Trajectory tracking control method and system for air release of child-mother type unmanned aerial vehicle

Abstract

The invention discloses a trajectory tracking control method and system for air release of a child-mother type unmanned aerial vehicle, and belongs to the technical field of unmanned aerial vehicle control, and the method comprises the following steps: carrying out the real-time calculation based on the release state of the child-mother type unmanned aerial vehicle, obtaining a time-varying parameter, and constructing a time-varying dynamic model; respectively constructing a position third-order linear expansion state observer and three attitude second-order linear expansion state observers with parallel channels; the total thrust and the expected attitude are obtained; calculating to obtain a control torque; constructing a total cost function, and performing iterative optimization based on candidate solution vectors constructed by particle swarm optimization algorithm controller parameters to obtain real-time target controller parameters; and dynamically adjusting a trajectory tracking controller based on the real-time target controller parameters so as to continuously perform trajectory tracking on the process of dynamically releasing the son-mother type unmanned aerial vehicle in the son-mother type unmanned aerial vehicle. The system continuously performs trajectory tracking based on the method. According to the invention, the problem that the child-mother type unmanned aerial vehicle trajectory tracking control accuracy and stability are insufficient is solved.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) control technology, and particularly relates to a trajectory tracking control method and system for the aerial release of mother-daughter UAVs. Background Technology

[0002] With the rapid development of low-altitude economy, emergency rescue, and intelligent logistics, multi-UAV collaborative systems have demonstrated immense application value. Mothership-Daughter UAV Systems, as a novel aerial operation platform, leverage the combined advantages of a long-endurance mother aircraft and highly flexible daughter aircraft to achieve multi-angle, wide-area, and rapid mission deployment. However, the technical bottleneck of this system lies in the flight stability and trajectory tracking accuracy of the mother aircraft during the dynamic release of the daughter aircraft. To ensure the safety of the mother aircraft and guarantee the accurate deployment of the daughter aircraft in the correct location, the controller must maintain high-precision tracking of the desired trajectory throughout the entire mission cycle—before, during, and after the release of the daughter aircraft.

[0003] When the mother aircraft dynamically releases the daughter aircraft in the air, it will encounter extremely violent transient disturbances and dynamic changes within the system, which poses a core challenge to trajectory tracking control. These complex disturbances are mainly manifested in three aspects: (1) abrupt changes in the physical properties of the system, that is, the total mass, center of mass position and moment of inertia matrix of the mother aircraft will change instantaneously and abruptly; (2) mechanical recoil and aerodynamic shock, including the instantaneous shock generated by the release mechanism and the complex, non-analytical aerodynamic disturbance force and torque caused by the separation of the daughter aircraft; (3) strongly coupled nonlinear problems, that is, the changes in the center of mass and inertia significantly enhance the coupling effect between the translational and rotational channels. However, existing trajectory tracking control technologies are difficult to effectively solve the trajectory maintenance problem in this scenario, mainly due to the applicability limitations of the model or strategy. For dynamic modeling, existing research is mostly focused on sling load models, variable mass models, or fixed-wing airdrop models. In terms of control strategies, traditional intelligent control strategies have limited generalization capabilities to cope with sudden and dramatic changes in internal parameters, while classic model-based control strategies struggle to converge quickly under instantaneous step disturbances. Furthermore, active disturbance rejection control inherently suffers from compensation lag when handling step disturbances of known parameters. Therefore, there is an urgent need for a composite control method capable of actively handling sudden changes in known parameters and suppressing unknown transient shocks in real time. Summary of the Invention

[0004] To address the aforementioned shortcomings in the existing technology, this invention provides a trajectory tracking control method and system for in-flight release of mother-daughter UAVs, which solves the problems of insufficient accuracy and stability in trajectory tracking control of mother-daughter UAVs.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: On the one hand, the present invention provides a trajectory tracking control method for in-flight release of mother-daughter unmanned aerial vehicles, comprising the following steps: S1. Based on the release state of the mother-daughter UAV, time-varying parameters are calculated in real time, and a time-varying dynamic model is constructed. S2. Construct a position third-order linear extended state observer and three attitude second-order linear extended state observers with parallel channels respectively. S3. Based on the time-varying dynamics model and the third-order linear extended state observer, the composite control law is derived and the thrust is calculated to obtain the total thrust and the desired attitude. S4. Calculate the command angular acceleration based on the desired attitude, define the error signal, and calculate the control torque based on three parallel attitude second-order linear extended state observers, nonlinear dynamic inverse and time-varying dynamic model. S5. Construct a total cost function based on the total thrust and control torque, and iteratively optimize the candidate solution vector constructed from the controller parameters of the trajectory tracking controller using the particle swarm optimization algorithm to obtain the real-time target controller parameters; S6. The trajectory tracking controller is dynamically adjusted based on the real-time target controller parameters to continuously track the trajectory of the mother-daughter UAV during the dynamic release of the daughter unit.

[0006] Further, step S1 includes the following steps: S11. Based on the release state of the mother drone, set the release state vector of the mother drone with respect to the daughter drone; S12. Based on the release state vector of the mother machine with respect to the daughter machine, the time-varying parameters are calculated in real time, including the total mass, total center of mass and total moment of inertia matrix. The calculation expression for the time-varying parameter in S12 is as follows: , , , , , , in, express The total mass of the mother machine and all unreleased daughter machines at that moment. Indicates the mass of the mother machine. This indicates the total number of slave machines. express The mother machine of the moment about the first Release state vector of each submachine Indicates the first The quality of the individual machine Indicates belonging to, express The position of the center of mass of the mother machine and all unreleased daughter machines at any given moment. Indicates the position of the mother machine's center of gravity. Indicates the first The position of the center of gravity of the machine express The total moment of inertia of the mother machine and all unreleased slave machines at any given moment. This represents the moment of inertia of the mother machine about its own center of mass. The antisymmetric matrix representing the translation vector of the mother machine. Indicates the first The moment of inertia of a single machine about its own center of mass. Indicates the first The antisymmetric matrix of the translation vectors of each submachine. Describes the antisymmetric matrix of a matrix. This represents the translation vector of the mother machine. Indicates the first The translation vector of each submachine, where, when At that time, The first in the mother machine The machine was not released when At that time, The first in the mother machine The individual machine has been released; S13. Based on the total centroid in the time-varying parameters, and considering the centroid shift caused by the release of the slave unit by the mother machine and the rotor thrust, the quasi-static disturbance torque and transient impact disturbance are calculated. The transient impact disturbance includes net recoil force and net recoil torque. The calculation expressions for the quasi-static disturbance moment and transient impact disturbance are as follows: , , , in, This represents the quasi-static disturbance moment. This represents the total thrust generated by the rotor. This represents the net recoil force exerted on the center of mass when the submachine gun is released. This refers to the release mechanism of the slave unit on the mother machine. Indicates that the mother machine passes through the first The reverse impact force experienced by the release mechanism when releasing the submachine. Indicates the net recoil torque. Indicates the first The center of mass position of the release mechanism; S14. Define aerodynamic interference force and aerodynamic interference torque based on the flow field disturbed at the moment of release of the slave unit; S15, Based on the mother machine body coordinate system The net recoil force, aerodynamic disturbance force, and external disturbance force are combined to construct a composite disturbance force. The calculation expression for the composite disturbance force is as follows: , in, Indicates the combined disturbance force. This represents the rotation matrix from the mothership's body coordinate system to the ground inertial coordinate system. Indicates aerodynamic interference force. Indicates external disturbance force; S16. Based on Newton-Euler's equations, the time-varying inertia... Differentiating the derivative yields the impact torque caused by the sudden change in rotational inertia. S17, Based on the mother machine body coordinate system The quasi-static disturbance torque, net recoil torque, aerodynamic disturbance torque, and moment of inertia change impact torque are used to construct a composite disturbance torque; The calculation expression for the composite disturbance moment is as follows: , in, Indicates the combined disturbance torque. Indicates aerodynamic disturbance torque. This represents the impact torque caused by a sudden change in rotational inertia. S18. Construct a time-varying dynamic model based on composite perturbation force and composite perturbation moment; The calculation expression for the time-varying dynamic model is as follows: , , in, This indicates the acceleration of the mother machine. This represents the total thrust generated by the rotor. Represents the gravitational acceleration vector. Indicates centripetal acceleration. This represents the angular acceleration of the mother machine. Represents the coordinate system of the mother machine. The three-axis control torque below, This represents the angular velocity of the mother machine. This represents the angular velocity matrix of the mother machine.

