance and indirectly estimates and compensates for it through the state feedback of the UAV itself. However, the disturbance observer-based method has inherent drawbacks: 1) Estimation time lag: The disturbance observer is essentially a low-pass filter with phase lag. When the rigid force changes rapidly, its estimate cannot keep up in real time, resulting in poor compensation effect; 2) Performance contradiction: There is an inherent contradiction in the selection of the disturbance observer bandwidth. High bandwidth can improve tracking speed but amplify sensor noise, while low bandwidth has the opposite effect, making it difficult to achieve both; 3) Information loss: Processing all disturbances in a "one-size-fits-all" manner loses the structured information of the disturbance source, making it impossible to achieve optimal and refined compensation control. Summary of the Invention

[0004] In view of this, this application provides an adaptive rigid force response control method and apparatus for unmanned aerial vehicles.

[0005] Specifically, this application is achieved through the following technical solution: The first aspect of this application provides an adaptive rigid force response control method for a carrier unmanned aerial vehicle (UAV), the method comprising:

[0006] Obtain the real-time state of the carrier drone in the current calculation cycle. The real-time state includes at least the real-time attitude information of the carrier drone and the deterministic perturbation force and deterministic perturbation torque in the rigid force.

[0007] Based on the real-time attitude information and the basic controller, basic control commands corresponding to the desired attitude information are generated.

[0008] Based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller, the feedforward compensation control command corresponding to the deterministic disturbance component in the rigid force is adaptively generated.

[0009] Based on the flight state of the UAV after receiving the integrated control command in the previous calculation cycle and the feedforward compensation control command in the current calculation cycle, the feedback compensation control command corresponding to the uncertainty disturbance component in the rigid force in the current calculation cycle is adaptively predicted. The prediction is based at least on the uncertainty disturbance force and the uncertainty disturbance torque.

[0010] By integrating the basic control commands, the feedforward compensation control commands, and the feedback compensation control commands, an adaptive integrated control command for the carrier UAV is generated.

[0011] A second aspect of this application provides an adaptive rigid force response control device for a carrier unmanned aerial vehicle, the device comprising:

[0012] The acquisition module is used to acquire the real-time state of the carrier drone in the current calculation cycle. The real-time state includes at least the real-time attitude information of the carrier drone and the deterministic perturbation force and deterministic perturbation torque in the rigid force.

[0013] The basic control module is used to generate basic control commands corresponding to the desired attitude information based on the real-time attitude information and the basic controller.

[0014] The feedforward compensation module is used to adaptively generate feedforward compensation control commands corresponding to the deterministic disturbance component in the rigid force based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller.

[0015] The feedback compensation module is used to adaptively predict the feedback compensation control command corresponding to the uncertainty disturbance component in the rigid force in the current calculation cycle based on the flight state of the UAV after receiving the integrated control command in the previous calculation cycle and the feedforward compensation control command in the current calculation cycle. The prediction is based at least on the uncertainty disturbance force and the uncertainty disturbance torque.

[0016] The control module is used to integrate the basic control commands, the feedforward compensation control commands, and the feedback compensation control commands to generate adaptive integrated control commands for the carrier UAV.

[0017] The adaptive rigid force response control method and apparatus for transport drones provided in this application, in order to achieve accurate response control of external forces on the transport drone, considers the complex changes in the rigid force of the transport drone. It conceptually decomposes the rigid force on the transport drone into deterministic and indeterminate disturbances, and simultaneously generates corresponding control commands for different items based on force and torque. This achieves adaptive and accurate generation of control commands for each item, improving the adaptive response control capability for complex scenarios. When cargo is loaded on top of the transport drone, since the center of mass of the cargo cannot be guaranteed to coincide with the center of the transport platform, the cargo's gravity and other contact forces during transport will come from various directions. For these two types of disturbances, compared to the traditional method of quantifying the disturbances solely through weight, this application introduces force and torque for a more refined evaluation, adding orientation in addition to numerical values. This fully addresses control scenarios involving the special relative positional relationships of cargo at non-central locations such as the edge of the transport drone, improving the accuracy of the transport drone control. Attached Figure Description

[0018] Figure 1 A flowchart of the adaptive rigid force response control method for a launch vehicle provided in this application;

[0019] Figure 2 This is a schematic diagram of the adaptive rigid force response control device for unmanned aerial vehicles provided in this application. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0021] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0022] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0023] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0024] Figure 1 A flowchart illustrating the adaptive rigid force response control method for the unmanned aerial vehicle provided in this application. Please refer to... Figure 1 The method provided in this embodiment may include:

[0025] S101. Obtain the real-time state of the carrier drone in the current calculation cycle. The real-time state includes at least the real-time attitude information of the carrier drone and the deterministic perturbation force and deterministic perturbation moment in the rigid force.

