Multi-unmanned aerial vehicle suspension system control method, system and device based on layered architecture

Abstract

The present application belongs to the technical field of multi-unmanned aerial vehicle hoisting control, and particularly relates to a multi-unmanned aerial vehicle hoisting system control method, system and equipment based on a hierarchical architecture. The method constructs a load pose controller based on a dynamics model of the multi-unmanned aerial vehicle hoisting system, obtains a load expected resultant force and a load expected resultant moment for cable tension distribution through load pose control feedforward compensation, obtains cable expected tension vectors of each unmanned aerial vehicle through cable tension distribution and maps the same into expected positions, tracks the expected positions through an unmanned aerial vehicle SMC controller, obtains expected thrust scalar values and expected moments of each unmanned aerial vehicle and applies the same to the multi-unmanned aerial vehicle hoisting system. In the hierarchical control architecture, the upper layer load pose controller provides an expected resultant moment, the middle layer performs cable tension distribution, and the bottom layer SMC controller tracks the expected positions. The upper layer and the middle layer can be converted into QP problems for instant solution, thereby greatly improving the calculation efficiency while ensuring the control performance.

Description

Technical Field

[0001] This invention belongs to the field of multi-UAV hoisting control technology, specifically relating to a control method, system, equipment, and medium for a multi-UAV hoisting system based on a hierarchical architecture. Background Technology

[0002] Multi-rotor unmanned aerial vehicles (UAVs) are widely used in both military and civilian fields such as reconnaissance and surveillance, aerial patrol, disaster relief, and cargo transportation due to their superior vertical takeoff and landing capabilities, excellent maneuverability, and high adaptability to complex scenarios. Robust control strategies such as sliding mode control (SMC), two-stage active disturbance rejection control (ADRC), control Lyapunov function (CLF), and nonlinear model predictive control (NMPC) are currently key research directions in multi-UAV slinging control. The paper "A Pose Control Method for Multi-UAV Load Slinging Based on NMPC" achieves obstacle avoidance and formation cooperative control by constructing a cable tension redundancy optimization mechanism. However, this method relies on complex online iterative optimization calculations, resulting in high computational burden, limited real-time performance, and difficulty in guaranteeing effective obstacle avoidance, requiring further verification. Therefore, current technologies suffer from the problems of high computational burden, difficulty in meeting real-time requirements, and poor load safety associated with traditional methods such as NMPC. Summary of the Invention

[0003] The purpose of this invention is to address the aforementioned problems in the prior art by providing a control method, system, device, and medium for a multi-UAV sling system based on a hierarchical architecture that significantly improves computational efficiency while ensuring control performance.

[0004] To achieve the above objectives, the technical solution of the present invention is as follows:

[0005] In a first aspect, the present invention provides a control method for a multi-UAV sling-mounted system based on a hierarchical architecture, the control method comprising:

[0006] Establish a dynamic model for a multi-UAV sling system;

[0007] A load pose controller is constructed based on the dynamic model. Based on the desired motion state of the load, the load pose controller is used to perform feedforward compensation on the load pose to obtain the desired resultant force and desired resultant torque of the load.

[0008] Based on the expected resultant force and the expected resultant torque of the load, the cable tension is distributed to obtain the expected cable tension vector of each UAV in the multi-UAV sling system. The expected position of each UAV is obtained by mapping the expected cable tension vector of each UAV.

[0009] A drone SMC controller is constructed. The drone SMC controller tracks the desired position of each drone, obtains the desired thrust scalar value and desired torque of each drone, and applies them to the multi-drone sling system.

[0010] The expression for the dynamic model is:

[0011] ;

[0012] ;

[0013] In the above formula, For load quality, Let x be the acceleration of the load in the inertial coordinate system. The expected resultant force of the load generated by the tension of all cables; It is the acceleration due to gravity; Let be the moment of inertia in the load coordinate system. Let be the angular acceleration of the load in the load coordinate system. Let be the angular velocity of the load in the load coordinate system. The desired resultant torque of the load generated by the tension of all cables, For the first The quality of the drone For the first The acceleration of the drone in the inertial coordinate system; For the first The expected thrust scalar value of the drone; This is the first rotation matrix, used to transfer vectors from the UAV's body coordinate system to the inertial coordinate system. Used to represent the z-axis direction in the UAV body coordinate system. This is the second rotation matrix, used to transfer vectors from the load body coordinate system to the inertial coordinate system. In the load coordinate system, the first The tension vector of the cable; ω represents the attitude angular velocity in the UAV's body coordinate system; Let be the attitude angular acceleration in the UAV's body coordinate system. Let be the attitude angular velocity in the UAV's body coordinate system. Let be the moment of inertia in the UAV's body coordinate system. In the UAV body coordinate system, the first The expected torque of the drone.

[0014] The load pose controller includes a load position loop CLF-CBF controller, which performs feedforward compensation on the load position based on the load's desired motion state to obtain the load's desired resultant force.

[0015] The design steps for the load location loop CLF-CBF controller are as follows:

[0016] Define load position tracking error Virtual speed command and load speed tracking error They are respectively , , ,in , These represent the position of the load in the inertial coordinate system and the acceleration of the load in the inertial coordinate system, respectively. , These are the expected load location and the expected load speed, respectively. It is a third-order diagonal matrix;

[0017] Construct its CLF function for the load location loop. for:

[0018] ;

[0019] CLF function for constructing load location loop The constraints are:

[0020] ;

[0021] In the above formula, This is the first convergence rate control coefficient; This is the first relaxation term; The expected resultant force of the load generated by the tension of all cables; For load quality; It is the acceleration due to gravity; The expected acceleration for the load;

[0022] For multi-UAV sling systems, the first challenge is... The second-order CBF constraint for each obstacle is:

[0023] ;

[0024] In the above formula, Indicates the first The location of the obstacle; , All are gain coefficients; This is the safe distance threshold;

[0025] The expression for the load position loop CLF-CBF controller is obtained as follows:

[0026] ;

[0027] In the above formula, for The penalty coefficient; It is a positive definite weight matrix; , These are the minimum and maximum values ​​of the expected resultant force of the load, respectively.