[0007] Further, step S2 includes the following steps: S21. Based on the second-order integrator system of translational dynamics reconstruction, a third-order linear extended state observer is constructed, wherein the third-order linear extended state observer is used to predict the lumped disturbance acceleration in real time. S22. Based on the first-order integrator system of rotational dynamics reconstruction, construct three parallel second-order linear extended state observers for attitude, wherein the three parallel second-order linear extended state observers for attitude are used to predict lumped rotational disturbances in real time. The calculation expression for the attitude second-order linear extended state observer is as follows: , in, Indicates the first Predicted angular acceleration values ​​for each channel. Indicates the first Predicted values ​​of lumped disturbances for each channel. This represents the bandwidth of the attitude second-order linearly extended state observer. Indicates the first The angular velocity of each channel, Indicates the first Predicted values ​​of angular velocity for each channel. Indicates the first The rate of change of the predicted value of the lumped disturbance for each channel, where, =1, 2, 3.

[0008] Further, step S3 includes the following steps: S31. Based on the time-varying dynamic model, the translational dynamic model is decoupled with acceleration as the core to obtain the decoupled translational dynamic model of the position controller. The calculation expression for the decoupled translational dynamics model of the position controller is as follows: , , , , in, Indicates thrust acceleration. Indicates a known acceleration. Represents the lumped disturbance acceleration. It is defined as; S32. Calculate the desired speed reference command based on the position error and integral error according to the outer position loop; The calculation expression for the desired speed reference command is as follows: , in, Indicates positional error. Indicates the mother machine in the ground inertial coordinate system The lower position, Indicates the trajectory position command. Indicates the integral error. express Position error at any given time This represents the result of the time derivative. Indicates the desired speed reference command. The adjustment factor representing the positive definite diagonal gain at position. Indicates the mother machine in the ground inertial coordinate system The speed of the drop This represents the position positive definite diagonal gain matrix. This represents the integral positive definite diagonal gain matrix; S33. Based on the desired speed reference command, use the speed loop to calculate the ideal feedback acceleration based on the speed tracking error; The expression for calculating the ideal feedback acceleration is as follows: , , in, Represents ideal feedback acceleration. Indicates the dynamic speed error. This represents the positive definite diagonal gain matrix of the velocity. Indicates speed error; S34. Based on the nonlinear dynamic inverse and ideal feedback acceleration, the desired thrust acceleration is calculated using the decoupled dynamic model of the position controller. The expression for calculating the desired thrust acceleration is as follows: , in, This represents the desired thrust acceleration. This represents the predicted value of the lumped disturbance acceleration; S35. Calculate the desired thrust vector based on the desired thrust acceleration and the real-time total mass in the time-varying parameters. The expression for calculating the desired thrust vector is as follows: , in, Represents the desired thrust vector; S36. Decompose the desired thrust vector to obtain the desired total thrust magnitude and desired attitude; The calculation expressions for the desired total thrust amplitude and desired attitude are as follows: , , in, This represents the expected total thrust. This represents the expected total thrust amplitude. It expresses a desired attitude.

[0009] Further, step S4 includes the following steps: S41. Differentiate the desired attitude that changes with time to calculate the command angular velocity, and then calculate the command angular acceleration by filtering and differentiating the command angular velocity. S42. Define the error signal, which includes attitude error and angular velocity tracking error; S43. Construct the target angular acceleration based on the command angular acceleration and error signal; The expression for calculating the target angular acceleration is as follows: , in, Indicates the target angular acceleration. Indicates the command angular acceleration. Represents the angular acceleration gain matrix. Indicates angular velocity tracking error. The adjustment factor representing the attitude gain. Represents the attitude gain matrix. Indicates attitude error; S44. Based on the target angular acceleration, the controller input angular acceleration of each channel is calculated using three parallel second-order linear extended state observers. The calculation expressions for the controller input angular acceleration of each channel are as follows: , in, Indicates the first The controller input angular acceleration for each channel, Indicates the first The target desired angular acceleration for each channel; S45. Based on the controller input angular acceleration, target angular acceleration and time-varying dynamic model of each channel, the control torque is calculated based on the nonlinear dynamic inverse and rotational dynamic model. The formula for calculating the control torque is as follows: , in, This represents the predicted value of the lumped disturbance. Indicates a known disturbance torque; Further, step S5 includes the following steps: S51. Construct the total cost function based on the total thrust and control torque; The total cost function is calculated as follows: , , , , , in, Represents particles Total cost Indicates instantaneous performance cost. Indicates instantaneous energy consumption cost. Indicates the end time of the mother machine release control. Indicates the start time of the mother machine release control. Indicates the derivative with respect to time. Indicates the basic weight of the location. Indicates the first One axis, express Axial direction, express Axial direction, express Axial direction, Indicates the first Axial weights for each axis, Indicates the first The absolute value of the axial position error in each axis. Indicates pose weights, Indicates the first One attitude angle, Indicates the roll angle. Indicates pitch angle, Indicates the yaw angle. Indicates the first The attitude weights for each attitude angle. Indicates the first The absolute value of the attitude error for each attitude angle. Indicates the global energy consumption weight. Indicates thrust weight. Indicates total thrust. Indicates the first Weighting of the control torque along each axis Indicates the first Control torque in each axis; S52. Based on the physical symmetry of the mother-daughter UAV in the XY plane, construct the candidate solution vector of the trajectory tracking controller based on the controller parameters of the trajectory tracking controller; S53. Treat each candidate solution vector of the trajectory tracking controller as a particle. ; S54. Based on the particle swarm optimization algorithm, the optimal solution vector of the real-time trajectory tracking controller is obtained by continuous iteration with the total cost function as the objective function. S55. Use the controller parameters corresponding to the optimal solution vector of the real-time trajectory tracking controller as the real-time target controller parameters.

