A rope tension estimation method for collaborative transport by multiple unmanned helicopters

Abstract

The present invention discloses a method for estimating the rope tension of collaborative transport by multiple unmanned helicopters, which is used to solve the problem that the accuracy of trajectory tracking control is affected by the inability to directly measure the rope tension of the drones during collaborative transport. The present invention regards the rope tension as a system disturbance and designs an echo state network estimator based on minimum learning parameters. The total disturbance in the system is estimated through the flight state information of the drones, thereby more accurately estimating the rope tension in the system. The echo state network proposed in the present invention can accurately obtain the estimated value of the rope tension of the system, which is convenient for targeted compensation control in the subsequent controller design, enhances the anti-interference ability of the system, effectively suppresses the impact of the swing of objects on the drones during flight, and improves the trajectory tracking accuracy of the collaborative transport system of multiple unmanned helicopters.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) flight technology, and in particular to a method for estimating the tension of a rope carried by cooperative transport of multiple unmanned helicopters. Background Art

[0002] Helicopter external suspension transport is primarily used for missions such as emergency rescue, military operations, and forest fire prevention, and is one of the primary applications of helicopters in both the military and civilian sectors. However, traditional single-handed helicopters experience body oscillation during external suspension flight, which can lead to coupled oscillation between the helicopter and the suspended load. This severely impacts the stability and flight safety of both the helicopter and the suspended load. Reducing this coupled oscillation and improving the handling qualities of helicopter suspension systems are hot topics of research for scholars both domestically and internationally.

[0003] In recent years, with the development of multi-drone collaborative technology, related research has increased significantly, with the application of multiple drones for suspended load transport becoming a popular topic. Compared to a single drone, collaborative handling by multiple drones can better leverage their advantages. These include greater maneuverability, greater load carrying capacity, enhanced mission execution, improved fault tolerance and robustness, and greater affordability. The number of drones can be flexibly adjusted according to the payload requirements of different missions, achieving efficient resource allocation. Furthermore, increasing the number of drones makes the overall collaborative handling system more stable, effectively avoiding the sling swing problem associated with a single drone. Currently, most research on collaborative drone handling technology uses quadcopters, but the limited payload size and weight of quadcopters limit their transport capacity. Compared to rotorcraft, collaborative unmanned helicopters offer greater transport capabilities, making them more advantageous in the field of large-mass payload transportation. Therefore, the use of multi-drone rope-suspended handling technology holds great potential.

[0004] Of note, the coupled dynamics between the drone and the payload make the rope tension difficult to measure. Consequently, existing control methods often assume it is unmeasurable, resulting in controllers designed to lack robust interference immunity. Furthermore, the rope tension acting on the unmanned helicopter significantly increases the complexity of the coupled payload transport system and significantly impacts its dynamics and stability. Therefore, a rope tension estimation method is urgently needed to improve the stability of multi-UAV collaborative transport systems and mitigate the effects of object sway on the drones during flight, thereby achieving high-precision trajectory tracking control. Summary of the Invention

[0005] Purpose of the invention: The present invention provides a method for estimating the tension of ropes used in collaborative transport by multiple unmanned helicopters. The method uses an echo state network estimator with minimum learning parameters to reconstruct the unmeasurable tension between the unmanned helicopter and the object. The reconstructed tension is then used in the design of the controller to compensate for the lack of corresponding anti-interference capability of existing controllers.

[0006] To achieve the above objectives, the present invention provides a technical solution: a method for estimating rope tension in collaborative transport by multiple unmanned helicopters. This method uses rope tension as a disturbance and, based on the flight state information of the drones, designs an echo state network estimator with minimum learning parameters to estimate the unmeasurable lumped disturbance in the system, thereby estimating rope tension. The method comprises the following steps:

[0007] Step S1: The multi-UAV collaborative transport system consists of N UAVs, a payload, and ropes connecting the payload to the UAVs. Based on the constraints between the UAVs and the ropes, the corresponding constraint equations are obtained using the Baumgarte method.