[0028] The load pose controller includes a load attitude loop (CBF) controller, which performs feedforward compensation on the load attitude based on the load's desired motion state to obtain the load's desired resultant torque.

[0029] The design steps for the load attitude loop CBF controller are as follows:

[0030] Define attitude angle error term ,in To represent quaternion multiplication, The desired attitude angle of the load. To determine the attitude of the load in the load coordinate system, the attitude angle error term is... Represented as ,in For the attitude angle error term, scalar The vector of the attitude angle error term is used to construct the desired attitude angular velocity under load. for:

[0031] ;

[0032] In the above formula, , These are the proportional gain matrix and the integral gain matrix, respectively. It is a time variable;

[0033] Define angular velocity error , Given the load's attitude angular velocity in the load coordinate system, construct its CLF function for the load attitude loop. for: ;

[0034] CLF function for constructing load attitude loop The constraints are:

[0035] ;

[0036] ;

[0037] In the above formula, As an intermediate quantity, The desired resultant torque of the load generated by the tension of all cables, Let be the moment of inertia of the load coordinate system. The desired attitude angular acceleration under load. This is the second convergence rate control coefficient. This is the second relaxation term;

[0038] The expression for the load attitude loop CLF-CBF controller is obtained as follows:

[0039] ;

[0040] In the above formula, , These are the minimum and maximum values ​​of the expected resultant torque of the load, respectively.

[0041] The cable tension distribution is based on the zero-space component of the cable tension, and the cable tension distribution steps include:

[0042] The expected resultant force of the load generated by all cable tensions Transform to the load coordinate system to obtain , The second rotation matrix is ​​used to transfer vectors from the load's body coordinate system to the inertial coordinate system, and to determine the desired resultant force of the load. Resultant torque with load The union is the desired resultant force vector. ;

[0043] Definition of the first Tension vector of the cable in the load coordinate system Then there is , ,in For the first The position vector of the suspension point of the cable in the load coordinate system. Let be the number of drones, where the number of drones is equal to the number of cables. for The antisymmetric matrix forms a row vector from all cable tension vectors in the load coordinate system. ,but , ;

[0044] definition It is a constant full-rank matrix. Given a 3x3 identity matrix, then ,in A constant full-rank matrix The pseudo-inverse matrix; definition A constant full-rank matrix The zero-space constant basis matrix, If is a null space coordinate vector, then ,in Represents the zero spatial component of cable tension; definition Let be the desired tension vector of all cables in the load coordinate system, then we have ,in In the load coordinate system, the first The expected tension vector of the cable;

[0045] definition For the first The desired direction vector of the first cable in the load coordinate system is obtained. Position of the drone in the inertial coordinate system ,in Let this be the position of the load in the inertial coordinate system. For the first The length of the cable;

[0046] Using null space coordinate vector To optimize the variables, an optimization model for the expected thrust scalar value is set up. Solving the optimization model yields the optimal null coordinate vector that satisfies the maximum cable tension constraint and the spatial obstacle avoidance constraint of the UAV. Based on the optimal null coordinate vector, the optimal cable tension vector for each UAV is calculated. The expression of the expected thrust scalar value optimization model is as follows:

[0047] ;

[0048] In the above formula, For the nominal case, the first The maximum cable tension that a drone can provide; ; For the first The maximum thrust of the drone; , The first frame and the first The position of the drone in the inertial coordinate system; For load quality; It is the acceleration due to gravity; , These are the safe distance thresholds between drones and the safe distance threshold between a drone and an obstacle, respectively. For the first The position of an obstacle in an inertial coordinate system.

[0049] The construction steps of the UAV SMC controller include:

[0050] Define virtual control variables and UAV position tracking error Take the sliding surface ,in For the first The expected thrust scalar value of the drone. This is the first rotation matrix used to transfer vectors from the UAV's body coordinate system to its inertial coordinate system. Used to represent the z-axis direction in the UAV body coordinate system. For the first The position of the drone in the inertial coordinate system. For the first The desired position of the drone in the inertial coordinate system. For UAV position tracking error The derivative, To achieve sliding mode surface gain, the UAV position controller was designed as follows:

[0051] ;

[0052] In the above formula, For the virtual control quantity of the UAV position controller; For the first The quality of the drone; It is the acceleration due to gravity; For the first The desired acceleration of the drone in the inertial coordinate system. The thickness of the boundary layer at the synovial surface. For lumped perturbation related gain Boundary layer thickness;

[0053] Virtual control quantity based on UAV position controller Extract the first Expected thrust scalar value of the drone and the desired thrust direction : , The desired thrust direction Split into ,in , , They are respectively The components along the x, y, and z axes of the inertial coordinate system are used to solve the first... Desired pitch angle of the drone and roll angle They are respectively: , ;

[0054] Let the first Desired yaw angle of the drone The desired yaw angle Desired pitch angle and roll angle Transform into multivariate representation Take the first The actual attitude angle of the drone is Calculate the attitude angle error term between the desired attitude angle and the actual attitude angle. ,Will Represented as ,in This is the scalar part of the attitude angle error term. It is the vector part of the attitude angle error term; it generates the desired attitude angular velocity of the UAV. : ,in , These are the proportional gain matrix and the integral gain matrix, respectively.