[0010] Further, S54 includes the following steps: S541. Based on the particle swarm algorithm and the parameter space of the candidate solutions of the trajectory tracking controller, initialize the particle swarm and the current iteration number to 0. S542. Based on the position and velocity of each particle in the particle swarm, execute the mother machine release control program respectively, and extract the three axial position errors, the three attitude angle errors, the total thrust and the three axial control torques as the simulation output of the particle. S543. Based on the simulation output of each particle, evaluate the total cost of each particle in the particle swarm based on the total cost function. S544. Based on the total cost of each particle, update the individual historical best value and the global historical best value in the particle swarm. S545. Update the velocity and position of each particle based on the individual historical best value in the particle swarm and the global historical best value. S546. Determine whether the current iteration count has reached the iteration count threshold. If yes, proceed to S547; otherwise, increment the current particle iteration count by 1 and return to S542. S547. The candidate solution vector corresponding to the global historical best value is used as the optimal solution vector of the trajectory tracking controller.

[0011] The beneficial effects of this invention are as follows: This invention provides a trajectory tracking control method for in-flight release of mother-daughter UAVs. It introduces nonlinear dynamic inverse NDI control, acquiring time-varying parameters in real time. The nonlinear dynamic inverse controller utilizes this precise time-varying model information to actively and instantly decouple and feedforward compensate for gravitational acceleration, centripetal acceleration, gyro coupling torque, and centroid offset torque in the control law derivation. Through the feedforward combined with the active compensation mechanism of nonlinear dynamic inverse, it avoids the prediction errors and delays of the extended state observer when dealing with known step jumps, significantly improving the trajectory smoothness and accuracy of the system during the deployment transient process. This invention utilizes a dual-channel linear extended state observer for position and attitude, focusing passive suppression on these unmodeled transient dynamics, treating all residual disturbances as lumped disturbances for real-time prediction and compensation. Its high bandwidth characteristics ensure protection against instantaneous impacts. The rapid acquisition and decay of the linear extended state observer ensures that even with severe model mismatch, it can still accurately identify and compensate for steady-state gravity disturbances in real time, while keeping the position error within the centimeter level, thus ensuring the excellent robustness of the trajectory tracking controller under strong uncertainties. The composite trajectory tracking control scheme provided by this invention includes up to 16 highly coupled key parameters such as inner loop, outer loop gain, and ESO bandwidth. By introducing a particle swarm optimization algorithm and adopting a position-attitude joint optimization strategy, the parameter tuning problem is transformed into an optimization problem with a total cost function as the objective. This total cost function balances trajectory tracking performance and control energy consumption. Through the global optimization capability of the particle swarm optimization algorithm, it successfully converges to the target controller parameters in 16-dimensional space. This not only breaks through the parameter tuning bottleneck but also ensures that the trajectory tracking control scheme can achieve optimal performance in complex tasks.

[0012] On the other hand, the present invention also provides a trajectory tracking control system based on a trajectory tracking control method for in-flight release of mother-daughter UAVs, comprising: The parameter update module is used to calculate time-varying parameters in real time based on the release state of the mother-daughter UAV and to build a time-varying dynamic model. The channel expansion observer module is used to construct a position third-order linear expansion state observer and three attitude second-order linear expansion state observers in parallel channels. The outer loop module is used to derive the composite control law and calculate the thrust based on the time-varying dynamics model and the third-order linear extended state observer of the position, so as to obtain the total thrust and the desired attitude. The attitude inner loop module is used to calculate the command angular acceleration based on the desired attitude and define the error signal. It also calculates the control torque based on three parallel attitude second-order linear extended state observers, nonlinear dynamic inverse and time-varying dynamic model. The parameter optimization module is used to construct the total cost function based on the total thrust and control torque, and to iteratively optimize the candidate solution vector constructed from the controller parameters of the trajectory tracking controller based on the particle swarm optimization algorithm to obtain the real-time target controller parameters. The trajectory tracking module is used to dynamically adjust the trajectory tracking controller based on the real-time target controller parameters in order to continuously track the trajectory of the mother-daughter UAV during the dynamic release of the daughter unit.

[0013] The beneficial effects of the present invention are as follows: The trajectory tracking control system provided by the present invention is a system corresponding to the above-mentioned trajectory tracking control method for aerial release of mother and daughter UAVs. It is used to realize continuous and accurate trajectory tracking control of mother and daughter UAVs, including aerial release daughter UAVs, under the above-mentioned trajectory tracking control method for aerial release of mother and daughter UAVs.

[0014] Other advantages of the present invention will be analyzed in more detail in the following embodiments. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating the steps of a trajectory tracking control method for aerial release of a mother-daughter UAV in an embodiment of the present invention.

[0017] Figure 2 This is a block diagram of a trajectory tracking control system based on a trajectory tracking control method for aerial release of a mother-daughter UAV, according to an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0019] In this scheme, to accurately describe the motion of the mother UAV in three-dimensional space within a mother-daughter UAV, two coordinate systems are defined: a ground inertial coordinate system and a ground-based inertial coordinate system. and the mother machine body coordinate system ; Ground inertial coordinate system Fixed to the ground, it describes the global position and trajectory of the mother-daughter UAV, using the North-Earth-East (NED) convention, where, For the preset fixed position on the ground, The axis points due north. The axis points due east. The axis points towards the Earth's center; Mother machine body coordinate system Fixed to the mainframe, using a front-right-bottom configuration, wherein... Located at the geometric center of the mother machine, The shaft points towards the head of the mother machine. The shaft points to the right side of the mother machine. Pointing downwards from the mother machine; the angular velocity and control torque of the mother machine are defined in the mother machine's body coordinate system.

[0020] In this scheme, without considering the loading and unloading of the submachine, the mother machine can be regarded as a standard six-degree-of-freedom (6-DOF) rigid body. Based on the dynamic model of the mother machine, the state vector of the mother machine is defined as follows: , in, Indicates the mother machine in the ground inertial coordinate system The lower position, Indicates the mother machine in the ground inertial coordinate system The speed of the drop Represents Euler angles. This represents the angular velocity of the mother machine, where the Euler angle includes the roll angle. Pitch angle and yaw angle ; The dynamic model of the machine tool includes translational kinematics and rotational kinematics. The calculation expression of the dynamic model of the machine tool is as follows: , , in, Indicates the mother machine in the ground inertial coordinate system The derivative at the lower position, This indicates the coordinate system of the mother machine in the mother machine body coordinate system. The transformation matrix from angular velocity to Euler angle rate of change; Based on the dynamic model of the mother machine, the dynamic equations of the mother machine can be constructed using the Newton-Euler method: , , in, Indicates the mother machine in the ground inertial coordinate system The acceleration below, Indicates the mass of the mother machine. This represents the rotation matrix from the mothership's body coordinate system to the ground inertial coordinate system. This represents the total thrust generated by the rotor. Represents the gravitational acceleration vector. Indicates external disturbance force. This represents the angular acceleration of the mother machine. The matrix representing the inverse of the mother machine's moment of inertia matrix. This represents the control torque generated by the rotor. The matrix representing the moment of inertia of the mother machine; like Figure 1 As shown, in one embodiment of the present invention, the present invention provides a trajectory tracking control method for mid-air release of a mother-daughter unmanned aerial vehicle, comprising the following steps: S1. Based on the release state of the mother-daughter UAV, time-varying parameters are calculated in real time, and a time-varying dynamic model is constructed. This scheme establishes a six-degree-of-freedom (6-DOF) nonlinear time-varying aerodynamic model that accurately describes the entire process of mother aircraft release in a mother-daughter UAV, and updates the system physical properties required for control in real time.

[0021] S1 includes the following steps: S11. Based on the release state of the mother drone, set the release state vector of the mother drone with respect to the daughter drone; In this scheme, the release status of the mother drone in a mother-daughter drone system can reflect whether the daughter drones carried by the mother drone have been released.