[0008] Step S2: The unmeasurable tension-related item F pi As a system disturbance, according to the echo state network approximation principle, the echo state network state / interference estimator pair F is constructed. pi Make estimates;

[0009] Step S3: designing a weight vector adaptive update law to determine the nominal weight vector in the echo state network;

[0010] Step S4: Determine the relevant parameters of the echo state network through Simulink experimental simulation, and complete the design of the minimum learning parameter echo state network estimator.

[0011] Furthermore, in step S1, the constraint equation is:

[0012] g i (L i ,l i )=||L i || 2 -l i 2 =0,i=1,2,...,N

[0013] Where, L i =X i1 -X L is the direction vector from the payload to the unmanned helicopter i; X i1 is the trajectory vector of the unmanned helicopter; X L is the trajectory vector of the load; l iis the nominal length of rope i. However, this method has numerical integration errors, so the Baumgarte method is used to describe the constraint formula as:

[0014]

[0015] in(·) + is the Moore-Penrose generalized inverse; α and β are feedback gains; a u is the unconstrained acceleration of the unmanned helicopter and the payload when the tension is ignored; M is given by M=diag(m1I3,m2I3,…,m N I3) composition; W=[W1,…,W N ] T Depend on Composition; V=[V1,…,V N ] T By V i =2L i T[0 3*3(i-1) I3 0 3*3(i-1) -I3]; I3 is the third-order unit matrix;

[0016] Furthermore, in step S2, the tension on the rope is difficult to measure due to the coupled dynamic relationship between the UAV and the payload, so the unmeasurable tension-related term F pi It is regarded as a system disturbance, and from the UAV six-degree-of-freedom model, it can be known that the rope tension related terms only act on the velocity loop, so it is only necessary to design the rope tension estimator for the velocity loop;

[0017] The present invention adopts the echo state network (ESN) which has a strong learning ability and superior approximation ability in the neural network. The process of designing the ESN state / interference estimator is as follows:

[0018] First, obtain the UAV flight status information output by the system, including speed, attitude angle, attitude angular velocity, etc. Then, based on the continuous function separation technology and echo state network approximation technology, the input X2 and output F are mapped as follows:

[0019]

[0020] Where: It represents the estimation error, and its upper bound is X∈R m ,H∈R k×1 ,y back ∈R l×1 are the input, state and output vectors respectively; Γ * ∈R k×l is the ideal weight vector, Γ in ∈R k×m , Γ∈Rk×k , Γ back ∈R k×l The weight matrices from the input layer to the hidden layer, from the hidden layer to the hidden layer, and from the output layer to the hidden layer are H(ψ2)=[h1(ψ2),h2(ψ2),...,h k (ψ2)] T ∈R k is a Gaussian basis function, and

[0021]

[0022] Where: ψ∈R m is the input vector; c=[c1,c2,...,c m ] T ∈R m is the center vector; b j The variable represents the basis function width;

[0023] Then, the echo state network state / interference estimator of the speed loop is designed according to the above formula and

[0024]

[0025] Where: is the ideal weight Γ 0,2 The estimated value of H2(ψ2)=[h 2x (ψ2),h 2y (ψ2),h 2z (ψ2)] T is the Gaussian basis function; X i2 is the input of the estimator; D2=[D 2x ,D 2y ,D 2z ] T is the estimator gain to be designed; is the state estimation error.

[0026] Furthermore, the six-degree-of-freedom model of the drone is specifically:

[0027] According to Newton's second law, the UAV dynamics equation is:

[0028]

[0029]

[0030] From the above dynamic equations, the UAV's six-degree-of-freedom affine nonlinear state equation can be obtained:

[0031]

[0032] Where:

[0033] X 1i =P i ,X 2i =V i ,F pi =-T Li / m i ,U i1 =R i eΘ i / m i -g

[0034] Where N is the number of unmanned helicopters, and the internal state variables of unmanned helicopter i include displacement P i =[x i ,y i ,z i ] T ;speed Θ i is the total distance of the unmanned helicopter; T Li is the force exerted by the load on the unmanned helicopter i; e is the unit vector; m i is the mass of the drone; g = [0,0,g] T is the acceleration due to gravity; R i is the transformation matrix from the unmanned helicopter's body coordinate system to the ground coordinate system.