[0055] Define the angular velocity tracking error of the UAV , Construct an integral sliding mode surface for the attitude angular velocity in the UAV body coordinate system. ,in For integral sliding surface gain, Assuming time is the variable, the UAV attitude controller is designed as follows:

[0056] ;

[0057] In the above formula, To control the gain of sliding mode approximation, Let be the moment of inertia in the UAV's body coordinate system; The desired attitude angle acceleration for the UAV; The integral of the sliding surface boundary layer thickness; For the first The desired torque of the drone in the drone's body coordinate system; ω represents the attitude angular velocity in the UAV's body coordinate system.

[0058] The desired motion state of the load includes the desired position, desired velocity, desired acceleration of the load, and desired attitude angle, desired attitude angular velocity, and desired attitude angular acceleration of the load.

[0059] Secondly, the present invention provides a multi-UAV sling-mounted control system based on a hierarchical architecture, which is based on the aforementioned multi-UAV sling-mounted system control method. The multi-UAV sling-mounted control system includes:

[0060] The dynamics model building module is used to build dynamics models of multi-UAV sling systems;

[0061] The upper-layer load pose compensation module is used to construct a load pose controller based on the dynamic model, and to perform feedforward compensation on the load pose based on the load's desired motion state through the load pose control to obtain the load's desired resultant force and load's desired resultant torque.

[0062] The middle layer cable tension distribution module is used to distribute cable tension based on the expected resultant force and the expected resultant torque of the load, to obtain the expected cable tension vector of each UAV in the multi-UAV sling system, and to obtain the expected position of each UAV by mapping the expected cable tension vector of each UAV.

[0063] The lower-level UAV control module is used to construct the UAV SMC controller. The UAV SMC controller tracks the desired position of each UAV, obtains the desired thrust scalar value and desired torque of each UAV, and applies them to the multi-UAV sling system.

[0064] Thirdly, the present invention provides a multi-UAV sling system control device based on a hierarchical architecture, the multi-UAV sling system control device including a memory and a processor; the memory is used to store computer program code and transmit the computer program code to the processor; the processor is used to execute the aforementioned multi-UAV sling system control method according to the instructions in the computer program code.

[0065] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned multi-UAV sling system control method.

[0066] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0067] 1. The cable tension distribution and hierarchical control method described in this invention adopts a hierarchical control architecture. The upper layer uses the load pose controller as input to provide the load's expected resultant torque and the expected torque of each UAV. The middle layer uses the load's expected resultant torque and the UAV's expected torque as input to distribute the cable tension, obtaining the expected tension vector of each cable and the expected position of each UAV. The lower layer uses the UAV SMC controller to track the expected position of each UAV, obtaining the expected thrust scalar value and expected torque of each UAV, which are ultimately applied to the multi-UAV sling system. In the above design, the load pose tracking and cable tension distribution problems can be transformed into traditional QP problems for real-time solution, avoiding multi-step iterative optimization. Compared with the traditional NMPC method, which has a large computational burden and is difficult to meet real-time requirements, the above design significantly improves computational efficiency while ensuring control performance.

[0068] 2. The cable tension distribution and hierarchical control method described in this invention applies CBF constraints to the load position loop of the overload posture controller, using CBF constraints to ensure safe obstacle avoidance for the load; in addition, the zero-space component of the cable tension is introduced for cable tension distribution, and the redundant tension provided by the zero-space component is used to adjust the position of the UAV, achieving effective distribution of cable tension while ensuring obstacle avoidance safety for both the load and the UAV; the above design can ensure obstacle avoidance accuracy and improve load safety. Attached Figure Description

[0069] Figure 1 This is a flowchart of the method described in this invention.

[0070] Figure 2 This is a schematic diagram illustrating the control of the method described in this invention.

[0071] Figure 3 For example, Algorithm 1 is paired with... and A schematic diagram of the root mean square error of observation.

[0072] Figure 4 This is a comparison chart of the load position tracking performance of the two algorithms in the example.

[0073] Figure 5 The graph shows a comparison of the load attitude tracking performance of the two algorithms in the example.

[0074] Figure 6 The cable tension distribution results are shown for the two algorithms in the example.

[0075] Figure 7 The image shows the position trajectory of the UAV and the payload in the XY plane according to Algorithm 1.

[0076] Figure 8 This is a schematic diagram of the pose tracking error of UAV No. 1 in Algorithm 1.

[0077] Figure 9 This is a structural block diagram of the system described in this invention.

[0078] Figure 10 This is a structural block diagram of the device described in this invention. Detailed Implementation

[0079] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0080] Example 1:

[0081] See Figure 1 A control method for a multi-UAV sling-mounted system based on a hierarchical architecture is performed in the following steps:

[0082] S1. Establish a dynamic model of the multi-UAV hoisting system.

[0083] Specifically, the expression for the dynamic model is:

[0084] ;

[0085] ;

[0086] In the above formula, For load quality, Let x be the acceleration of the load in the inertial coordinate system. The expected resultant force of the load generated by the tension of all cables; It is the acceleration due to gravity; Let be the moment of inertia in the load coordinate system. Let be the angular acceleration of the load in the load coordinate system. Let be the angular velocity of the load in the load coordinate system. The desired resultant torque of the load generated by the tension of all cables, For the first The quality of the drone For the first The acceleration of the drone in the inertial coordinate system; For the first The expected thrust scalar value of the drone; This is the first rotation matrix, used to transfer vectors from the UAV's body coordinate system to the inertial coordinate system. Used to represent the z-axis direction in the UAV body coordinate system. This is the second rotation matrix, used to transfer vectors from the load body coordinate system to the inertial coordinate system. In the load coordinate system, the first The tension vector of the cable; Let be the attitude angular acceleration in the UAV's body coordinate system. Let be the attitude angular velocity in the UAV's body coordinate system. Let be the moment of inertia in the UAV's body coordinate system. In the UAV body coordinate system, the first The expected torque of the drone.

