Intelligent control method for rope-driven parallel system considering rope spatial synchronization

Abstract

The invention relates to an intelligent control method for a rope-driven parallel system considering the synchronization of rope space, and belongs to the field of control of multi-rope parallel drive systems. The invention solves the problem that the prior art ignores the influence of the synchronization of the lengths of each rope in the rope space on the control accuracy and stability of the end effector. The invention first constructs a theoretical deviation coupling error vector in the rope space, and uses the theoretical deviation coupling error vector to construct an intelligent synchronous controller; a deep reinforcement learning algorithm is used to train a deep neural network in combination with the theoretical deviation coupling error vector; the trained deep neural network recognizes the actual deviation coupling error vector constructed, and outputs the current action on the K in the intelligent synchronous controller. cp and K cd The optimization is performed, and the actual deviation control signal generated by using the actual deviation coupling error vector is combined with the control signal output by the basic controller to generate a synchronous control signal. The present invention is mainly used to control the motor in the rope-driven parallel system.

Description

Technical Field

[0001] The invention belongs to the field of multi-rope parallel drive system control, and in particular relates to an intelligent control method for a rope-driven parallel system taking into account the spatial synchronization of ropes. Background Art

[0002] The rope-driven parallel system (i.e., multi-rope parallel drive system) is widely used in large-span cargo handling, loading and unloading, and some specific scenarios due to its special flexible characteristics. However, since the rope-driven parallel system uses ropes as the driving unit, it is complicated and difficult to accurately control the position and posture of the end effector, which greatly limits the practical application of the rope-driven parallel system. When designing the control scheme of the rope-driven parallel system, an important problem that is currently ignored is that in the process of adjusting the length of multiple ropes, the asynchronous adjustment of the rope length will cause the multiple ropes to oscillate and thus affect the control effect of the end effector. First, since the rope-driven parallel system is a parallel distributed configuration: n motors are placed at n positions in the workspace, and each motor independently controls the retraction and release of a rope. Secondly, the rope-driven parallel system actually controls the position and posture of the end effector indirectly by controlling n independent motors to drive n independent ropes. Finally, because each position and posture point of the end effector corresponds to a set of rope lengths, only when the n ropes reach their respective expected lengths at the same time can the end effector reach the expected point. Therefore, the synchronization of the rope length in the rope space will directly affect the control accuracy and stability of the end effector, and this problem needs to be solved. Summary of the invention

[0003] The purpose of the present invention is to solve the problem that the prior art ignores the influence of the synchronization of the lengths of each rope in the rope space on the control accuracy and stability of the end effector. The present invention provides an intelligent control method for a rope-driven parallel system considering the synchronization of the rope space.

[0004] An intelligent control method for a rope-driven parallel system considering rope spatial synchronization comprises the following steps:

[0005] Step 1: Use the dynamic model of the rope-driven parallel system to calculate the theoretical length of each rope in the rope-driven parallel system; according to the theoretical length of each rope in the rope-driven parallel system, construct the theoretical deviation coupling error vector e in the rope space c-rc ;

[0006] Step 2: Use the theoretical deviation coupling error vector e c-rc Construct an intelligent synchronization controller that takes into account the spatial synchronization of ropes;

[0007] Step 3: Use deep reinforcement learning algorithm and combine theoretical deviation coupling error vector e c-rc, train the deep neural network to obtain the trained deep neural network;

[0008] Step 4: Use the actual length of each rope in the rope-driven parallel system collected at the current moment to construct the actual deviation coupling error vector e′ in the rope space c-rc ;

[0009] Step 5: Couple the actual deviation error vector e′ c-rc As the input of the trained deep neural network, the actual deviation coupling error vector e′ is obtained by the trained deep neural network. c-rc After recognition, output the current action a′;

[0010] Step 6: Use the current action a′ to calculate the proportional parameter matrix K of the intelligent synchronous controller obtained in step 2 cp and the differential parameter matrix K cd Update and obtain the updated scale parameter matrix K′ cp and the updated differential parameter matrix K′ cd ;