[0035] Furthermore, in step S3, since the expected weight vector in the echo state network is usually unknown, it is necessary to additionally design a weight vector adaptive update law

[0036]

[0037] Where: υ2=diag(υ 2x ,υ 2y ,υ 2z ) is the adaptive gain matrix; γ2=diag(γ 2x ,γ 2y ,γ 2z ) is the parameter to be adjusted;

[0038] Since the number of nominal weight vectors is large and they are updated in real time, in order to reduce the computational burden, the minimum parameter technique is adopted, that is, the modulus of the nominal weight vector || Γ is used. * || 2 Replace the nominal weight vector, thereby significantly reducing the computational burden of the echo state network.

[0039] Furthermore, in step S4, the relevant parameters are mainly the estimator gain D2, the adaptive gain matrix υ2, and the parameter to be adjusted γ2. To more accurately obtain the specific values ​​of these parameters, numerical simulation using Simulink is used to determine the estimator parameters. The determination of the estimator parameters completes the design of the echo state network, thereby realizing rope tension estimation.

[0040] Beneficial effects:

[0041] The echo state network estimator designed in this invention, based on minimum learning parameters, can accurately estimate the rope tension between the unmanned helicopter and the object, so that during the controller design stage, control compensation can be performed for the rope tension disturbance experienced by the system to improve the system's anti-interference ability, effectively suppressing the impact of the object's swing during flight on the drone, thereby achieving more precise trajectory tracking control. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A block diagram of a trajectory tracking control system based on an echo state network estimator provided by the present invention;

[0043] Figure 2 This is a diagram of a multi-unmanned helicopter collaborative transport model provided by the present invention;

[0044] Figure 3 The three-dimensional trajectory curve of the collaborative transportation of multiple unmanned helicopters provided by the present invention;

[0045] Figure 4 The rope tension estimation curve of the UAV 1 based on the echo state network estimator provided by the present invention;

[0046] Figure 5 The UAV 2 rope tension estimation curve based on the echo state network estimator provided by the present invention DETAILED DESCRIPTION

[0047] The present invention will be further described below with reference to the accompanying drawings.

[0048] This embodiment provides a rope tension estimation method based on an echo state network estimator with minimum learning parameters. The relevant physical parameters of the unmanned helicopter are: the number of drones N = 4, the rope length l i =10m, UAV mass m u =500kg, object weight m L = 100kg, the angle between the rope connecting the drone and the object is 45°. Figure 1As shown, the present invention uses the UAV flight state output by the system as an estimation signal, designs an echo state network estimator, and estimates the rope tension based on the UAV flight state. This is then used in the controller feedforward compensation design to compensate for the effect of rope tension on the UAV during flight. It includes the following steps:

[0049] Step S1: The collaborative transport system of multiple unmanned helicopters consists of N unmanned helicopters, a payload, and ropes connecting the payload to the unmanned helicopters. Considering the constraint relationship between the unmanned helicopters and the ropes, the following constraint formula can be written:

[0050] g i (L i ,l i )=||L i || 2 -l i 2 =0,i=1,2,...,N

[0051] Where, L i =X i1 -X L is the direction vector from the payload to the unmanned helicopter i, l i is the nominal length of rope i. However, this method has numerical integration errors, so the Baumgarte method is used to describe the constraint formula as:

[0052]

[0053] in(·) + is the Moore-Penrose generalized inverse; α and β are feedback gains; a u is the unconstrained acceleration of the unmanned helicopter and the payload when the tension is ignored; M is given by M=diag(m1I3,m2I3,…,m N I3) composition; W=[W1,…,W N ] T Depend on Composition; V=[V1,…,V N ] T By V i =2L i T [0 3*3(i-1) I3 0 3*3(i-1) -I3]; I3 is the third-order unit matrix;

[0054] Step S2: According to Newton's second law, the UAV dynamics equation is:

[0055]

[0056]