[0026] The method provided by this invention is used for real-time flight attitude control based on the force conditions of the transport drone. As a specific implementation, the top of the transport drone includes a transport platform for placing items to be transported to a designated destination. Since the goods may be placed at different positions on the transport platform under different transfer conditions, such as at the edge of the transport platform, the external interference force on the transport drone is no longer exactly the same as the mass of the goods. In order to improve the accurate response to interference forces, the method provided by this invention installs multiple force measurement sensors at the bottom of the transport platform to measure the real-time force conditions on the transport platform, so as to calculate the force and its torque.

[0027] During the operation of a UAV, the disturbance forces it experiences can be divided into two aspects: First, there is the measurable disturbance component, which is the deterministic disturbance within the rigid force. This component is the part that the sensor described in this application can directly and reliably capture. The deterministic disturbance includes at least the projections of the inertial force of the payload, gravity, and contact force with the environment onto the sensor's detection axis. The detection axes of each sensor are not exactly the same. Second, there is the unmeasurable disturbance component, which is the uncertain disturbance. This component is the part that the sensor cannot capture or that is not covered when the UAV model is established. The uncertain disturbance includes at least: 1) forces acting on non-detection axes of the sensor; 2) complex aerodynamic disturbances, such as ground effects and changes in downwash airflow; 3) internal friction of the system not considered in the UAV model. Preferably, the deterministic disturbance is much larger than the uncertain disturbance, the amplitude of the deterministic disturbance component is greater than the amplitude of the uncertain disturbance, the frequency of the deterministic disturbance component is less than the frequency of the uncertain disturbance, and the uncertain disturbance has random and unstructured characteristics.

[0028] The method provided in this application is used for real-time attitude control of a transport drone during cargo delivery, enabling the drone to maintain a stable attitude. At each time step, real-time attitude and force information of the transport drone are acquired and used as control inputs for a real-time controller for real-time adjustment. The deterministic disturbance component in the rigid force detected by the sensors has a large amplitude and low frequency. The frequency of the drone's attitude control is higher than that of the deterministic disturbance component, improving the drone's response efficiency to sudden, large-amplitude disturbances. The instruction for each time step includes three parts: a basic control instruction generated based on the real-time difference between the current attitude and the ideal attitude; an active disturbance compensation control instruction corresponding to the deterministic disturbance component detected by the sensors, i.e., a feedforward compensation control instruction; and a control instruction that performs hysteresis-based feedback compensation for the uncertain disturbance estimated from the drone's state variables in the previous calculation cycle, i.e., a feedback compensation control instruction. Based on three commands, the system conceptually classifies and controls the comprehensive external force disturbance from the perspective of whether it is measurable, without the need for complex mathematical calculations. Finally, the system achieves comprehensive control of the carrier UAV through the simple fusion of control commands, which simplifies the response control calculation under complex and variable disturbance conditions, while improving the speed of disturbance response and the accuracy of compensation.

[0029] As an optional embodiment, the transport drone includes at least a transport platform, and the bottom surface of the transport platform is equipped with multiple sensors. Specifically, the multiple sensors can be multiple single-axis force detection sensors or multi-axis force detection sensors. Each sensor includes at least one detection axis, such as an axis parallel to the transport platform or an axis perpendicular to the plane of the transport platform. The detection axes of each sensor are not identical. Preferably, multiple sensors are arranged on the plane under the transport platform, with multiple sensors included in the direction of each detection axis. The union of the original detection ranges of the multiple sensors corresponding to each detection axis is greater than or equal to the area of ​​the transport platform. Before obtaining the real-time status of the transport drone in the current calculation cycle, the method at least further includes: determining the position of the transport object on the transport platform; determining the target detection range of each sensor corresponding to the type of object based on the position; the union of the target detection ranges of all sensors under the same detection axis is equal to the area of ​​the transport platform; and there is no overlap in the target detection ranges of the sensors under the same detection axis. Preferably, determining the target detection range of each sensor includes at least: acquiring the position of the transport object on the transport platform; determining the distribution characteristics of the disturbance force caused by the gravity of the transport object on each detection axis based on the type of the transport object and the position; determining the target detection range of the sensor on the corresponding detection axis according to the distribution characteristics on each detection axis; the greater the distribution of the disturbance force caused by the gravity of the transport object on the detection axis, the smaller the target detection range of each sensor on the corresponding detection axis. It should be noted that the target detection range of the same sensor is different for different transport objects; the target detection range of the same sensor is also different for different positions of the same transport object on the transport platform. The transport object can be various transportable items such as express delivery boxes and food; the force distribution of the transport platform under the gravity of the object varies depending on its position on the transport platform. The degree of influence on the attitude and balance control of the transport drone also varies. Before determining the force disturbance component each time the transport drone is detected, the detection range of the sensor is adjusted according to the characteristics of the transport object to achieve high-precision, full-coverage disturbance force detection.