[0011] Step 7: The intelligent synchronous controller is based on the updated proportional parameter matrix K′ cp , the updated differential parameter matrix K′ cd , and the actual deviation coupling error vector e′ obtained in the rope space in step 4 c-rc Generate actual deviation control signal u′ crcc ; Then control the signal u′ according to the actual deviation crcc Generate synchronization control signal u tc-crcc Control the multi-rope parallel drive system;

[0012] Among them, u tc-crcc =u tc -u′ crcc ,u tc It is the control signal output by the basic controller.

[0013] Preferably, in step 1, e c-rc =[e c-rc1 ...e c-rcn ] T ; Among them, e c-rci is the theoretical deviation coupling error corresponding to the i-th rope, i = 1, 2, ..., n;

[0014] The theoretical deviation coupling error vector e c-rc All theoretical deviation coupling errors in form a ring topology structure, specifically:

[0015]

[0016] in,

[0017] is the expected length of the i-th rope; e ci is the theoretical tracking error of the length of the i-th rope, and

[0018] is the theoretical length of the i-th rope.

[0019] Preferably, in step 2, the model of the intelligent synchronous controller is constructed as follows:

[0020]

[0021] K cp =diag{K cp1 ,K cp2 ,,...,K cpn};

[0022] K cd =diag{K cd1 ,K cd2 ,,...,K cdn};

[0023] Among them, u crcc is the theoretical deviation control signal output by the intelligent synchronous controller. is the theoretical deviation coupling error vector e c-rc The derivative of K cpi is the scale parameter matrix K cp The i-th component in K cdi is the differential parameter matrix K cd The i-th component in , i = 1, 2, ..., n.

[0024] Preferably, in step 3, a deep reinforcement learning algorithm is used in combination with the theoretical deviation coupling error vector e c-rc , train the deep neural network, and obtain the implementation methods of the trained deep neural network including:

[0025] Step 31: Construct a Markov decision process, specifically:

[0026] First, the theoretical deviation coupling error vector e c-rc Defined as the state of the Markov decision process x(m) = e c-rc (m); where e c-rc (m) represents the theoretical deviation coupling error vector e c-rc The result of the mth step after discretization; x(m) is the state of the mth control cycle;

[0027] Secondly, using Kcp and K cd The vector composed of the components in is defined as the action a(m) of the Markov decision process = [K cp1 ...K cpn K cd1 ...K cdn ] T ; Among them, K cpi is the scale parameter matrix K cp The i-th component in K cdi is the differential parameter matrix K cd The i-th component in , i = 1, 2, ..., n; a(m) is the action of the m-th control cycle;

[0028] Finally, based on the state x(m) and action a(m) of the Markov decision process, the Markov decision process is described as:

[0029]

[0030] The control cost function of the Markov decision process Discount factor γ∈[0,1) for the Markov decision process;

[0031] in, represents the state space, represents the action space;

[0032] represents the probability of state x(m) transferring to x(m+1) when action a(m) is taken in state x(m); W is a positive definite weight matrix;

[0033] Step 32: Use reinforcement learning algorithm to train the deep neural network. During the training process, the state x(m) of the Markov decision process is used as the input of the deep neural network, and the action a(m) of the Markov decision process is used as the output of the deep neural network.

[0034] Preferably, in step 5, e′ c-rc =[e′ c-rc1 ...e′ c-rcn ] T ;

[0035] Among them, e′ c-rci is the actual deviation coupling error corresponding to the i-th rope, i = 1, 2, ..., n;

[0036] In the actual deviation coupling error vector e′ c-rc All actual deviation coupling errors in form a ring topology structure, specifically:

[0037]

[0038] in,

[0039] is the expected length of the i-th rope; e′ ci is the actual tracking error of the length of the i-th rope, and is the actual length of the i-th rope.