[0087] S2. Construct a load pose controller based on the dynamic model, and perform feedforward compensation on the load pose based on the load's desired motion state through the load pose control to obtain the load's desired resultant force and load's desired resultant torque.

[0088] Specifically, the desired motion state of the load includes the desired position, desired velocity, desired acceleration of the load, as well as the desired attitude angle, desired attitude angular velocity, and desired attitude angular acceleration of the load.

[0089] Specifically, the load pose controller includes a load position loop CLF-CBF controller, which performs feedforward compensation on the load position based on the load's desired motion state to obtain the load's desired resultant force.

[0090] The design steps for the load location loop CLF-CBF controller are as follows:

[0091] Define load position tracking error Virtual speed command and load speed tracking error They are respectively , , ,in , These represent the position of the load in the inertial coordinate system and the acceleration of the load in the inertial coordinate system, respectively. , These are the expected load location and the expected load speed, respectively. It is a third-order diagonal matrix;

[0092] Construct its CLF function for the load location loop. for:

[0093] ;

[0094] CLF function for constructing load location loop The constraints are:

[0095] ;

[0096] In the above formula, This is the first convergence rate control coefficient. This is the first relaxation term; The expected acceleration for the load;

[0097] For multi-UAV sling systems, the first challenge is... The second-order CBF constraint for each obstacle is:

[0098] ;

[0099] In the above formula, Indicates the first The location of the obstacle , All are gain coefficients. This is the safe distance threshold;

[0100] The expression for the load position loop CLF-CBF controller is obtained as follows:

[0101] ;

[0102] In the above formula, for The penalty coefficient, It is a positive definite weight matrix. , These are the minimum and maximum values ​​of the expected resultant force of the load, respectively.

[0103] Specifically, the load pose controller includes a load pose loop (CBF) controller, which performs feedforward compensation on the load pose based on the load's desired motion state to obtain the load's desired resultant torque.

[0104] The design steps for the load attitude loop CBF controller are as follows:

[0105] Define attitude angle error term ,in To represent quaternion multiplication, The desired attitude angle of the load. To determine the attitude of the load in the load coordinate system, the attitude angle error term is... Represented as ,in For the attitude angle error term, scalar The vector of the attitude angle error term is used to construct the desired attitude angular velocity under load. for:

[0106] ;

[0107] In the above formula, , These are the proportional gain matrix and the integral gain matrix, respectively. It is a time variable;

[0108] Define angular velocity error , Given the load's attitude angular velocity in the load coordinate system, construct its CLF function for the load attitude loop. for: ;

[0109] CLF function for constructing load attitude loop The constraints are:

[0110] ;

[0111] ;

[0112] In the above formula, As an intermediate quantity, The desired resultant torque of the load generated by the tension of all cables, Let be the moment of inertia of the load coordinate system. The desired attitude angular acceleration under load. This is the second convergence rate control coefficient. This is the second relaxation term;

[0113] The expression for the load attitude loop CLF-CBF controller is obtained as follows:

[0114] ;

[0115] In the above formula, , These are the minimum and maximum values ​​of the expected resultant torque of the load, respectively.

[0116] S3. Based on the expected resultant force and the expected resultant torque of the load, the cable tension is distributed to obtain the expected cable tension vector of each UAV in the multi-UAV sling system, and the expected position of each UAV is obtained by mapping the expected cable tension vector of each UAV.

[0117] Specifically, the cable tension distribution is based on the zero-space component of the cable tension, and the cable tension distribution steps include:

[0118] The expected resultant force of the load generated by all cable tensions Transform to the load coordinate system to obtain The combined load expected force Resultant torque with load The union is the desired resultant force vector. ;

[0119] Definition of the first Tension vector of the cable in the load coordinate system Then there is , ,in For the first Let the position vector of the suspension point of the cable in the load coordinate system be... for The antisymmetric matrix forms a row vector from all cable tension vectors in the load coordinate system. , Let the number of drones be equal to the number of cables. , ;

[0120] definition It is a constant full-rank matrix. Given a 3x3 identity matrix, where all diagonal elements are 1 and all other elements are 0, then... ,in A constant full-rank matrix The pseudo-inverse matrix; definition A constant full-rank matrix The zero-space constant basis matrix, If is a null space coordinate vector, then ,in Represents the zero spatial component of cable tension; definition Let be the desired tension vector of all cables in the load coordinate system, then we have ,in In the load coordinate system, the first The expected tension vector of the cable;

[0121] definition For the first The desired direction vector of the first cable in the load coordinate system is obtained. Position of the drone in the inertial coordinate system , For the first The length of the cable;

[0122] Using null space coordinate vector To optimize the variables, an optimization model for the expected thrust scalar value is set up. Solving the optimization model yields the optimal null coordinate vector that satisfies the maximum cable tension constraint and the spatial obstacle avoidance constraint of the UAV. Based on the optimal null coordinate vector, the optimal cable tension vector for each UAV is calculated. The expression of the expected thrust scalar value optimization model is as follows:

[0123] ;

[0124] In the above formula, For the nominal case, the first The maximum cable tension that a drone can provide, , For the first The maximum thrust of the drone , The first frame and the first The position of the drone in the inertial coordinate system. , These are the safe distance thresholds between drones and the safe distance threshold between a drone and an obstacle, respectively. For the first The position of the obstacle in the inertial coordinate system.