[0030] The step of obtaining the real-time status of the carrier drone in the current calculation cycle includes:

[0031] The multiple sensors simultaneously detect changes in force above the transport platform. Specifically, the multiple sensors can be a subset of all sensors arranged on the transport platform, or all sensors arranged on the transport platform. If the sensors involved in detection are a subset of all sensors, before the multiple sensors simultaneously detect changes in force above the transport platform, the method at least includes: determining the minimum detection range of each sensor; determining the distribution characteristics of the disturbance force caused by the gravity of the transport object on each detection axis based on the type and location of the transport object; determining the target detection range of the sensor on the corresponding detection axis as the maximum detection range of the sensor based on the distribution characteristics on each detection axis; solving for the optimal number of sensors and the optimal detection range on each detection axis with the objective function of minimizing the number of sensors involved in detection on each detection axis; determining the target sensor involved in the force change of the current calculation cycle based on the optimal number of sensors; and the target sensor detecting the force on the transport platform according to the optimal detection range. As an optional embodiment, the sensor detection frequency is greater than the time window of the current calculation cycle. In this case, multiple sensor values ​​are obtained within the current calculation cycle, and the force change of the transport platform is determined based on the values ​​detected by multiple sensors. In this application, the detectable force change is regarded as a deterministic disturbance force in the rigid force. In the current calculation cycle, the deterministic disturbance force in the current calculation cycle is detected before control, so that the control command in the current calculation cycle already includes compensation for the deterministic disturbance force, thereby improving the response speed of the carrier UAV control.

[0032] The position and mass information of the transported object above the transport platform are located based on the force changes. As an optional implementation force, the force distribution on the transport platform in each detection axis direction is determined based on the force changes of multiple sensors corresponding to each detection axis; the position information of the transported object on the transport platform is calculated based on the force distribution in each detection axis direction; and the mass information of the transported object is calculated based on the detection values ​​of each sensor. Specifically, in a transport scenario, the transported object is usually placed in the edge area of ​​the transport platform. If the existing technology is still used, only the gravity of the transported object is calculated, which results in a large deviation in the calculation of the deterministic disturbance provided by the transported object to the UAV, leading to deviations in the attitude control of the transported UAV and causing the cargo to slip. Therefore, the position information of the transported object is calculated based on the detection values ​​of multiple sensors, and then deterministic disturbance force and deterministic disturbance torque are calculated simultaneously based on the position information. This provides torque information for subsequent attitude control of the transported UAV, improving the accuracy of attitude control and ensuring the safety of the transported cargo. For details on the implementation of calculating the position of the force application point based on multi-point force detection values, please refer to existing technologies; further details will not be provided here.

[0033] The deterministic perturbation force of the carrier drone is calculated based on the mass information. Specifically, the gravity of the carrier object is calculated based on the mass information, and the value of the deterministic perturbation force relative to the center of mass of the carrier drone is determined according to the distribution of the gravity and the detection values ​​of the multiple sensors. Here, since the sensors detect the force on the entire carrier platform, there is still a certain gap between the gravity and the actual deterministic force relative to the center of mass. At this time, the degree of this deviation is determined according to the distribution of the detected forces of the multiple sensors, and the final force value relative to the center of mass is determined based on the degree of deviation and the gravity. The deterministic perturbation moment is calculated based on the position information and the deterministic perturbation force. The deterministic perturbation moment is specifically the product of the deterministic perturbation force and the deterministic perturbation force arm. The deterministic perturbation force arm is a spatial vector with the reference origin being the center of mass of the carrier drone and the endpoint being the actual point of action of the deterministic perturbation force on the drone body. The endpoint is, for example, the center of mass of the carrier object or the main connection point between the carrier object and the drone platform. The lever arm vector is determined based on the pre-set or real-time acquired position information of the transport object relative to the UAV's body coordinate system. This position information specifically provides the precise spatial coordinates of the point of application of the deterministic disturbance force relative to the UAV's center of mass. In this way, the lever arm vector can be accurately established for subsequent precise calculation of the deterministic disturbance torque.

[0034] The final deterministic disturbance component is a vector, specifically including deterministic disturbance force and deterministic disturbance moment. The method provided by this invention, on the one hand, performs feedforward compensation for the deterministic disturbance force experienced by the UAV during the current calculation cycle, before the control command is issued, achieving zero-delay disturbance response control for the UAV and greatly ensuring the stability of the UAV's attitude. On the other hand, when performing feedforward compensation for the deterministic disturbance component, in addition to the deterministic disturbance force, the deterministic disturbance moment is also calculated, providing sufficient information input for the subsequent feedforward compensation controller. Even when the transport object is in the edge region of the transport platform, high-precision deterministic disturbance force feedforward compensation control can still be achieved, improving the robustness of the UAV's attitude control.