[0040] Preferably, in step 5, a′=[K′ cp1 ...K′ cpn K′ cd1 ...K′ cdn ] T ;

[0041] K′ cpi is the scale parameter matrix K cp The optimal value of the i-th component in K′ cdi is the differential parameter matrix K cd The optimal value of the i-th component in .

[0042] Preferably, the implementation method of step 1, calculating the theoretical length of each rope in the rope-driven parallel system by using the rope-driven parallel system dynamics model comprises:

[0043] Step 1: According to the dynamic model of the rope-driven parallel system, the speed of the end effector in the rope-driven parallel system is obtained. And the speed of the end effector Integrate to obtain the theoretical position of the end effector X = [b Τ ψ Τ ] Τ ;in,

[0044] b=[x o y o z o ] Τ , b is the translation vector of the end effector, x o ,y o and z o are the positions of the end effector on the X, Y and Z axes of the world coordinate system O-XYZ, respectively;

[0045] ψ=[α o β o γ o ] Τ , ψ is the rotation vector, α o , β o and γ o They are the rotation angles around the X, Y and Z axes of the world coordinate system O-XYZ;

[0046] Step 1 and 2: According to the theoretical position X of the end effector, obtain the theoretical position vector l of the rope i ;

[0047] l i =b+R·a i -p i ;

[0048] Among them, l i is the theoretical position vector of the i-th rope, R is the rotation matrix between the world coordinate system O-XYZ and the body coordinate system B-xyz of the end effector, a i Point B i Coordinate vector in the body coordinate system B-xyz, i=1,2,...,n, B i is the connection point of the i-th rope on the end effector, p i For point P i Coordinate vector in the world coordinate system O-XYZ, P i is the exit point of the i-th rope on the pulley;

[0049] Step 13: According to the rope theory position vector l i , get the theoretical length of each rope Among them, ||·||2 is the two-norm.

[0050] Preferably, the rope-driven parallel system dynamics model is:

[0051]

[0052] in,

[0053]

[0054] u is the motor torque as the input of the system, M is the positive definite symmetric inertia matrix, J is the Jacobian matrix, I m and F v are the inertia matrix and viscous friction matrix of the rope winding device, R T is the transmission ratio from motor angle to rope length, is the Coriolis matrix, G is the gravity vector, and are the velocity and acceleration vectors of the end effector, (J T ) + It's J T The left pseudo-inverse matrix, (J T ) + =(JJ T ) -1 J, M eq , C eq and H eqare the first to third intermediate variable matrices respectively.

[0055] Preferably, in step seven,

[0056] in, is e′ c-rc The derivative of .

[0057] Advantages of the present invention:

[0058] (1) The present invention takes into account the influence of rope spatial synchronization on the multi-rope length adjustment process and constructs a multi-rope length coupling error vector based on ring topology deviation coupling. The intelligent synchronization controller designed based on this coupling error can effectively improve the synchronization of multi-rope length adjustment and significantly reduce the rope oscillation in the multi-rope length adjustment process.

[0059] (2) Compared with the traditional control method, the intelligent control method of the rope-driven parallel system designed by the present invention considering the spatial synchronization of the rope can effectively improve the control accuracy of the end effector and weaken its own oscillation. When the intelligent control method of the rope-driven parallel system considering the spatial synchronization of the rope is used for tracking, the X-axis tracking error is reduced by 80%, the Y-axis tracking error is reduced by 78.4%, and the Z-axis tracking error is reduced by 62.7%.

[0060] (3) The present invention takes into account that the parameter adjustment of traditional controllers is very cumbersome and difficult during the design process. However, deep reinforcement learning has strong robustness and excellent optimization performance. The present invention adopts a deep reinforcement learning algorithm to perform offline training and online optimization on the parameters of the controller to achieve rationalization and optimization of parameter selection. Due to the unsupervised learning characteristics of reinforcement learning, it can autonomously generate parameter optimization strategies during offline training without human intervention, reducing the difficulty of manually setting parameters. At the same time, the optimization strategy obtained by offline training can adjust the controller parameters online according to the actual scenario, greatly improving the robustness and adaptability of the controller.