[0125] S4. Construct a UAV SMC controller, track the desired position of each UAV through the UAV SMC controller, obtain the desired thrust scalar value and desired torque of each UAV, and apply them to the multi-UAV sling system.

[0126] Specifically, the construction steps of the UAV SMC controller include:

[0127] Define virtual control variables and UAV position tracking error Take the sliding surface ,in For the first The expected thrust scalar value of the drone. This is the first rotation matrix used to transfer vectors from the UAV's body coordinate system to its inertial coordinate system. Used to represent the z-axis direction in the UAV body coordinate system. For the first The position of the drone in the inertial coordinate system. For the first The desired position of the drone in the inertial coordinate system. For UAV position tracking error The derivative, To achieve sliding mode surface gain, the UAV position controller was designed as follows:

[0128] ;

[0129] In the above formula, For the virtual control quantity of the UAV position controller; For the first The desired acceleration of the drone in the inertial coordinate system. The thickness of the boundary layer at the synovial surface. For lumped perturbation related gain Boundary layer thickness;

[0130] Virtual control quantity based on UAV position controller Extract the first Expected thrust scalar value of the drone and the desired thrust direction : , The desired thrust direction Split into ,in , , They are respectively The components along the x, y, and z axes of the inertial coordinate system are used to solve the first... Desired pitch angle of the drone and roll angle They are respectively: , ;

[0131] Let the first Desired yaw angle of the drone The desired yaw angle Desired pitch angle and roll angle Transform into multivariate representation Take the first The actual attitude angle of the drone is Calculate the attitude angle error term between the desired attitude angle and the actual attitude angle. ,Will Represented as ,in This is the scalar part of the attitude angle error term. It is the vector part of the attitude angle error term; it generates the desired attitude angular velocity of the UAV. : ,in , These are the proportional gain matrix and the integral gain matrix, respectively.

[0132] Define the angular velocity tracking error of the UAV , Construct an integral sliding mode surface for the attitude angular velocity in the UAV body coordinate system. ,in For integral sliding surface gain, Assuming time is the variable, the UAV attitude controller is designed as follows:

[0133] ;

[0134] In the above formula, To control the gain of sliding mode approximation, Let be the moment of inertia in the UAV's body coordinate system. For the desired attitude angle acceleration of the UAV, The integral of the sliding surface boundary layer thickness. For the first The desired torque of the drone in the drone's body coordinate system. ω represents the attitude angular velocity in the UAV's body coordinate system.

[0135] Performance verification:

[0136] To verify the robustness and control performance of the proposed method, this example uses two different algorithms built on the Matlab platform for comparative study. Algorithm 1 is the method proposed in this invention, and Algorithm 2 is an optimization method for cable redundancy tension allocation using NMPC as the upper-level controller and combined with the null space method. During the simulation, the geometric control in Algorithm 2 is uniformly replaced with the UAV SMC controller proposed in Algorithm 1 to ensure the fairness of the performance comparison between the upper-level load pose controller and the middle-level cable tension allocation.

[0137] The experimental parameters are set as shown in Table 1. The simulation environment is set as follows: the initial position and attitude angle of the load are (1, 0, 1.5) m and (0, 0, 0) m, respectively; the desired trajectory of the load is a circle with a radius of 1.0 m centered at (0, 0, 1.5) m, a desired velocity of 0.1 rad / s, and a desired attitude angle of (0, 0, 0) rad. The simulation environment and control parameter configuration are shown in Table 1, where N is the prediction time domain, Q... X Q XN Q represents the system state weights and terminal weights, respectively, both being 13-dimensional diagonal matrices; U The input weights are represented by a 12-dimensional diagonal matrix. Three static obstacles were placed within the simulation area: obstacles 1 and 2 were located near the desired trajectory of the load, and obstacle 3 was located near the UAV's flight area. Constant and nonlinear bounded perturbations were applied to the system during the periods of 5-10s and 10-20s, respectively. Four UAVs were used. To verify the cable tension distribution performance under UAV thrust-limited fault conditions, the fault of UAV 1 was triggered at t=10s.

[0138] Table 1 Experimental parameter settings

[0139] ;

[0140] The following are the results of the comparative analysis:

[0141] (1) Algorithm 1 respectively for and Observations were performed, and the root mean square error (RMSE) of the observations was extracted as follows: Figure 3 As shown in Figures (1) and (2). Due to the introduction of constant and nonlinear bounded perturbations at t=5s and 10s, the values ​​at the aforementioned times are... and The RMSE values ​​all fluctuated at t=5s. and The values ​​are 0.25 N and 0.006 N•m, respectively, at t=10 s. and The error values ​​were 0.04N and 0.001N•m respectively, both relatively small, and the tracking deviation at other times was almost zero. This indicates that Algorithm 1... and Effective observation was achieved. Figure 3 The red dashed line in the middle represents the upper bound of the observation error analysis.

[0142] (2) Comparison of load position tracking performance of the two algorithms Figure 4 As shown, Figure 4 Figures (a) and (b) in the middle correspond to Algorithm 1 and Algorithm 2, respectively. The performance comparison of load attitude tracking under the two algorithms is as follows: Figure 5 As shown, Figure 5 Figures (c) and (d) correspond to Algorithm 1 and Algorithm 2, respectively. Figure 4 and Figure 5 The yellow-green and red fluorescent dashed lines represent the constant and nonlinear bounded perturbations introduced at t=5s and 10s, respectively. It can be seen that the tracking errors of both algorithms fluctuate at t=5s and 10s, but Algorithm 1 exhibits significantly smaller error fluctuations. Specifically, for load position tracking, the maximum tracking error of Algorithm 1 is (0.001m, 0.0005m, 0.015m), while that of Algorithm 2 is (0.06m, 0.04m, 0.035m). For load attitude tracking, the maximum tracking error of Algorithm 1 is (0.004rad, 0.012rad, 0.022rad), while that of Algorithm 2 is (0.1rad, 0.4rad, 0.42rad). This indicates that compared to Algorithm 2, Algorithm 1 can effectively compensate for the impact of environmental disturbances and UAV malfunctions on system performance, demonstrating superior accuracy in both load position and load attitude tracking.