[0035] S102. Generate basic control commands corresponding to the desired attitude information based on the real-time attitude information and the basic controller.

[0036] Specifically, the real-time attitude information refers to the flight attitude parameters of the carrier UAV, including but not limited to parameters affecting flight attitude such as flight acceleration, angular velocity, and altitude. The basic controller is specifically a controller that calculates and controls the real-time attitude of the carrier UAV based on the desired attitude information. The front end of the basic controller includes at least a carrier UAV model. Based on the control target and the carrier UAV model, the desired attitude information is predicted and generated. The desired attitude information includes at least the parameter values ​​of various variables that determine the real-time attitude of the carrier UAV, such as desired acceleration, desired angular velocity, and desired flight altitude. The basic controller calculates basic control commands based on the difference between the real-time attitude and the desired attitude information, and the control algorithm within the basic controller. The control target of the basic control commands is to maintain consistency between the real-time attitude and the desired attitude information. The content of the basic control commands may include control variable values ​​issued to the carrier UAV controller, such as total thrust and torque information.

[0037] The step of generating basic control commands corresponding to the desired attitude information based on the real-time attitude information and the basic controller includes:

[0038] The real-time attitude information and the desired attitude information are input to the base controller. Specifically, the real-time attitude information is detected jointly by multiple types of sensors mounted on the UAV, such as a three-axis accelerometer, angular velocity sensor, and linear velocity sensor. The desired attitude information is predicted by the UAV model. The base controller embeds the SE3 control algorithm to calculate the optimal base control command based on the real-time attitude information and the desired attitude information. Specifically, the desired attitude information includes at least the desired acceleration and desired heading; the real-time attitude information includes at least the current attitude, current deterministic perturbation force, and deterministic perturbation torque; the current attitude includes at least the current actual spatial orientation and rotational motion state of the UAV, such as attitude and angular velocity. The desired attitude information is information generated by the UAV planning or prediction that indicates the target spatial motion trend that the UAV expects to reach, and includes at least the desired linear acceleration and desired heading angle. The desired linear acceleration is used to drive the UAV to a specified position at a specified speed, and the desired heading angle is used to specify the UAV's direction of travel to the specified position. The basic controller is the core module of the control system of the UAV, embedding the SE(3) control algorithm. It calculates the original control output to drive the UAV to achieve basic flight and attitude stability based on the deviation between real-time attitude information and desired attitude information. The SE(3) control algorithm simultaneously processes the three-dimensional translation and rotation of the UAV, calculating the optimal total thrust and three-axis torque to achieve precise tracking of the desired trajectory and attitude. Specifically, the basic control command, calculated and output by the basic controller, contains the original command of the UAV's total thrust and three-axis attitude control torque, which is the basis for driving the UAV motors to generate thrust and torque. The basic controller calculates the total thrust based on the force balance relationship of the UAV. It determines the force balance relationship between multiple external forces acting on the UAV based on the real-time attitude information and desired attitude information, and calculates the desired thrust based on the force balance relationship.

[0039] m represents the mass of the drone being transported. Let g be the acceleration of the drone being transported, and g be the acceleration due to gravity. [0,0,1] T The standard unit vector, Let be the rotation matrix from the machine body to the world coordinate system. The force generated is applied to the cargo measured in the body coordinate system.

[0040] Calculate the total thrust based on the expected thrust. .

[0041] Further, based on the desired attitude information, the desired z-axis of the carrier UAV is calculated: ; Calculate the desired x-axis of the carrier UAV based on the desired attitude information: , This represents the desired yaw angle for the drone.

[0042] Construct a rotation matrix based on the described force balance relationship: ,in, , The attitude error vector is calculated based on the rotation matrix: Where V is the Vee graph transformation, which converts an antisymmetric matrix into a vector representation. The superscript T indicates transpose.

[0043] Calculate the angular velocity error vector based on the real-time attitude information. , This indicates an estimated angular velocity. Representing the desired angular velocity, the attitude error vector and the angular velocity error vector are fused to generate the ideal torque: , , Given a constant coefficient, J is the moment of inertia of the UAV. To estimate the body's angular velocity, for The derivative of the given torque is used to calculate the foundation torque based on the difference between the ideal torque and the deterministic disturbance torque. , For ideal torque, To determine the disturbance torque, the total thrust and the basic torque are used as the basic control command.

[0044] S103. Based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller, adaptively generate the feedforward compensation control command corresponding to the deterministic disturbance component in the rigid force.