[0022] S12. Based on the release state vector of the mother machine with respect to the daughter machine, the time-varying parameters are calculated in real time, including the total mass, total center of mass and total moment of inertia matrix. The calculation expression for the time-varying parameter in S12 is as follows: , , , , , , in, express The total mass of the mother machine and all unreleased daughter machines at that moment. Indicates the mass of the mother machine. This indicates the total number of slave machines. express The mother machine of the moment about the first Release state vector of each submachine Indicates the first The quality of the individual machine Indicates belonging to, express The position of the center of mass of the mother machine and all unreleased daughter machines at any given moment. Indicates the position of the mother machine's center of gravity. Indicates the first The position of the center of gravity of the machine express The total moment of inertia of the mother machine and all unreleased slave machines at any given moment. This represents the moment of inertia of the mother machine about its own center of mass. The antisymmetric matrix representing the translation vector of the mother machine. Indicates the first The moment of inertia of a single machine about its own center of mass. Indicates the first The antisymmetric matrix of the translation vectors of each submachine. Describes the antisymmetric matrix of a matrix. This represents the translation vector of the mother machine. Indicates the first The translation vector of each submachine, where, when At that time, The first in the mother machine The machine was not released when At that time, The first in the mother machine The individual machine has been released; In this scheme, the total mass in the time-varying parameters is defined as the sum of the mass of the mother unit and the unreleased daughter units; the instantaneous total center of mass of the mother unit and daughter units reflects the overall rotational dynamics of the mother-daughter UAV. This scheme places the mother unit and all unreleased daughter units in the mother unit's body coordinate system. The weighted average position is used as the total centroid. In this scheme, according to the parallel axis theorem and the principle of rigid body superposition, the total moment of inertia is obtained around the instantaneous total centroid. The calculation of the total moment of inertia takes into account the contribution of the mother machine itself and the contribution of all unreleased daughter machines. Since the centroid of the daughter machine is non-zero offset from the total centroid of the mother-daughter UAV, the introduced cross inertia term makes the total moment of inertia a non-diagonal, dense and time-varying inertia matrix. The real-time update of the total moment of inertia is the basis for the attitude loop nonlinear dynamic inverse (NDI) compensation of the gyroscope coupling torque. S13. Based on the total centroid in the time-varying parameters, and considering the centroid shift caused by the release of the slave unit by the mother machine and the rotor thrust, the quasi-static disturbance torque and transient impact disturbance are calculated. The transient impact disturbance includes net recoil force and net recoil torque. The calculation expressions for the quasi-static disturbance moment and transient impact disturbance are as follows: , , , in, This represents the quasi-static disturbance moment. This represents the total thrust generated by the rotor. This represents the net recoil force exerted on the center of mass when the submachine gun is released. This refers to the release mechanism of the slave unit on the mother machine. Indicates that the mother machine passes through the first The reverse impact force experienced by the release mechanism when releasing the submachine. Indicates the net recoil torque. Indicates the first The center of mass position of the release mechanism; In this scheme, the quasi-static disturbance torque is the torque generated by the combined action of the center of mass offset and the rotor thrust. After the slave unit is released, the total center of mass of the mother-daughter UAV no longer coincides with the geometric center of the body. The resulting eccentricity forms a lever arm, thereby generating a continuous disturbance torque that is proportional to the current thrust, i.e., the quasi-static disturbance torque. S14. Define aerodynamic interference force and aerodynamic interference torque based on the flow field disturbed at the moment of release of the slave unit; In this scheme, the sudden separation of the sub-unit will disturb the flow field and generate complex aerodynamic interference forces and aerodynamic interference torques. Since the flow field induced by the instantaneous separation of the sub-unit is extremely complex and strongly coupled with the flight state, this scheme treats the disturbance as a bounded unknown disturbance, which can be suppressed by the controller.

[0023] S15, Based on the mother machine body coordinate system The net recoil force, aerodynamic disturbance force, and external disturbance force are combined to construct a composite disturbance force. The calculation expression for the composite disturbance force is as follows: , in, Indicates the combined disturbance force. This represents the rotation matrix from the mothership's body coordinate system to the ground inertial coordinate system. Indicates aerodynamic interference force. This represents the external disturbance force; in this scheme, the external disturbance force refers to disturbance forces such as wind disturbance that are not directly modeled. S16. Based on Newton-Euler's equations, the time-varying inertia... Differentiating the derivative yields the impact torque caused by the sudden change in rotational inertia. S17, Based on the mother machine body coordinate system The quasi-static disturbance torque, net recoil torque, aerodynamic disturbance torque, and moment of inertia change impact torque are used to construct a composite disturbance torque; The calculation expression for the composite disturbance moment is as follows: , in, Indicates the combined disturbance torque. Indicates aerodynamic disturbance torque. This represents the impact torque caused by a sudden change in rotational inertia. S18. Construct a time-varying dynamic model based on composite perturbation force and composite perturbation moment; The calculation expression for the time-varying dynamic model is as follows: , , in, This indicates the acceleration of the mother machine. This represents the total thrust generated by the rotor. Represents the gravitational acceleration vector. Indicates centripetal acceleration. This represents the angular acceleration of the mother machine. Represents the coordinate system of the mother machine. The three-axis control torque below, This represents the angular velocity of the mother machine. This represents the angular velocity matrix of the mother machine.

[0024] In this scheme, a robust controller is designed to construct a time-varying dynamic model. By calculating the total thrust and three-axis control torque, the mother drone can accurately track the given desired trajectory throughout the entire deployment mission and quickly suppress various complex disturbances at the moment of deployment, thereby ensuring the smooth, safe and high-precision flight of the mother-daughter UAV.

[0025] S2. Construct a position third-order linear extended state observer and three attitude second-order linear extended state observers with parallel channels respectively. In this scheme, a linear extended state observer with dual position and attitude channels is used to predict and compensate for all remaining position dynamics and external disturbances that are difficult to model accurately, i.e., lumped disturbances, in real time. This scheme expands the lumped disturbances into new state variables, thereby transforming the complex position disturbances into the problem of predicting observable states. The observer gain of the extended state observer is determined by the bandwidth parameterization method.