[0143] (3) The cable tension distribution results of the two algorithms are as follows: Figure 6 As shown, Figure 6 Figures (e) and (f) correspond to Algorithm 1 and Algorithm 2, respectively. The red fluorescent dashed line indicates that the malfunction of UAV 1 occurred at t=10s. It can be seen that at t=5s, due to the sudden addition of a constant disturbance, the cable tension distribution of Algorithm 2 changed abruptly, while Algorithm 1 achieved a smooth cable tension distribution through real-time observation and compensation of disturbances using LESO. After the fault occurred (t=10s), the cable tension corresponding to UAV 1 and UAV 3 in Algorithm 1 decreased to about 0.7N, while the cable tension corresponding to UAV 2 and UAV 4 increased to about 1.3N. This indicates that Algorithm 1 actively reduced the given value of the cable tension corresponding to the faulty UAV. In addition, since UAV 1 and UAV 3 are diagonally distributed, under the torque balance constraint, when the cable tension corresponding to UAV 1 decreases, the cable tension corresponding to UAV 3 is also adjusted accordingly. Ultimately, UAV 2 and UAV 4 undertake the main tension compensation task, achieving effective fault-tolerant allocation. In contrast, the cable tension corresponding to UAV 1 in Algorithm 2 remained saturated at the upper limit of the fault tension of 1N, failing to achieve active fault tolerance, resulting in an unbalanced cable tension allocation.

[0144] (4) Statistics show that in the non-fault mode (0-10s), Algorithm 2 resulted in 116 load-obstacle collisions and 69 UAV-obstacle collisions; in the fault mode (10-20s), Algorithm 2 resulted in 261 load-obstacle collisions and 9 UAV-obstacle collisions. The position trajectory diagram of the UAV and load in the XY plane of Algorithm 1 is shown below. Figure 7 As shown, the load's motion trajectory ( Figure 7 The black dashed line in the diagram shows that the drone successfully avoided obstacle 1 within 0-10 seconds by passing within a distance of (0.5, 0.5, 1.6) m; it also successfully avoided obstacle 2 within 15-20 seconds by passing within a distance of (-0.3, 0.7, 1.6) m; and it successfully avoided obstacle 3 within 0-10 seconds by passing within a distance of (0.8, 0.8, 5.2) m. This demonstrates that Algorithm 1 can effectively ensure the safe operation of the system in both normal and fault modes.

[0145] (5) The pose tracking error curve of UAV No. 1 in Algorithm 1 is as follows: Figure 8 As shown in the figure, before the fault occurred (t < 10s), the position RMSE and attitude RMSE were stable at 0.05m and 0.3rad, respectively. Fluctuations only occurred at the moment of the fault (t = 10s), with the position RMSE and attitude RMSE reaching maximum values ​​of 0.5m and 1.5rad, respectively. After approximately 1.5s, both position RMSE and attitude RMSE rapidly decreased, reaching almost zero at t = 15s. The experimental results fully verify that Algorithm 1 has strong robustness in both normal and fault modes.

[0146] (6) The upper and middle layers of Algorithm 1 use 5-dimensional and 22-dimensional QP instantaneous solution methods respectively, with an average total time of 9.67ms; while the upper layer of Algorithm 2 uses 30-dimensional NMPC iterative optimization and the middle layer uses 16-dimensional QP instantaneous solution, with an average total time of 91.15ms. It can be seen that compared with Algorithm 2, Algorithm 1 has a computational efficiency improvement of 89.4% and has better real-time deployment capabilities.

[0147] Example 2:

[0148] See Figure 9A multi-UAV sling control system based on a hierarchical architecture includes a dynamic model construction module, an upper-layer load pose compensation module, a middle-layer cable tension distribution module, and a lower-layer UAV control module. The dynamic model construction module is used to establish a dynamic model of the multi-UAV sling system. Specifically, the dynamic model construction module executes S1 in Embodiment 1, which will not be elaborated here. The upper-layer load pose compensation module is used to construct a load pose controller based on the dynamic model. Based on the desired motion state of the load, the load pose controller performs feedforward compensation on the load pose to obtain the desired resultant force and desired resultant torque of the load. Specifically, the upper-layer load pose compensation module executes S2 in Embodiment 1, which will not be elaborated here. The middle-layer cable tension distribution module... The module is used to distribute cable tension based on the expected resultant force and the expected resultant torque of the load, to obtain the expected cable tension vector of each UAV in the multi-UAV sling system, and to map the expected position of each UAV according to the expected cable tension vector of each UAV. Specifically, the middle-layer cable tension distribution module is used to execute S3 in Embodiment 1, which will not be repeated here. The lower-layer UAV control module is used to construct a UAV SMC controller, and to track the expected position of each UAV through the UAV SMC controller, to obtain the expected thrust scalar value and expected torque of each UAV and to apply them to the multi-UAV sling system. Specifically, the lower-layer UAV control module is used to execute S4 in Embodiment 1, which will not be repeated here.