[0045] The deterministic disturbance feedforward controller is used to generate feedforward compensation control commands based on the deterministic disturbance components detected by the sensors, which, together with the basic control commands, serve as real-time compensation quantities. The generation of feedforward compensation control commands based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller includes:

[0046] The deterministic disturbance force and the deterministic disturbance torque are input to the deterministic disturbance feedforward controller. In addition to inputting the deterministic disturbance force to the controller, the method provided by this invention also synchronously inputs the deterministic disturbance force to the controller, further improving the accuracy of the compensation amount corresponding to the deterministic disturbance component estimated by the deterministic disturbance feedforward controller. The deterministic disturbance feedforward controller determines an allocation matrix corresponding to the UAV. The allocation matrix includes at least a force allocation matrix and a torque allocation matrix, used to characterize the conversion relationship between the force and torque from the output space of each independent actuator to the UAV in the body coordinate system. The actuator is, for example, a propeller, etc. The force is the total thrust experienced by the UAV, and the torque is the three-axis attitude torque such as roll, pitch, and yaw. Based on the allocation matrix, the deterministic disturbance feedforward controller maps and detects to obtain the deterministic disturbance components experienced at each position, including the deterministic disturbance force and the deterministic disturbance torque. The deterministic disturbance component information is converted into forces and moments of the UAV in the body coordinate system based on the allocation matrix. Then, corresponding feedforward compensation is calculated based on the converted forces and moments. The feedforward gain is the inverse of the allocation matrix. The function of the feedforward gain is to directly convert the deterministic disturbance force and torque, which are external disturbance inputs, into compensation control commands that can be applied to the various actuators of the UAV. This ensures that the generated compensation commands can accurately produce an effect on the UAV body that is equal in magnitude and opposite in direction to the disturbance force and torque, thereby achieving immediate cancellation of rigid disturbances. Specifically, multiple historical compensation information of the UAV is determined; the force allocation matrix and torque allocation matrix of the UAV are learned based on the multiple historical compensation information; and the force allocation matrix and torque allocation matrix are corrected based on the compensation margin of the next compensation information after each historical compensation information. Historical compensation information includes at least the compensation input, namely the force and torque information of the deterministic disturbance components detected by the sensors, and at least the compensation output, namely the compensation command obtained after feedforward gain conversion. Based on the historical compensation command, a deterministic disturbance feedforward controller specific to the UAV is learned. Then, the deterministic disturbance feedforward controller can directly convert multiple measurable components (deterministic disturbance components) in the actuator output space into components in the UAV body coordinate system according to the allocation matrix. Based on this component, a feedforward compensation control command is generated and fused with the basic command to obtain a comprehensive command. From the perspective of the UAV, all commands are issued at once in the form of a comprehensive command, which automatically cancels the disturbance caused by the deterministic disturbance components while completing the basic control. The torque command is generated in the same way, and will not be described in detail again.After obtaining the force allocation matrix and torque allocation matrix, the method further includes modifying the allocation matrix. Specifically, the instruction issued is a learned compensation instruction, but its actual compensation effect cannot be accurately measured, i.e., it cannot be determined whether the learned input and output is the optimal compensation control content. Therefore, based on the learned allocation matrix, the predicted historical compensation instruction corresponding to the next historical compensation information of the current historical compensation information is predicted. The compensation margin is determined based on the difference between the predicted historical compensation instruction and the actual historical compensation instruction. The allocation matrix is ​​modified based on the compensation margin, so that the compensation instruction generated by the deterministic disturbance feedforward controller according to the transformation matrix is ​​equal to the predicted historical compensation instruction. The feedforward compensation control instruction is calculated based on the product of the allocation matrix and the deterministic disturbance force and the deterministic disturbance torque. Kff is the feedforward gain calculated based on the UAV control allocation matrix. Since the forces and torques of the UAV are measured and calculated in real time, the compensation has almost no delay and can instantly offset most of the rigid force disturbances.

[0047] S104. Based on the flight state of the UAV after receiving the integrated control command in the previous calculation cycle and the feedforward compensation control command in the current calculation cycle, adaptively predict the feedback compensation control command corresponding to the uncertainty disturbance component in the rigid force in the current calculation cycle.

[0048] The previous calculation cycle refers to the calculation cycle preceding the current calculation cycle. The control of each calculation cycle includes the instruction corresponding to the control target of the current cycle, the instruction corresponding to the real-time detection of uncertainty disturbance components in the current calculation cycle, and the instruction corresponding to the uncertainty disturbance obtained based on the state prediction of the previous calculation cycle.