[0061] Therefore, when the rope-driven parallel system intelligent control method of the present invention considers the rope spatial synchronization to control the motor in the rope-driven parallel system, the control accuracy and stability of the rope-driven parallel system on the end effector are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 The schematic diagram of the principle of a rope-driven parallel system with 6 degrees of freedom controlled by n ropes; i is the tension vector corresponding to the i-th rope, g is the gravitational acceleration vector;

[0063] Figure 2The control method implemented by the rope-driven parallel system intelligent control method considering the rope spatial synchronization of the present invention and the APD controller is a tracking comparison diagram in the XY plane when tracking a circular trajectory; wherein EXP is the expected trajectory of the circular trajectory in the XY plane;

[0064] Figure 3 This is a control method implemented by the rope-driven parallel system intelligent control method considering the rope spatial synchronization of the present invention and the APD controller, and a tracking error comparison diagram on the X-axis;

[0065] Figure 4 This is a control method implemented by the rope-driven parallel system intelligent control method considering the rope spatial synchronization of the present invention and the APD controller, and a tracking error comparison diagram on the Y axis;

[0066] Figure 5 A comparison chart of tracking errors on the Z axis is shown in the figure. The figure shows a control method implemented by the rope-driven parallel system intelligent control method considering the rope spatial synchronization of the present invention and the APD controller. DETAILED DESCRIPTION

[0067] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0068] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0069] Example 1: Combined with Figure 1 Embodiment 1 is described below. The method for intelligent control of a rope-driven parallel system considering rope spatial synchronization described in this embodiment includes the following steps:

[0070] Step 1: Use the dynamic model of the rope-driven parallel system to calculate the theoretical length of each rope in the rope-driven parallel system; according to the theoretical length of each rope in the rope-driven parallel system, construct the theoretical deviation coupling error vector e in the rope space c-rc ;

[0071] Step 2: Use the theoretical deviation coupling error vector e c-rc Construct an intelligent synchronization controller that takes into account the spatial synchronization of ropes;

[0072] Step 3: Use deep reinforcement learning algorithm and combine theoretical deviation coupling error vector e c-rc , train the deep neural network to obtain the trained deep neural network;

[0073] Step 4: Use the actual length of each rope in the rope-driven parallel system collected at the current moment to construct the actual deviation coupling error vector e′ in the rope space c-rc ;

[0074] Step 5: Couple the actual deviation error vector e′ c-rc As the input of the trained deep neural network, the actual deviation coupling error vector e′ is obtained by the trained deep neural network. c-rc After recognition, output the current action a′;

[0075] Step 6: Use the current action a′ to calculate the proportional parameter matrix K of the intelligent synchronous controller obtained in step 2 cp and the differential parameter matrix K cd Update and obtain the updated scale parameter matrix K′ cp and the updated differential parameter matrix K′ cd ; Step 7, the intelligent synchronous controller is based on the updated proportional parameter matrix K′ cp , the updated differential parameter matrix K′ cd , and the actual deviation coupling error vector e′ obtained in the rope space in step 4 c-rc Generate actual deviation control signal u′ crcc ; Then control the signal u′ according to the actual deviation crcc Generate synchronization control signal u tc-crcc Control the multi-rope parallel drive system;

[0076] Among them, u tc-crcc =u tc -u′ crcc ,u tc It is the control signal output by the basic controller.

[0077] When applying specifically, in, is e′ c-rc The basic controller is a controller formed by any control method in the prior art. In order to ensure the stability and convergence of the motor control in the rope-driven parallel system, it is only necessary to ensure that the multi-rope parallel drive system is stable, and the basic controller generates a control signal u tc The specific implementation scheme can be implemented by any control method in the prior art. Specifically, the control signal u output by the basic controller tc The control signal u can be generated according to the position and speed of the end effector and the length and speed of the rope in the prior art. tc , which is also an existing method.