[0149] Example 3:

[0150] See Figure 10 A multi-UAV sling system control device based on a hierarchical architecture is disclosed. The multi-UAV sling system control device includes a memory and a processor. The memory is used to store computer program code and transmit the computer program code to the processor. The processor is used to execute the multi-UAV sling system control method described in Embodiment 1 according to the instructions in the computer program code.

[0151] Example 4:

[0152] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the multi-UAV sling system control method described in Embodiment 1.

[0153] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program goods. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0154] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0155] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0156] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A control method for a multi-UAV sling-mounted system based on a hierarchical architecture, characterized in that: The control method for the multi-UAV sling-mounted system includes: Establish a dynamic model for a multi-UAV sling system; A load pose controller is constructed based on the dynamic model. Based on the desired motion state of the load, the load pose controller is used to perform feedforward compensation on the load pose to obtain the desired resultant force and desired resultant torque of the load. Based on the expected resultant force and the expected resultant torque of the load, the cable tension is distributed to obtain the expected cable tension vector of each UAV in the multi-UAV sling system. The expected position of each UAV is obtained by mapping the expected cable tension vector of each UAV. A drone SMC controller is constructed. The drone SMC controller tracks the desired position of each drone, obtains the desired thrust scalar value and desired torque of each drone, and applies them to the multi-drone sling system.

2. The control method for a multi-UAV sling-mounted system based on a hierarchical architecture according to claim 1, characterized in that: The expression for the dynamic model is: ； ； In the above formula, For load quality, Let x be the acceleration of the load in the inertial coordinate system. The expected resultant force of the load generated by the tension of all cables; It is the acceleration due to gravity; Let be the moment of inertia in the load coordinate system; This represents the angular acceleration of the load in the load coordinate system. Let be the angular velocity of the load in the load coordinate system; The expected resultant torque of the load generated by all cable tensions; For the first The quality of the drone; For the first The acceleration of the drone in the inertial coordinate system; For the first The expected thrust scalar value of the drone; This is the first rotation matrix, used to transfer vectors from the UAV body coordinate system to the inertial coordinate system; Used to represent the z-axis direction in the UAV body coordinate system; This is the second rotation matrix, used to transfer vectors from the load body coordinate system to the inertial coordinate system; In the load coordinate system, the first The tension vector of the cable; ω represents the attitude angular velocity in the UAV's body coordinate system; This refers to the attitude angular acceleration in the UAV's body coordinate system; Let be the moment of inertia in the UAV's body coordinate system. In the UAV body coordinate system, the first The expected torque of the drone.

3. A control method for a multi-UAV sling-mounted system based on a hierarchical architecture according to claim 1 or 2, characterized in that: The load pose controller includes a load position loop CLF-CBF controller, which performs feedforward compensation on the load position based on the load's desired motion state to obtain the load's desired resultant force. The design steps for the load location loop CLF-CBF controller are as follows: Define load position tracking error Virtual speed command and load speed tracking error They are respectively , , ,in , These represent the position of the load in the inertial coordinate system and the acceleration of the load in the inertial coordinate system, respectively. , These are the expected load location and the expected load speed, respectively. It is a third-order diagonal matrix; Construct its CLF function for the load location loop. for: ； CLF function for constructing load location loop The constraints are: ； In the above formula, This is the first convergence rate control coefficient; This is the first relaxation term; The expected resultant force of the load generated by the tension of all cables; For load quality; It is the acceleration due to gravity; The expected acceleration for the load; For multi-UAV sling systems, the first challenge is... The second-order CBF constraint for each obstacle is: ； In the above formula, Indicates the first The location of the obstacle; , All are gain coefficients; This is the safe distance threshold; The expression for the load position loop CLF-CBF controller is obtained as follows: ； In the above formula, for The penalty coefficient, It is a positive definite weight matrix. , These are the minimum and maximum values ​​of the expected resultant force of the load, respectively.

4. A control method for a multi-UAV sling-mounted system based on a hierarchical architecture according to claim 1 or 2, characterized in that: The load posture controller includes a load posture loop CBF controller, which performs feedforward compensation on the load posture based on the load's desired motion state to obtain the load's desired resultant torque. The design steps for the load attitude loop CBF controller are as follows: Define attitude angle error term ,in To represent quaternion multiplication, The desired attitude angle of the load. To determine the attitude of the load in the load coordinate system, the attitude angle error term is... Represented as ,in For the attitude angle error term, scalar The vector of the attitude angle error term is used to construct the desired attitude angular velocity under load. for: ； In the above formula, , These are the proportional gain matrix and the integral gain matrix, respectively. It is a time variable; Define angular velocity error , Given the load's attitude angular velocity in the load coordinate system, construct its CLF function for the load attitude loop. for: ; CLF function for constructing load attitude loop The constraints are: ； ； In the above formula, This is an intermediate quantity; The desired attitude angular acceleration for the load; The expected resultant torque of the load generated by all cable tensions; Let be the moment of inertia of the load coordinate system; This is the second convergence rate control coefficient. This is the second relaxation term; The expression for the load attitude loop CLF-CBF controller is obtained as follows: ； In the above formula, , These are the minimum and maximum values ​​of the expected resultant torque of the load, respectively.