[0049] Specifically, the flight state of the UAV after receiving the integrated control command in the previous calculation cycle is detected; the feedforward compensation control command calculated by the UAV in the current calculation cycle is obtained; the flight state and the feedforward compensation control command are input to the feedback compensation controller; the feedback compensation controller estimates the total compensation amount of the disturbance component based on the flight state, and removes the compensation amount corresponding to the feedforward compensation control command from the total compensation amount to obtain the compensation amount corresponding to the uncertainty disturbance component. In order to obtain the true state of the UAV after various control strategies are modified, so as to evaluate the uncertainty disturbance component experienced by the UAV, the real-time state of the UAV after all control commands in the previous cycle are detected as the initial attitude for attitude control. Preferably, after receiving the real-time status of the previous cycle, the status information of the previous cycle and the feedforward compensation control command already calculated in the current calculation cycle are jointly input into the feedback compensation controller. As an optional embodiment, the feedback compensation controller is an L1 adaptive controller. Since the command for the current calculation cycle has not yet been issued, the input to the feedback compensation controller is the real-time status of the UAV after the command for the previous calculation cycle was issued, and the feedforward compensation amount already calculated for deterministic disturbance components in the current calculation cycle, i.e., the feedforward compensation control command. The output is the compensation amount for unmeasurable disturbances, i.e., the compensation amount for uncertain disturbance components. Because uncertain disturbance components cannot be directly measured, the feedback compensation controller infers them by observing their effect on the system, i.e., the state error, and performs compensation accordingly.

[0050] The feedback compensation controller, based on the total compensation amount of the predicted disturbance component in the flight state, removes the compensation amount corresponding to the feedforward compensation control command from the total compensation amount to obtain the compensation amount corresponding to the uncertain disturbance component. This includes: establishing a state predictor for the UAV, which predicts the real-time state of the UAV at the next moment based on the real-time information of the UAV; using the difference between the state predicted by the state predictor and the detected real-time state of the UAV as the total influence margin of the disturbance component; using the product of the update law and the current real-time state information of the UAV as the adaptive law of the feedback compensation controller, predicting the total disturbance component based on the total influence margin; removing the deterministic disturbance component corresponding to the feedforward compensation control command from the total disturbance component to obtain the non-deterministic disturbance component; and generating a feedback compensation control command for the non-deterministic disturbance component based on a filter. A state predictor parallel to the carrier UAV is constructed. The state prediction error is defined as the difference between the state predicted by the state predictor and the state detected by the carrier UAV in real time. When a disturbance component actually exists, the state prediction error is not zero. The state prediction error caused by the disturbance component is then parameterized into the adaptive law of the feedback controller. Where B(R) is a regression matrix, including the rotation matrix, mass, and moment of inertia of the UAV, used to map the estimated uncertainty to the prediction error space, and h is the update law, specifically:

[0051] ;

[0052] in, Let I be the predefined Hurwitz matrix, and let I represent the identity matrix, which in turn represents the gain and convergence rate of the adaptive law. For state prediction error, Compensation for time-related changes.

[0053] Using the above method, the total disturbance component experienced by the UAV is derived from the state. The previously measured deterministic disturbance component is then removed from the total disturbance component. Preferably, after the filter generates the command, the control command corresponding to the deterministic disturbance component is subtracted at the control command level. Alternatively, before the filter generates the command, the deterministic disturbance force is subtracted from the predicted total disturbance force at the disturbance force level. Subsequently, the control command corresponding to the disturbance component is calculated. ,in, It is the cutoff frequency of the low-pass filter, used to balance the response speed and smoothness of the disturbance estimation. For adaptive laws, This is the compensation instruction obtained from the previous step's length calculation.

[0054] S5. Integrate the basic control command, the feedforward compensation control command, and the feedback compensation control command to generate an adaptive integrated control command for the carrier UAV.

[0055] The control terms in the basic control command, the feedforward compensation control command, and the feedback compensation control command are determined. The control values ​​corresponding to the same control term are fused, and all control terms are combined to generate the comprehensive control command. Specifically, the fusion involves summing the three control commands. The control terms corresponding to each control command are not entirely the same. The fusion specifically includes: summing the control values ​​corresponding to the same control term; combining the control values ​​corresponding to different control terms; if the directions of the corresponding control values ​​are the same, the fusion is the sum of the absolute values ​​of the control values, and the control direction is consistent with the direction of any control value. If the control directions of the control values ​​of the same control term are opposite, the fusion is the difference of the absolute values ​​of the control values, and the direction of the control value in the fusion result is consistent with the direction of the control value with the larger absolute value. By superimposing these three control quantities, a hierarchical and collaborative control strategy is achieved. The basic controller ensures tracking, the feedforward controller quickly cancels large disturbances, and the L1 adaptive feedback controller accurately eliminates residual small disturbances and model uncertainties, ultimately achieving high robustness and high precision control of the UAV under complex physical interactions.