[0078] In specific applications, the deep reinforcement learning algorithm is an actor-critic reinforcement learning algorithm based on Lyapunov. Figure 1 The schematic diagram of a rope-driven parallel system with n ropes controlling 6 degrees of freedom is given in the figure. The dynamic model of the rope-driven parallel system is:

[0079]

[0080] in,

[0081]

[0082] u is the motor torque as the input of the system, M is the positive definite symmetric inertia matrix, J is the Jacobian matrix, is the derivative of J, I m and F v are the inertia matrix and viscous friction matrix of the rope winding device, R T is the transmission ratio from motor angle to rope length, is the Coriolis matrix, G is the gravity vector, and are the velocity and acceleration vectors of the end effector, (J T ) + It's J T The left pseudo-inverse matrix, (J T ) + =(JJ T ) -1 J, M eq , C eq and H eq are the first to third intermediate variable matrices respectively.

[0083] Furthermore, in step 1, the implementation method of calculating the theoretical length of each rope in the rope-driven parallel system using the rope-driven parallel system dynamics model includes:

[0084] Step 1: According to the dynamic model of the rope-driven parallel system, the speed of the end effector in the rope-driven parallel system is obtained. And the speed of the end effector Integrate to obtain the theoretical position of the end effector X = [b Τ ψ Τ ] Τ ;in,

[0085] b=[x o y o z o ] Τ , b is the translation vector of the end effector, x o ,y o and z oare the positions of the end effector on the X, Y and Z axes of the world coordinate system O-XYZ, respectively;

[0086] ψ=[α o β o γ o ] Τ , ψ is the rotation vector, α o , β o and γ o They are the rotation angles around the X, Y and Z axes of the world coordinate system O-XYZ;

[0087] Step 1 and 2: According to the theoretical position X of the end effector, obtain the theoretical position vector l of the rope i ;

[0088] l i =b+R·a i -p i ;

[0089] Among them, l i is the theoretical position vector of the i-th rope, R is the rotation matrix between the world coordinate system O-XYZ and the body coordinate system B-xyz of the end effector, a i Point B i Coordinate vector in the body coordinate system B-xyz, i=1,2,...,n, B i is the connection point of the i-th rope on the end effector, p i For point P i Coordinate vector in the world coordinate system O-XYZ, P i is the exit point of the i-th rope on the pulley;

[0090] Step 13: According to the rope theory position vector l i , get the theoretical length of each rope Among them, ||·||2 is the two-norm.

[0091] In specific applications, when a rope-driven parallel system with three degrees of freedom is controlled, the dynamic model of the rope-driven parallel system is shown in Formula 1. Since there are three degrees of freedom, the theoretical position X of the end effector in Formula 1 degenerates into a translation vector b, and the rotation vector ψ is 0.

[0092] The synchronization of rope length directly affects the control accuracy of the end effector. In order to make multiple ropes reach their respective desired lengths at the same time, we designed the theoretical deviation coupling error vector e in rope space: c-rc and the actual deviation coupling error vector e′ c-rc , and the deviation coupling error vectors are all multi-rope length coupling error vectors based on ring topology deviation coupling.

[0093] Furthermore, in step 1, e c-rc =[e c-rc1 ...e c-rcn ] T ;

[0094] Among them, e c-rci is the theoretical deviation coupling error corresponding to the i-th rope, i = 1, 2, ..., n;

[0095] The theoretical deviation coupling error vector e c-rc All theoretical deviation coupling errors in form a ring topology structure, specifically:

[0096]

[0097] in,

[0098] is the expected length of the i-th rope; e ci is the theoretical tracking error of the length of the i-th rope, and

[0099] is the theoretical length of the i-th rope.