[0026] S2 includes the following steps: S21. Based on the second-order integrator system of translational dynamics reconstruction, a third-order linear extended state observer is constructed, wherein the third-order linear extended state observer is used to predict the lumped disturbance acceleration in real time. In this scheme, the lumped disturbance acceleration includes all unmodeled and unpredictable translational disturbance terms; In this scheme, the position third-order linear extended state observer is in axis, shaft and The axes can be completely decoupled, and the corresponding calculation expressions have the same structure in each axis direction. In this embodiment, we take... Taking the shaft as an example, then The calculation expression for the third-order linear extended state observer in the axial direction is as follows: , in, This indicates that the mother machine and all unreleased slave machines are in The derivative of the predicted position along the axis with respect to time. This indicates that the mother machine and all unreleased slave machines are in Predicted value of axial velocity. This represents the bandwidth of the third-order linearly extended state observer at the location. This indicates that the mother machine and all unreleased slave machines are in The actual position in the axial direction, This indicates that the mother machine and all unreleased slave machines are in Predicted value of axial position. This indicates that the mother machine and all unreleased slave machines are in The derivative of the predicted axial velocity with respect to time. This indicates that the mother machine and all unreleased slave machines are in Predicted values ​​of lumped disturbance acceleration in the axial direction. This indicates that the mother machine and all unreleased slave machines are in Thrust acceleration in the axial direction, This indicates that the mother machine and all unreleased slave machines are in Known acceleration in the axial direction, Indicates the observer bandwidth. This indicates that the mother machine and all unreleased slave machines are in The time derivative of the predicted disturbance value in the axial direction; In this scheme, both thrust acceleration and known acceleration are control terms. Thrust acceleration is calculated by the controller, and known acceleration is a nonlinear dynamic inverse (NDI) compensation term. Known acceleration is the sum of the known gravitational acceleration of the mother machine and the unreleased daughter machine and the centripetal acceleration caused by the total center of mass shift. S22. Based on the first-order integrator system of rotational dynamics reconstruction, construct three parallel second-order linear extended state observers for attitude, wherein the three parallel second-order linear extended state observers for attitude are used to predict lumped rotational disturbances in real time. In this scheme, the three attitude second-order linear extended state observers in parallel channels are set with attitude angles in the three directions corresponding to the attitude; the predicted value of the lumped rotational disturbance includes all the remaining dynamics that have not been compensated by the nonlinear dynamic inverse (NDI), mainly including: external disturbance torque and internal dynamic transient term, wherein the external disturbance torque is aerodynamic disturbance torque and mechanical recoil torque, and the internal dynamic transient term is variable inertia impact torque, wherein the mechanical recoil torque is the net recoil torque; The calculation expression for the attitude second-order linear extended state observer is as follows: , in, Indicates the first Predicted angular acceleration values ​​for each channel. Indicates the first Predicted values ​​of lumped disturbances for each channel. This represents the bandwidth of the attitude second-order linearly extended state observer. Indicates the first The angular velocity of each channel, Indicates the first Predicted values ​​of angular velocity for each channel. Indicates the first The rate of change of the predicted value of the lumped disturbance for each channel, where, =1, 2, 3.

[0027] S3. Based on the time-varying dynamics model and the third-order linear extended state observer, the composite control law is derived and the thrust is calculated to obtain the total thrust and the desired attitude. This proposal suggests a trajectory tracking controller based on Nonlinear Dynamic Inverse (NDI), Linear Extended State Observer (LESO), and cascade feedback. The controller employs a cascaded structure of an outer position loop and an inner attitude loop. It uses the LESO to estimate and compensate for concentrated disturbances in real time to achieve high-precision and robust tracking. In this proposal, the active feedforward characteristics of the nonlinear dynamic inverse are used to accurately compensate for time-varying parameters. Similarly, the observation capability of the LESO is utilized to estimate and suppress composite disturbances in real time. S3 includes the following steps: S31. Based on the time-varying dynamic model, the translational dynamic model is decoupled with acceleration as the core to obtain the decoupled translational dynamic model of the position controller; in this scheme, according to the time-varying dynamic model, the calculation expression of the translational dynamic model can be obtained as follows: ; The calculation expression for the decoupled translational dynamics model of the position controller is as follows: , , , , in, Indicates thrust acceleration. Indicates a known acceleration. Represents the lumped disturbance acceleration. It is defined as; In this scheme, thrust acceleration is the control term, which is solved by the composite trajectory tracking controller; known acceleration is a nonlinear dynamic inverse compensation term, including known gravitational acceleration and acceleration that can be calculated in real time based on the current angular velocity and total centroid of the mother machine; lumped disturbance acceleration is the prediction term of the extended state observer, which includes all unmodeled and difficult-to-predict disturbances, and is the object that the extended state observer needs to identify and supplement.

[0028] S32. Calculate the desired speed reference command based on the position error and integral error according to the outer position loop; The calculation expression for the desired speed reference command is as follows: , in, Indicates positional error. Indicates the mother machine in the ground inertial coordinate system The lower position, Indicates the trajectory position command. Indicates the integral error. express Position error at any given time This represents the result of the time derivative. Indicates the desired speed reference command. The adjustment factor representing the positive definite diagonal gain at position. Indicates the mother machine in the ground inertial coordinate system The speed of the drop This represents the position positive definite diagonal gain matrix. This represents the integral positive definite diagonal gain matrix; S33. Based on the desired speed reference command, use the speed loop to calculate the ideal feedback acceleration based on the speed tracking error; The expression for calculating the ideal feedback acceleration is as follows: , , in, Represents ideal feedback acceleration. Indicates the dynamic speed error. This represents the positive definite diagonal gain matrix of the velocity. Indicates speed error; S34. Based on the nonlinear dynamic inverse and ideal feedback acceleration, the desired thrust acceleration is calculated using the decoupled dynamic model of the position controller. In this scheme, the actual acceleration of the mother machine is used. The ideal acceleration command issued by the feedback controller is accurately tracked, and the control term, namely thrust acceleration, is obtained by inverse solving based on the decoupled dynamic model. The unmeasurable lumped disturbance acceleration is replaced by the predicted lumped disturbance acceleration obtained through the position third-order linear extended state observer to obtain the desired thrust acceleration. The expression for calculating the desired thrust acceleration is as follows: , in, This represents the desired thrust acceleration. This represents the predicted value of the lumped disturbance acceleration; S35. Calculate the desired thrust vector based on the desired thrust acceleration and the real-time total mass in the time-varying parameters. The expression for calculating the desired thrust vector is as follows: , in, Represents the desired thrust vector; S36. Decompose the desired thrust vector to obtain the desired total thrust magnitude and desired attitude; The calculation expressions for the desired total thrust amplitude and desired attitude are as follows: , , in, This indicates the expected total thrust. This represents the expected total thrust amplitude. It expresses a desired attitude.

[0029] S4. Calculate the command angular acceleration based on the desired attitude, define the error signal, and calculate the control torque based on three parallel attitude second-order linear extended state observers, nonlinear dynamic inverse and time-varying dynamic model. This scheme designs a composite control law in the attitude inner loop to quickly and accurately track the desired attitude calculated by the position outer loop; S4 includes the following steps: S41. Differentiate the desired attitude that changes over time to calculate the command angular velocity, and then calculate the command angular acceleration by filtering and differentiating the command angular velocity. In this scheme, to avoid numerical noise, a low-pass filter or trajectory planning is used to generate a smooth command angular acceleration. S42. Define the error signal, which includes attitude error and angular velocity tracking error; In this scheme, when the rigid body attitude is represented by a unit quaternion, the attitude error can be extracted from the vector part of the error quaternion. The error quaternion can be obtained from the given desired attitude quaternion and the current attitude quaternion, i.e. ,in, Represents the error quaternion. Represents the expected pose quaternion. To represent quaternion multiplication, The quaternion represents the conjugate of the current attitude quaternion; the attitude error can accurately characterize the proportional relationship between the minimum rotation axis and the amount of rotation required to rotate from the current attitude to the desired attitude; if the error quaternion is represented as ,in, Indicates the rotation angle. If the unit rotation axis is represented, then the attitude error... When the attitude error is small, that is hour, It is approximately linearly related to the rotation vector, which facilitates controller design.