[0056] The adaptive rigid force response control method for unmanned aerial vehicles (UAVs) provided in this embodiment achieves high-precision and robust adaptive control of UAV attitude in complex transportation environments by conceptually deconstructing the sources of disturbances and adopting differentiated compensation strategies. Specifically, the rigid force is explicitly decomposed into measurable deterministic disturbances and unmeasurable nondeterministic disturbances. The deterministic disturbance component is directly fed into the feedforward controller as a known input, so that the state prediction of feedback control only needs to handle nondeterministic disturbances, significantly reducing the uncertainty of feedback control and fundamentally improving the fidelity of state prediction during feedback control. This effectively avoids system divergence caused by prediction mismatch in strong impact scenarios such as large swings or position shifts of cargo. This application uses multi-axis force sensors distributed at the bottom of the transportation platform to detect and calculate deterministic disturbance forces and torques in real time, directly inputting them into the feedforward controller to generate compensation commands. This bypasses the inherent time delay of traditional observers and achieves zero-delay response of measurement and compensation for known disturbances, greatly improving the real-time performance and accuracy of control and ensuring the stability of the transportation process. From the perspective of the overall controller structure, this application adopts a composite control architecture of feedforward and feedback. Feedforward control focuses on compensating for measurable, large-amplitude, and slow-dynamic deterministic disturbances; feedback control, on the other hand, adaptively estimates and compensates for unmeasurable and highly random nondeterministic disturbances through state prediction models and filters, separating the main disturbance sources from random noise, ensuring that the system has good performance predictability in both transient and steady-state conditions, and significantly enhancing overall robustness.

[0057] Corresponding to the aforementioned embodiments of the adaptive rigid force response control method for unmanned aerial vehicles (UAVs), this application also provides an embodiment of an adaptive rigid force response control device for unmanned aerial vehicles (UAVs).

[0058] Figure 2 This is a structural schematic diagram of the adaptive rigid force response control device for the unmanned aerial vehicle provided in this application. Please refer to... Figure 2 The apparatus provided in this embodiment includes:

[0059] The acquisition module 210 is used to acquire the real-time state of the carrier drone in the current calculation cycle. The real-time state includes at least the real-time attitude information of the carrier drone and the deterministic perturbation force and deterministic perturbation torque in the rigid force.

[0060] The basic control module 220 is used to generate basic control commands corresponding to the desired attitude information based on the real-time attitude information and the basic controller.

[0061] The feedforward compensation module 230 is used to adaptively generate feedforward compensation control commands corresponding to the deterministic disturbance component in the rigid force based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller.

[0062] Feedback compensation module 240 is used to adaptively predict the feedback compensation control command corresponding to the uncertainty disturbance component in the rigid force in the current calculation cycle based on the flight state of the UAV after receiving the integrated control command in the previous calculation cycle and the feedforward compensation control command in the current calculation cycle.

[0063] The control module 250 is used to integrate the basic control commands, the feedforward compensation control commands, and the feedback compensation control commands to generate adaptive integrated control commands for the carrier UAV.

[0064] The apparatus of this embodiment can be used to perform... Figure 1 The steps of the method embodiment shown are similar in principle and process, and will not be repeated here.

[0065] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0066] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0067] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An adaptive rigid force response control method for a carrier unmanned aerial vehicle, characterized in that, The method includes: Obtain the real-time state of the carrier drone in the current calculation cycle. The real-time state includes at least the real-time attitude information of the carrier drone and the deterministic perturbation force and deterministic perturbation torque in the rigid force. Based on the real-time attitude information and the basic controller, basic control commands corresponding to the desired attitude information are generated. Based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller, the feedforward compensation control command corresponding to the deterministic disturbance component in the rigid force is adaptively generated. Based on the flight state of the UAV after receiving the integrated control command in the previous calculation cycle and the feedforward compensation control command in the current calculation cycle, the feedback compensation control command corresponding to the uncertainty disturbance component in the rigid force in the current calculation cycle is adaptively predicted. The prediction is based at least on the uncertainty disturbance force and the uncertainty disturbance torque. By integrating the basic control commands, the feedforward compensation control commands, and the feedback compensation control commands, an adaptive integrated control command for the carrier UAV is generated.

2. The method according to claim 1, characterized in that, The transport drone includes at least a transport platform, and the bottom surface of the transport platform is equipped with multiple sensors; obtaining the real-time status of the transport drone in the current calculation cycle includes: The multiple sensors simultaneously detect changes in force above the transport platform; The position and mass information of the object being transported above the transport platform are determined based on the force changes. The deterministic disturbance force of the carrier UAV is calculated based on the quality information; The deterministic disturbance moment is calculated based on the location information and the deterministic disturbance force.

3. The method according to claim 1, characterized in that, The step of generating basic control commands corresponding to the desired attitude information based on the real-time attitude information and the basic controller includes: The real-time attitude information and the desired attitude information are input to the basic controller; The basic controller calculates the total thrust based on the force balance relationship of the carrier UAV; Construct a rotation matrix based on the force balance relationship; Calculate the attitude error vector based on the rotation matrix; Calculate the angular velocity error vector based on the real-time attitude information; By fusing the attitude error vector and the angular velocity error vector, an ideal torque is generated. The foundation torque is calculated based on the difference between the ideal torque and the deterministic disturbance torque; The total thrust and the basic torque are used as the basic control commands.