[0100] Furthermore, in step 5, e′ c-rc =[e′ c-rc1 ...e′ c-rcn ] T ;

[0101] Among them, e′ c-rci is the actual deviation coupling error corresponding to the i-th rope, i = 1, 2, ..., n;

[0102] In the actual deviation coupling error vector e′ c-rc All actual deviation coupling errors in form a ring topology structure, specifically:

[0103]

[0104] in,

[0105] is the expected length of the i-th rope; e′ ci is the actual tracking error of the length of the i-th rope, and is the actual length of the i-th rope.

[0106] Furthermore, in step 2, the model of the intelligent synchronous controller is constructed as follows:

[0107]

[0108] K cp =diag{K cp1 ,K cp2, ,...,K cpn};

[0109] K cd =diag{K cd1 ,K cd2, ,...,K cdn};

[0110] Among them, u crcc is the theoretical deviation control signal output by the intelligent synchronous controller. is the theoretical deviation coupling error vector e c-rc The derivative of K cpi is the scale parameter matrix K cp The i-th component in K cdi is the differential parameter matrix K cd The i-th component in , i = 1, 2, ..., n.

[0111] Furthermore, in step 3, a deep reinforcement learning algorithm is used and combined with the theoretical deviation coupling error vector e c-rc , train the deep neural network, and obtain the implementation methods of the trained deep neural network including:

[0112] Step 31: Construct a Markov decision process, specifically:

[0113] First, the theoretical deviation coupling error vector e c-rc Defined as the state of the Markov decision process x(m) = e c-rc (m); where e c-rc (m) represents the theoretical deviation coupling error vector e c-rc The result of the mth step after discretization; x(m) is the state of the mth control cycle;

[0114] Secondly, using K cp and K cd The vector composed of the components in is defined as the action a(m) of the Markov decision process = [K cp1 ...K cpn K cd1 ...K cdn ] T ; Among them, K cpi is the scale parameter matrix K cp The i-th component in K cdi is the differential parameter matrix K cd The i-th component in , i = 1, 2, ..., n; a(m) is the action of the m-th control cycle;

[0115] Finally, based on the state x(m) and action a(m) of the Markov decision process, the Markov decision process is described as:

[0116]

[0117] The control cost function of the Markov decision process Discount factor γ∈[0,1) for the Markov decision process;

[0118] in, represents the state space, represents the action space;

[0119] represents the probability of state x(m) transferring to x(m+1) when action a(m) is taken in state x(m); W is a positive definite weight matrix;

[0120] Step 32: Use reinforcement learning algorithm to train the deep neural network. During the training process, the state x(m) of the Markov decision process is used as the input of the deep neural network, and the action a(m) of the Markov decision process is used as the output of the deep neural network.

[0121] Furthermore, in step 5, a′=[K′ cp1 ...K′ cpn K′ cd1 ...K′ cdn ] T ;

[0122] K′ cpi is the scale parameter matrix K cp The optimal value of the i-th component in K′ cdi is the differential parameter matrix K cd The optimal value of the i-th component in .

[0123] In this preferred embodiment, when optimizing the intelligent synchronous controller, the two parameter matrices K of the intelligent synchronous controller obtained in step 2 are optimized by using the current action a′. cp and K cd Update to achieve K cp and K cd Optimization, that is, optimization of the intelligent synchronous controller.

[0124] The rope-driven parallel system intelligent control method considering the rope spatial synchronization of the present invention, in order to ensure the stability and convergence of the motor control in the rope-driven parallel system, uses the intelligent synchronous controller constructed by the present invention in combination with a basic controller formed by any control method in the prior art, and only needs to ensure that the rope-driven parallel system is stable. The two are used in combination to generate a control signal u tc-crcc =u tc -u′ crcc , acting on the motor in the rope-driven parallel system.