5. A control method for a multi-UAV sling-mounted system based on a hierarchical architecture according to claim 1 or 2, characterized in that: The cable tension distribution is based on the zero-space component of the cable tension, and the cable tension distribution steps include: The expected resultant force of the load generated by all cable tensions Transform to the load coordinate system to obtain , This is the second rotation matrix used to transfer vectors from the load body coordinate system to the inertial coordinate system; it represents the desired resultant force of the load. Resultant torque with load The union is the desired resultant force vector. ; Definition of the first Tension vector of the cable in the load coordinate system Then there is , ,in For the first The position vector of the suspension point of the cable in the load coordinate system. Let be the number of drones, where the number of drones is equal to the number of cables. for The antisymmetric matrix forms a row vector from all cable tension vectors in the load coordinate system. ,but , ; definition It is a constant full-rank matrix. Given a 3x3 identity matrix, then ,in A constant full-rank matrix The pseudo-inverse matrix; definition A constant full-rank matrix The zero-space constant basis matrix, If is a null space coordinate vector, then ,in Represents the zero spatial component of cable tension; definition Let be the desired tension vector of all cables in the load coordinate system, then we have ,in In the load coordinate system, the first The expected tension vector of the cable; definition For the first The desired direction vector of the first cable in the load coordinate system is obtained. Position of the drone in the inertial coordinate system ,in Let this be the position of the load in the inertial coordinate system. For the first The length of the cable; Using null space coordinate vector To optimize the variables, an optimization model for the expected thrust scalar value is set up. Solving the optimization model yields the optimal null coordinate vector that satisfies the maximum cable tension constraint and the spatial obstacle avoidance constraint of the UAV. Based on the optimal null coordinate vector, the optimal cable tension vector for each UAV is calculated. The expression of the expected thrust scalar value optimization model is as follows: ； In the above formula, For the nominal case, the first The maximum cable tension that a drone can provide, , For the first The maximum thrust of the drone , The first frame and the first The position of the drone in the inertial coordinate system; For load quality; It is the acceleration due to gravity; , These are the safe distance thresholds between drones and the safe distance thresholds between drones and obstacles, respectively. For the first The position of an obstacle in an inertial coordinate system.

6. A control method for a multi-UAV sling-mounted system based on a hierarchical architecture according to claim 1 or 2, characterized in that: The construction steps of the UAV SMC controller include: Define virtual control variables and UAV position tracking error Take the sliding surface ,in For the first The expected thrust scalar value of the drone. This is the first rotation matrix used to transfer vectors from the UAV's body coordinate system to its inertial coordinate system. Used to represent the z-axis direction in the UAV body coordinate system. For the first The position of the drone in the inertial coordinate system. For the first The desired position of the drone in the inertial coordinate system. For UAV position tracking error The derivative of To achieve sliding mode surface gain, the UAV position controller was designed as follows: ； In the above formula, For the virtual control quantity of the UAV position controller; For the first The quality of the drone; It is the acceleration due to gravity; For the first The desired acceleration of the drone in the inertial coordinate system. The thickness of the boundary layer at the synovial surface; For lumped disturbance related gain; Boundary layer thickness; Virtual control quantity based on UAV position controller Extract the first Expected thrust scalar value of the drone and the desired thrust direction : , The desired thrust direction Split into ,in , , They are respectively The components along the x, y, and z axes of the inertial coordinate system are used to calculate the first component. Desired pitch angle of the drone and roll angle They are respectively: , ; Let the first Desired yaw angle of the drone The desired yaw angle Desired pitch angle and roll angle Transform into multivariate representation Take the first The actual attitude angle of the drone is Calculate the attitude angle error term between the desired attitude angle and the actual attitude angle. ,Will Represented as ,in This is the scalar part of the attitude angle error term. It is the vector part of the attitude angle error term; it generates the desired attitude angular velocity of the UAV. : ,in , These are the proportional gain matrix and the integral gain matrix, respectively. Define the angular velocity tracking error of the UAV , Construct an integral sliding mode surface for the attitude angular velocity in the UAV body coordinate system. ,in For integral sliding surface gain, Assuming time is the variable, the UAV attitude controller is designed as follows: ； In the above formula, To control the gain for sliding mode approximation; Let be the moment of inertia in the UAV's body coordinate system; The desired attitude angle acceleration for the UAV; The integral of the sliding surface boundary layer thickness; For the first The desired torque of the drone in the drone's body coordinate system; ω represents the attitude angular velocity in the UAV's body coordinate system.

7. The control method for a multi-UAV sling-mounted system based on a hierarchical architecture according to claim 1, characterized in that: The desired motion state of the load includes the desired position, desired velocity, desired acceleration of the load, and desired attitude angle, desired attitude angular velocity, and desired attitude angular acceleration of the load.

8. A multi-UAV sling-mounted control system based on a hierarchical architecture, implemented according to any one of claims 1-7, characterized in that: The multi-UAV sling-mounted control system includes: The dynamics model building module is used to build dynamics models of multi-UAV sling systems; The upper-layer load pose compensation module is used to construct a load pose controller based on the dynamic model, and to perform feedforward compensation on the load pose based on the load's desired motion state through the load pose control to obtain the load's desired resultant force and load's desired resultant torque. The middle layer cable tension distribution module is used to distribute cable tension based on the expected resultant force and the expected resultant torque of the load, to obtain the expected cable tension vector of each UAV in the multi-UAV sling system, and to obtain the expected position of each UAV by mapping the expected cable tension vector of each UAV. The lower-level UAV control module is used to construct the UAV SMC controller. The UAV SMC controller tracks the desired position of each UAV, obtains the desired thrust scalar value and desired torque of each UAV, and applies them to the multi-UAV sling system.

9. A control device for a multi-UAV sling-mounted system based on a hierarchical architecture, characterized in that: The multi-UAV sling system control device includes a memory and a processor; the memory is used to store computer program code and transmit the computer program code to the processor; the processor is used to execute the multi-UAV sling system control method as described in any one of claims 1-7 according to the instructions in the computer program code.

10. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the multi-UAV sling system control method as described in any one of claims 1-7.

Images