4. The method according to claim 1, characterized in that, Based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller, a feedforward compensation control command is generated, including: The deterministic disturbance force and the deterministic disturbance torque are input to the deterministic disturbance feedforward controller; The deterministic perturbation feedforward controller determines the allocation matrix corresponding to the UAV; The feedforward compensation control command is calculated based on the product of the allocation matrix and the deterministic disturbance force and the deterministic disturbance torque, respectively.

5. The method according to claim 1, characterized in that, Based on the flight state of the UAV after receiving integrated control commands in the previous calculation cycle and the feedforward compensation control commands in the current calculation cycle, the feedback compensation control commands corresponding to the uncertainty disturbances in the current calculation cycle are predicted, including: Detect the flight status of the carrier UAV after receiving the integrated control command in the previous calculation cycle; Obtain the feedforward compensation control command calculated by the carrier UAV in the current calculation cycle; The flight status and the feedforward compensation control command are input to the feedback compensation controller; The feedback compensation controller, based on the total compensation amount of the predicted disturbance component of the flight state, removes the compensation amount corresponding to the feedforward compensation control command from the total compensation amount to obtain the compensation amount corresponding to the uncertainty disturbance component.

6. The method according to claim 5, characterized in that, The feedback compensation controller, based on the total compensation amount of the predicted disturbance component in the flight state, subtracts the compensation amount corresponding to the feedforward compensation control command from the total compensation amount to obtain the compensation amount corresponding to the uncertain disturbance component, including: A state predictor for the transport drone is established, which is used to predict the real-time state of the transport drone at the next moment based on the real-time information of the transport drone. The difference between the state predicted by the state predictor and the real-time state of the carrier UAV detected is used as the total influence margin of the disturbance component. The product of the update law and the real-time state information of the current UAV is used as the adaptive law of the feedback compensation controller, and the total disturbance component is predicted based on the total influence margin. The deterministic disturbance component corresponding to the feedforward compensation control command is removed from the total disturbance component to obtain the uncertain disturbance component; The filter generates feedback compensation control commands for the uncertain disturbance components.

7. The method according to claim 1, characterized in that, The process of integrating the basic control command, the feedforward compensation control command, and the feedback compensation control command to generate a comprehensive control command includes: Determine the control items in the basic control command, the feedforward compensation control command, and the feedback compensation control command; The control values ​​corresponding to the same control item are merged, and all control items are combined to generate the comprehensive control instruction.

8. The method according to claim 1, characterized in that, Before obtaining the real-time status of the carrier drone in the current calculation cycle, the method further includes at least: Determine the position of the object being transported on the transport platform; Determine the target detection range of each sensor corresponding to the type of the transported object based on the location; Wherein, the union of the target detection ranges of all sensors under the same detection axis is equal to the area of ​​the transport platform, and there is no overlap in the target detection ranges of the various sensors under the same detection axis.

9. The method according to claim 4, characterized in that, The allocation matrix includes at least a force allocation matrix and a torque allocation matrix. The deterministic perturbation feedforward controller determines the allocation matrix corresponding to the UAV, including: Determine multiple historical compensation information for the drone; The force distribution matrix and torque distribution matrix of the UAV are learned based on the multiple historical compensation information. The force allocation matrix and the torque allocation matrix are modified based on the compensation margin of the next compensation information according to each historical compensation information.

10. An adaptive rigid force response control device for a transport unmanned aerial vehicle, characterized in that, The device includes: The acquisition module is used to acquire the real-time state of the carrier drone in the current calculation cycle. The real-time state includes at least the real-time attitude information of the carrier drone and the deterministic perturbation force and deterministic perturbation torque in the rigid force. The basic control module is used to generate basic control commands corresponding to the desired attitude information based on the real-time attitude information and the basic controller. The feedforward compensation module is used to adaptively generate feedforward compensation control commands corresponding to the deterministic disturbance component in the rigid force based on the deterministic disturbance force, the deterministic disturbance torque, and the deterministic disturbance feedforward controller. The feedback compensation module is used to adaptively predict the feedback compensation control command corresponding to the uncertainty disturbance component in the rigid force in the current calculation cycle based on the flight state of the UAV after receiving the integrated control command in the previous calculation cycle and the feedforward compensation control command in the current calculation cycle. The prediction is based at least on the uncertainty disturbance force and the uncertainty disturbance torque. The control module is used to integrate the basic control commands, the feedforward compensation control commands, and the feedback compensation control commands to generate adaptive integrated control commands for the carrier UAV.

Images