[0057] From the above dynamic equations, the UAV's six-degree-of-freedom affine nonlinear state equation can be obtained:

[0058]

[0059] Where:

[0060] X 1i =P i ,X 2i =V i ,F pi =-T Li / m i ,U i1 =R i eΘ i / m i -g

[0061] Where N is the number of unmanned helicopters, and the internal state variables of unmanned helicopter i include displacement P i =[x i ,y i ,z i ] T ;speed Θ i is the total distance of the unmanned helicopter; T Li is the force exerted by the load on the unmanned helicopter i; e is the unit vector; m i is the mass of the drone; g = [0,0,g] T is the acceleration due to gravity; R i is the transformation matrix from the unmanned helicopter's body coordinate system to the ground coordinate system.

[0062] The unmeasured tension-related term F pi are regarded as lumped disturbances respectively, and according to the echo state network approximation principle, the echo state network state / interference estimator pair F can be constructed pi Make an estimate.

[0063]

[0064] Where: is the weight vector; H2(ψ2)=[h 2x (ψ2),h 2y (ψ2),h 2z (ψ2)] T is the Gaussian basis function; X i2 is the input of the estimator; D2=[D 2x ,D 2y ,D 2z ] T is the estimator gain to be designed; is the state estimation error.

[0065] Step S3: Since the nominal weight vector in the echo state network is usually unknown, an additional weight vector adaptive update law needs to be designed to estimate the ideal weight Γ 0,2 .

[0066]

[0067] Where: υ2=diag(υ 2x ,υ 2y ,υ 2z ) is the adaptive gain matrix; γ2=diag(γ 2x ,γ 2y ,γ 2z ) are the parameters to be adjusted.

[0068] Step S4: Determine the relevant parameters of the echo state network through mathematical derivation and Simulink experimental simulation, and finally complete the design of the minimum learning parameter echo state network estimator, as follows:

[0069] The relevant parameters are mainly the estimator gain D2, the adaptive gain matrix υ2, and the parameter to be adjusted γ2. By adjusting the parameters, the estimator gain of the velocity loop is determined to be D2 = 30 × diag (1, 1, 1), the adaptive gain matrix is ​​υ2 = diag (2, 2, 2), and the parameter to be adjusted is γ2 = diag (1, 0.5, 0.5). Since the four drones are symmetrically distributed, the simulation results based on the above parameters give the comparison results of the estimated and actual values ​​of the pulling force of drone 1 and drone 2, as shown in the figure. Figure 4 and Figure 5 shown.