[0030] The expression for calculating angular velocity tracking error is: ,in, Indicates angular velocity tracking error. This indicates the commanded angular velocity; in this scheme, the angular velocity tracking control error directly reflects the dynamic tracking performance. S43. Construct the target angular acceleration based on the command angular acceleration and error signal; The expression for calculating the target angular acceleration is as follows: , in, Indicates the target angular acceleration. Indicates the command angular acceleration. Represents the angular acceleration gain matrix. Indicates angular velocity tracking error. The adjustment factor representing the attitude gain. Represents the attitude gain matrix. Indicates attitude error; S44. Based on the target angular acceleration, the controller input angular acceleration of each channel is calculated using three parallel second-order linear extended state observers. The calculation expressions for the controller input angular acceleration of each channel are as follows: , in, Indicates the first The controller input angular acceleration for each channel, Indicates the first The target desired angular acceleration for each channel; S45. Based on the controller input angular acceleration, target angular acceleration and time-varying dynamic model of each channel, the control torque is calculated based on the nonlinear dynamic inverse and rotational dynamic model. In this scheme, based on the time-varying dynamics model, the calculation expression for the translational dynamics model can be obtained as follows: ; The formula for calculating the control torque is as follows: , in, This represents the predicted value of the lumped disturbance. Indicates a known disturbance torque; In this plan, As a feedforward of expectation, to compensate for the angular acceleration of the expected settling; As proportional-derivative feedback, it is used to eliminate attitude and angular velocity errors; As a known disturbance feedforward, it actively cancels the modelable disturbance torque; ESO compensation is a crucial part of disturbance mitigation, used to estimate and suppress all unknown disturbance moments. The predicted value of the lumped disturbance is the prediction of the unknown disturbance by the three parallel attitude second-order linear extended state observers. The known disturbance moments include all moments that can be calculated in real time based on the current state and parameters, mainly including gyro-coupled moments. and the quasi-static disturbance moment generated by the centroid shift Unknown disturbance torques include all disturbance torques that are difficult to model accurately, such as net recoil torque, aerodynamic disturbance torque, and sudden change in moment of inertia impact torque. In this scheme, based on the aforementioned composite control law, the attitude loop is dynamically linearized into a proportional-differential system that is only affected by bounded estimation errors.

[0031] S5. Construct a total cost function based on the total thrust and control torque, and iteratively optimize the candidate solution vector constructed from the controller parameters of the trajectory tracking controller using the particle swarm optimization algorithm to obtain the real-time target controller parameters; This scheme transforms controller parameter tuning into a multi-dimensional, constrained nonlinear optimization problem, and introduces particle swarm optimization algorithm for automatic optimization. To overcome the limitations of traditional inner and outer loop decoupling optimization strategies, this scheme adopts a position-attitude joint optimization strategy.

[0032] S5 includes the following steps: S51. Construct the total cost function based on the total thrust and control torque; The cost function designed in this scheme aims to balance trajectory tracking performance and control energy consumption, and can quantify multi-dimensional performance indicators, such as tracking error, overshoot, and energy consumption, into a single scalar cost value. The total cost function is calculated as follows: , , , , , in, Represents particles Total cost Indicates instantaneous performance cost. Indicates instantaneous energy consumption cost. Indicates the end time of the mother machine release control. Indicates the start time of the mother machine release control. Indicates the derivative with respect to time. Indicates the basic weight of the location. Indicates the first One axis, express Axial direction, express Axial direction, express Axial direction, Indicates the first Axial weights for each axis, Indicates the first The absolute value of the axial position error in each axis. Indicates pose weights, Indicates the first One attitude angle, Indicates the roll angle. Indicates pitch angle, Indicates the yaw angle. Indicates the first The attitude weights for each attitude angle. Indicates the first The absolute value of the attitude error for each attitude angle. Indicates the global energy consumption weight. Indicates thrust weight. Indicates total thrust. Indicates the first Weighting of the control torque along each axis Indicates the first Control torque in each axis; In this scheme, instantaneous performance cost is used to penalize the controller's tracking errors in position and attitude; attitude weight is used to balance the dimensions to balance the contributions of position error and attitude error in the same cost function; instantaneous energy cost is used to penalize excessive control output, reflecting consideration of system energy consumption and actuator losses, where actuators include controllable components such as motors. S52. Based on the physical symmetry of the mother-daughter UAV in the XY plane, construct the candidate solution vector of the trajectory tracking controller based on the controller parameters of the trajectory tracking controller; In this scheme, based on the physical symmetry of the mother-daughter UAV in the XY plane, the same controller parameters of the X and Y axes in the outer position loop and the inner attitude loop are merged, so that they are assigned the same value during the optimization process, while the Z axis parameter ( Then, it is optimized independently; in this scheme, the candidate solution vector of the trajectory tracking controller is a 16-dimensional vector composed of 16 controller parameters to be optimized, which are shown in Table 1: Table 1 Controller Parameter Table In this plan, , , , , , and In respectively axis, shaft and The axis has corresponding values, because shaft and If the axes have the same symmetry, they are combined into a single value, while the Z-axis has an independent value. and It is a scalar adjustment factor for the gain, and it has only one value; S53. Treat each candidate solution vector of the trajectory tracking controller as a particle. ; S54. Based on the particle swarm optimization algorithm, the optimal solution vector of the real-time trajectory tracking controller is obtained by continuous iteration with the total cost function as the objective function. S54 includes the following steps: S541. Based on the particle swarm algorithm and the parameter space of the candidate solutions of the trajectory tracking controller, initialize the particle swarm and the current iteration number to 0. In this scheme, each potential solution of the trajectory tracking controller is regarded as a particle. Each particle has two core attributes in the search space: position and velocity. Position represents a candidate solution in the search space, i.e., a set of controller parameters of the trajectory tracking controller to be optimized. Velocity represents the direction and rate of change of the particle in the search space. During the search process, each particle dynamically adjusts its flight based on its own experience and the experience of other particles in the swarm. This adjustment depends on two key optimal values: the individual historical optimal value and the global historical optimal value. The individual historical optimal value refers to the best position experienced by each particle itself, and the global historical optimal value refers to the best position found by the entire particle swarm.

[0033] S542. Based on the position and velocity of each particle in the particle swarm, execute the mother machine release control program respectively, and extract the three axial position errors, the three attitude angle errors, the total thrust and the three axial control torques as the simulation output of the particle. S543. Based on the simulation output of each particle, evaluate the total cost of each particle in the particle swarm based on the total cost function. S544. Based on the total cost of each particle, update the individual historical best value and the global historical best value in the particle swarm. S545. Update the velocity and position of each particle based on the individual historical best value in the particle swarm and the global historical best value. In this scheme, the velocity and position of each particle are updated through a position and velocity update model; The calculation expression for the position and velocity update model is as follows: , , in, Indicates the updated number The position of each particle. Indicates the current number The position of each particle. Indicates the updated number The speed of each particle Indicates inertia weight, Indicates the current number The speed of each particle The learning factor represents the individual's historical best value. This represents the randomness coefficient of an individual's search. Indicates the current number The individual historical best value of each particle. This represents the learning factor for the global historical optimum. Indicates the random coefficient of the global search. Indicates the global historical best value; S546. Determine whether the current iteration count has reached the iteration count threshold. If yes, proceed to S547; otherwise, increment the current particle iteration count by 1 and return to S542. S547. The candidate solution vector corresponding to the global historical best value is used as the optimal solution vector of the trajectory tracking controller.

[0034] S55. Use the controller parameters corresponding to the optimal solution vector of the real-time trajectory tracking controller as the real-time target controller parameters.

[0035] S6. The trajectory tracking controller is dynamically adjusted based on the real-time target controller parameters to continuously track the trajectory of the mother-daughter UAV during the dynamic release of the daughter unit.

[0036] like Figure 2 As shown, on the other hand, the present invention also provides a trajectory tracking control system based on the above-described trajectory tracking control method for aerial release of mother-daughter UAVs, comprising: The parameter update module is used to calculate time-varying parameters in real time based on the release state of the mother-daughter UAV and to build a time-varying dynamic model. The channel expansion observer module is used to construct a position third-order linear expansion state observer and three attitude second-order linear expansion state observers in parallel channels. The outer loop module is used to derive the composite control law and calculate the thrust based on the time-varying dynamics model and the third-order linear extended state observer of the position, so as to obtain the total thrust and the desired attitude. The attitude inner loop module is used to calculate the command angular acceleration based on the desired attitude and define the error signal. It also calculates the control torque based on three parallel attitude second-order linear extended state observers, nonlinear dynamic inverse and time-varying dynamic model. The parameter optimization module is used to construct the total cost function based on the total thrust and control torque, and to iteratively optimize the candidate solution vector constructed from the controller parameters of the trajectory tracking controller based on the particle swarm optimization algorithm to obtain the real-time target controller parameters. The trajectory tracking module is used to dynamically adjust the trajectory tracking controller based on the real-time target controller parameters in order to continuously track the trajectory of the mother-daughter UAV during the dynamic release of the daughter unit.

[0037] This solution uses multi-scenario, progressive simulation experiments, including adverse conditions such as high dynamic trajectory, 1.5kg model mismatch, transient impact, and continuous random noise, to verify that the composite control strategy can still maintain the trajectory tracking accuracy and closed-loop stability of the mother-daughter UAV in the air under extreme conditions, fully demonstrating its high bandwidth disturbance suppression capability and engineering scalability.

[0038] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A trajectory tracking control method for sub-munition UAV air release, characterized in that, Includes the following steps: S1. Based on the release state of the mother-daughter UAV, time-varying parameters are calculated in real time, and a time-varying dynamic model is constructed. S2. Construct a position third-order linear extended state observer and three attitude second-order linear extended state observers with parallel channels respectively. S3. Based on the time-varying dynamics model and the third-order linear extended state observer, the composite control law is derived and the thrust is calculated to obtain the total thrust and the desired attitude. S4. Calculate the command angular acceleration based on the desired attitude, define the error signal, and calculate the control torque based on three parallel attitude second-order linear extended state observers, nonlinear dynamic inverse and time-varying dynamic model. S5. Construct a total cost function based on the total thrust and control torque, and iteratively optimize the candidate solution vector constructed from the controller parameters of the trajectory tracking controller using the particle swarm optimization algorithm to obtain the real-time target controller parameters; S6. The trajectory tracking controller is dynamically adjusted based on the real-time target controller parameters to continuously track the trajectory of the mother-daughter UAV during the dynamic release of the daughter unit.

2. The trajectory tracking control method for sub-parent drone aerial release according to claim 1, wherein, S1 includes the following steps: S11. Based on the release state of the mother drone, set the release state vector of the mother drone with respect to the daughter drone; S12. Based on the release state vector of the mother machine with respect to the daughter machine, the time-varying parameters are calculated in real time, including the total mass, total center of mass and total moment of inertia matrix. The calculation expression for the time-varying parameter in S12 is as follows: ， ， ， ， ， ， in, express The total mass of the mother machine and all unreleased daughter machines at that moment. Indicates the mass of the mother machine. This indicates the total number of slave machines. express The mother machine of the moment about the first Release state vector of each submachine Indicates the first The quality of the individual machine Indicates belonging to, express The position of the center of mass of the mother machine and all unreleased daughter machines at any given moment. Indicates the position of the mother machine's center of gravity. Indicates the first The position of the center of gravity of the machine express The total moment of inertia of the mother machine and all unreleased slave machines at any given moment. This represents the moment of inertia of the mother machine about its own center of mass. The antisymmetric matrix representing the translation vector of the mother machine. Indicates the first The moment of inertia of a single machine about its own center of mass. Indicates the first The antisymmetric matrix of the translation vectors of each submachine. Describes the antisymmetric matrix of a matrix. This represents the translation vector of the mother machine. Indicates the first The translation vector of each submachine, where, when At that time, The first machine in the time machine The machine was not released when At that time, The first machine in the time machine The individual machine has been released; S13. Based on the total centroid in the time-varying parameters, and considering the centroid shift caused by the release of the slave unit by the mother machine and the rotor thrust, the quasi-static disturbance torque and transient impact disturbance are calculated. The transient impact disturbance includes net recoil force and net recoil torque. The calculation expressions for the quasi-static disturbance moment and transient impact disturbance are as follows: ， ， ， wherein, represents the quasi-static disturbance moment, represents the total thrust generated by the rotor, represents the net recoil force on the total center of mass when a submachine is released, represents the release mechanism of the submachine on the mother machine, represents the counter-impact force received by the mother machine when releasing the submachine through the first release mechanism, represents the net recoil moment, represents the center of mass position of the first release mechanism; S14. Define aerodynamic interference force and aerodynamic interference torque based on the flow field disturbed at the moment of release of the slave unit; S15, based on the parent machine coordinate system under the net recoil force, aerodynamic interference force and external disturbance force, construct the composite disturbance force; The calculation expression for the composite disturbance force is as follows: ， wherein, represents a complex disturbance force, represents a rotation matrix from the parent aircraft body coordinate system to the ground inertial coordinate system, represents an aerodynamic interference force, represents an external disturbance force; S16, time-varying inertia according to Newton Euler equation derivation, get the moment of inertia mutation impact torque; S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system S17, based on the parent machine coordinate system The calculation expression for the composite disturbance moment is as follows: ， wherein, represents a compound disturbance torque, represents an aerodynamic interference torque, represents a moment of inertia jump impact torque; S18. Construct a time-varying dynamic model based on composite perturbation force and composite perturbation moment; The calculation expression for the time-varying dynamic model is as follows: ， ， in, This indicates the acceleration of the mother machine. This represents the total thrust generated by the rotor. Represents the gravitational acceleration vector. Indicates centripetal acceleration. This represents the angular acceleration of the mother machine. Represents the coordinate system of the mother machine. The three-axis control torque below, This represents the angular velocity of the mother machine. This represents the angular velocity matrix of the mother machine.