[0070] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of implementation of the present invention. Several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for estimating the tension of ropes used in collaborative transport by multiple unmanned helicopters, characterized by: The method uses rope tension as a disturbance and designs an echo state network estimator based on minimum learning parameters according to the flight state information of the UAV. The method estimates the unmeasurable lumped disturbance in the system, thereby estimating the rope tension. The method specifically includes the following steps: Step S1: According to the constraint relationship between the UAV and the rope, the corresponding constraint equation is obtained using the Baumgarte method; Step S2: The unmeasurable tension-related item F pi As a system disturbance, according to the echo state network approximation principle, the echo state network state / interference estimator pair F is constructed. pi Make estimates; Step S3: designing a weight vector adaptive update law to determine the nominal weight vector in the echo state network; Step S4: Determine the relevant parameters of the echo state network through Simulink experimental simulation, and complete the design of the minimum learning parameter echo state network estimator; In step S2, the tension on the rope is difficult to measure due to the coupled dynamic relationship between the drone and the payload, so the unmeasurable tension-related term F pi It is regarded as a system disturbance, and from the UAV six-degree-of-freedom model, it can be known that the rope tension related terms only act on the velocity loop, so it is only necessary to design the rope tension estimator for the velocity loop; The echo state network (ESN) is a neural network with strong learning and approximation capabilities. The process of designing the ESN state / interference estimator is as follows: First, the flight status information of the UAV is obtained from the system output, including speed, attitude angle, and attitude angular velocity. Then, based on the continuous function separation technology and the echo state network approximation technology, the input X2 and output F are mapped as follows: Where: It represents the estimation error, and its upper bound is X2∈R m ,H∈R k×1 ,y2 back ∈R l×1 are input, state and output vectors respectively; Γ * ∈R k×l is the ideal weight vector, Γ2∈R k×k , Γ2 back ∈R k×l The weight matrices from the input layer to the hidden layer, from the hidden layer to the hidden layer, and from the output layer to the hidden layer are H(ψ2)=[h1(ψ2),h2(ψ2),...,h k (ψ2)] T ∈R k is a Gaussian basis function, and Where: ψ∈R m is the input vector; c=[c1,c2,...,c m ] T ∈R m is the center vector; b j The variable represents the basis function width; Then, the echo state network state / interference estimator of the speed loop is designed according to the above formula and Where: is the ideal weight Γ 0,2 The estimated value of H2(ψ2)=[h 2x (ψ2),h 2y (ψ2),h 2z (ψ2)] T is the Gaussian basis function; X i2 is the input of the estimator; D2=[D 2x ,D 2y ,D 2z ] T is the estimator gain to be designed; is the state estimation error; In step S3, since the expected weight vector parameters in the echo state network are unknown, it is necessary to additionally design the weight vector adaptive update law Where: υ2=diag(υ 2x ,υ 2y ,υ 2z ) is the adaptive gain matrix; γ2=diag(γ 2x ,γ 2y ,γ 2z ) is the parameter to be adjusted; Since the number of nominal weight vectors is large and they are updated in real time, in order to reduce the computational burden, the minimum parameter technique is adopted, that is, the modulus of the nominal weight vector || Γ is used. * || 2 Replace the nominal weight vector, thereby significantly reducing the computational burden of the echo state network.

2. The method for estimating the tension of ropes in collaborative transport by multiple unmanned helicopters according to claim 1, characterized in that: In step S1, the constraint equation is: g i (L i ,l i )=||L i || 2 -l i 2 =0,i=1,2,...,N Where, L i =X i1 -X L is the direction vector from the payload to the unmanned helicopter i; X i1 is the trajectory vector of the unmanned helicopter; X L is the trajectory vector of the load; l i is the nominal length of rope i; however, this method has numerical integration errors, so the Baumgarte method is used to describe the constraint formula as: in(·) + is the Moore-Penrose generalized inverse; α and β are feedback gains; a u It is the unconstrained acceleration of the unmanned helicopter and the load when the tension is ignored; M is given by M=diag(m1I3,m2I3,···,m N I3) composition; W=[W1,···,W N ] T Depend on Composition; V=[V1,···,V N ] T Depend on Composition; I3 is the third-order identity matrix.

3. The method for estimating the tension of ropes in collaborative transport by multiple unmanned helicopters according to claim 1 is characterized in that the six-degree-of-freedom model of the unmanned helicopters is specifically: According to Newton's second law, the UAV dynamics equation is: From the above dynamic equations, the UAV's six-degree-of-freedom affine nonlinear state equation can be obtained: Where: X 1i =P i ,X 2i =V i ,F pi =-T Li / m i ,U i1 =R i eΘ i / m i -g Where N is the number of unmanned helicopters, and the internal state variables of unmanned helicopter i include displacement P i =[x i ,y i ,z i ] T ;speed Θ i is the total distance of the unmanned helicopter; T Li is the force exerted by the load on the unmanned helicopter i; e is the unit vector; m i is the mass of the drone; g = [0,0,g] T is the acceleration due to gravity; R i is the transformation matrix from the unmanned helicopter's body coordinate system to the ground coordinate system.

4. The method for estimating the tension of ropes in collaborative transport by multiple unmanned helicopters according to claim 1, characterized in that: In step S4, the relevant parameters include the estimator gain D2, the adaptive gain matrix υ2, and the parameter to be adjusted γ2. To more accurately obtain the specific values ​​of these parameters, numerical simulation using Simulink is used to determine the estimator parameters. Determining the estimator parameters completes the design of the echo state network, thereby realizing rope tension estimation.

Images