3. The trajectory tracking control method for sub-parent drone aerial release according to claim 2, characterized in that, S2 includes the following steps: S21. Based on the second-order integrator system of translational dynamics reconstruction, a third-order linear extended state observer is constructed, wherein the third-order linear extended state observer is used to predict the lumped disturbance acceleration in real time. S22. Based on the first-order integrator system of rotational dynamics reconstruction, construct three parallel second-order linear extended state observers for attitude, wherein the three parallel second-order linear extended state observers for attitude are used to predict lumped rotational disturbances in real time. The calculation expression for the attitude second-order linear extended state observer is as follows: ， wherein denotes the predicted value of the angular acceleration of the denotes the predicted value of the lumped disturbance of the denotes the bandwidth of the attitude second order linear extended state observer, denotes the angular velocity of the denotes the predicted value of the angular velocity of the denotes the rate of change of the predicted value of the lumped disturbance of the = 1,2,3.​​​​​ 4. The trajectory tracking control method for sub-parent drone aerial release according to claim 3, wherein, S3 includes the following steps: S31. Based on the time-varying dynamic model, the translational dynamic model is decoupled with acceleration as the core to obtain the decoupled translational dynamic model of the position controller. The calculation expression for the decoupled translational dynamics model of the position controller is as follows: ， ， ， ， wherein, represents the thrust acceleration, represents the known acceleration, represents the lumped disturbance acceleration, represents is defined as; S32. Calculate the desired speed reference command based on the position error and integral error according to the outer position loop; The calculation expression for the desired speed reference command is as follows: ， wherein, represents a position error, represents a position of the mother machine in a ground inertial coordinate system , represents a trajectory position command, represents an integral error, represents a position error at a time instant, represents a time derivative result, represents a desired velocity reference command, represents a tuning factor for a positive definite diagonal gain of position, represents a velocity of the mother machine in a ground inertial coordinate system , represents a positive definite diagonal gain matrix of position, represents a positive definite diagonal gain matrix of integral; S33. Based on the desired speed reference command, use the speed loop to calculate the ideal feedback acceleration based on the speed tracking error; The expression for calculating the ideal feedback acceleration is as follows: ， ， wherein, represents ideal feedback acceleration, represents velocity error dynamics, represents velocity positive definite diagonal gain matrix, represents velocity error; S34. Based on the nonlinear dynamic inverse and ideal feedback acceleration, the desired thrust acceleration is calculated using the decoupled dynamic model of the position controller. The expression for calculating the desired thrust acceleration is as follows: ， wherein, represents the desired thrust acceleration, represents the lumped disturbance acceleration prediction; S35. Calculate the desired thrust vector based on the desired thrust acceleration and the real-time total mass in the time-varying parameters. The expression for calculating the desired thrust vector is as follows: ， wherein, represents the desired thrust vector; S36. Decompose the desired thrust vector to obtain the desired total thrust magnitude and desired attitude; The calculation expressions for the desired total thrust amplitude and desired attitude are as follows: ， ， wherein, represents the desired total thrust magnitude, represents the desired total thrust magnitude, represents the desired attitude.

5. The trajectory tracking control method for sub-parent drone aerial release according to claim 4, characterized in that, S4 includes the following steps: S41. Differentiate the desired attitude that changes with time to calculate the command angular velocity, and then calculate the command angular acceleration by filtering and differentiating the command angular velocity. S42. Define the error signal, which includes attitude error and angular velocity tracking error; S43. Construct the target angular acceleration based on the command angular acceleration and error signal; The expression for calculating the target angular acceleration is as follows: ， wherein denotes a target angular acceleration, denotes a command angular acceleration, denotes an angular acceleration gain matrix, denotes an angular velocity tracking error, denotes a tuning factor for the attitude gain, denotes an attitude gain matrix, denotes an attitude error; S44. Based on the target angular acceleration, the controller input angular acceleration of each channel is calculated using three parallel second-order linear extended state observers. The calculation expressions for the controller input angular acceleration of each channel are as follows: ， wherein, represents the controller input angular acceleration of the represents the target desired angular acceleration of the channel.​ S45. Based on the controller input angular acceleration, target angular acceleration and time-varying dynamic model of each channel, the control torque is calculated based on the nonlinear dynamic inverse and rotational dynamic model. The formula for calculating the control torque is as follows: ， wherein, represents a predicted value of the collective disturbance, represents a known disturbance torque.

6. The trajectory tracking control method for sub-parent drone aerial release according to claim 5, wherein, S5 includes the following steps: S51. Construct the total cost function based on the total thrust and control torque; The total cost function is calculated as follows: ， ， ， ， ， in, Represents particles Total cost Indicates instantaneous performance cost. Indicates instantaneous energy consumption cost. Indicates the end time of the mother machine release control. Indicates the start time of the mother machine release control. Indicates the derivative with respect to time. Indicates the basic weight of the location. Indicates the first One axis, express Axial direction, express Axial direction, express Axial direction, Indicates the first Axial weights for each axis, Indicates the first The absolute value of the axial position error in each axis. Indicates pose weights, Indicates the first One attitude angle, Indicates the roll angle. Indicates pitch angle, Indicates the yaw angle. Indicates the first The attitude weights for each attitude angle. Indicates the first The absolute value of the attitude error for each attitude angle. Indicates the global energy consumption weight. Indicates thrust weight. Indicates total thrust. Indicates the first Weighting of the control torque along each axis Indicates the first Control torque in each axis; S52. Based on the physical symmetry of the mother-daughter UAV in the XY plane, construct the candidate solution vector of the trajectory tracking controller based on the controller parameters of the trajectory tracking controller; S53. Treat each candidate solution vector of the trajectory tracking controller as a particle. ; S54. Based on the particle swarm optimization algorithm, the optimal solution vector of the real-time trajectory tracking controller is obtained by continuous iteration with the total cost function as the objective function. S55. Use the controller parameters corresponding to the optimal solution vector of the real-time trajectory tracking controller as the real-time target controller parameters.

7. The trajectory tracking control method for in-flight release of a mother-daughter UAV according to claim 6, characterized in that, S54 includes the following steps: S541. Based on the particle swarm algorithm and the parameter space of the candidate solutions of the trajectory tracking controller, initialize the particle swarm and the current iteration number to 0. S542. Based on the position and velocity of each particle in the particle swarm, execute the mother machine release control program respectively, and extract the three axial position errors, the three attitude angle errors, the total thrust and the three axial control torques as the simulation output of the particle. S543. Based on the simulation output of each particle, evaluate the total cost of each particle in the particle swarm based on the total cost function. S544. Based on the total cost of each particle, update the individual historical best value and the global historical best value in the particle swarm. S545. Update the velocity and position of each particle based on the individual historical best value in the particle swarm and the global historical best value. S546. Determine whether the current iteration count has reached the iteration count threshold. If yes, proceed to S547; otherwise, increment the current particle iteration count by 1 and return to S542. S547. The candidate solution vector corresponding to the global historical best value is used as the optimal solution vector of the trajectory tracking controller.

8. A trajectory tracking control system based on the trajectory tracking control method for in-flight release of a mother-daughter UAV as described in any one of claims 1-7, characterized in that, include: The parameter update module is used to calculate time-varying parameters in real time based on the release state of the mother-daughter UAV and to build a time-varying dynamic model. The channel expansion observer module is used to construct a position third-order linear expansion state observer and three attitude second-order linear expansion state observers in parallel channels. The outer loop module is used to derive the composite control law and calculate the thrust based on the time-varying dynamics model and the third-order linear extended state observer of the position, so as to obtain the total thrust and the desired attitude. The attitude inner loop module is used to calculate the command angular acceleration based on the desired attitude and define the error signal. It also calculates the control torque based on three parallel attitude second-order linear extended state observers, nonlinear dynamic inverse and time-varying dynamic model. The parameter optimization module is used to construct the total cost function based on the total thrust and control torque, and to iteratively optimize the candidate solution vector constructed from the controller parameters of the trajectory tracking controller based on the particle swarm optimization algorithm to obtain the real-time target controller parameters. The trajectory tracking module is used to dynamically adjust the trajectory tracking controller based on the real-time target controller parameters in order to continuously track the trajectory of the mother-daughter UAV during the dynamic release of the daughter unit.

Images