[0125] Verification test:

[0126] The present invention controls the synchronous control signal u of the motor in the rope-driven parallel system tc-crcc for:

[0127]

[0128] Among them, e b =bb d =[e b1 e b2 e b3 ] Τ represents the tracking error of the end effector position, b d =[b d1 b d2 b d3 ] Τ is the desired translation vector, e b1 、e b2 and e b3 are the tracking errors of the end effector on the X, Y and Z axes of the world coordinate system O-XYZ, respectively. d1 、b d2 and b d3 They represent the expected values ​​of the end effector on the X, Y, and Z axes of the world coordinate system O-XYZ, respectively, and s is the error based on the current error e b The constructed sliding surface: u NTSMC is the control signal output by the basic controller, and the synchronous control signal u tc-crcc =u NTSM-CRCC , that is u NTSM-CRCC It is also a synchronization control signal generated by the method of the present invention.

[0129] In order to increase the comparison and better reflect the superiority of the intelligent control method of the rope-driven parallel system considering the rope spatial synchronization proposed in the present invention, a comparative experiment is carried out between the intelligent control method of the rope-driven parallel system considering the rope spatial synchronization of the present invention and the APD controller. The APD controller is a typical controller in the prior art and is used as the basic controller. The model of the APD controller is as follows:

[0130]

[0131] Among them, u a is the control signal output by the APD controller, T exp represents the desired tension in the rope, and represent the desired velocity and acceleration of the rope, respectively. and Represent the error of rope length and the speed of error change, K p and K d is the parameter of the APD controller, and K p is the first parameter matrix, K d is the second parameter matrix; R T is the transmission ratio matrix from drum to rope length, F v is the viscous friction matrix.

[0132] In this comparative experiment, the above-mentioned relevant parameters are selected as follows:

[0133] F v =diag{0.6345,0.61711,0.62511}, G=[0,0,-3·9.8] Τ , R T =diag{0.063,0.063,0.063},

[0134] I m =diag{0.0536,0.0526,0.0526}, M=diag{3,3,3}, q=7, p=9, ρ=10,

[0135] K1=diag{10,10,10},K2=diag{1,1,1},K p =diag{35,35,35},K d =diag{0.5,0.5,0.5}.

[0136] The expected trajectory of this comparative experiment is: Circular trajectory: X c =-1+0.4·cos((1 / 10)πt), Y c =1+0.4·sin((1 / 10)πt), Z c =1.

[0137] Figure 2: It is a comparison diagram of tracking the flat circle trajectory in the XY plane by two control schemes, wherein the dotted line represents the tracking result of the control method (APD) implemented by the existing APD controller, and the solid line with stars represents the tracking result of the rope-driven parallel system intelligent control method (NTSM-CRCC) considering the spatial synchronization of the ropes of the present invention.

[0138] Figures 3 to 5 The following are the tracking error comparison diagrams of the X-axis, Y-axis, and Z-axis, respectively. The dotted line represents the tracking error of the control method (APD) implemented by the existing APD controller on the X-axis, Y-axis, or Z-axis, and the dashed line represents the tracking error of the rope-driven parallel system intelligent control method (NTSM-CRCC) considering the spatial synchronization of the rope in the present invention on the X-axis, Y-axis, or Z-axis. Figures 3 to 5 It can be seen from the figure that the method proposed in the present invention has a very significant improvement in the tracking accuracy of the desired trajectory, and the tracking errors of each axis have been significantly reduced, which can prove the effectiveness and superiority of the intelligent control method of the rope-driven parallel system considering the spatial synchronization of the rope proposed in the present invention. Figures 3 to 5 It can be seen that compared with the prior art, the X-axis tracking error of the present invention is reduced by 80%, the Y-axis tracking error is reduced by 78.4%, and the Z-axis tracking error is reduced by 62.7%.

[0139] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. An intelligent control method for a rope-driven parallel system considering the spatial synchronization of ropes, characterized in that: The method comprises the following steps: Step 1: Use the dynamic model of the rope-driven parallel system to calculate the theoretical length of each rope in the rope-driven parallel system; according to the theoretical length of each rope in the rope-driven parallel system, construct the theoretical deviation coupling error vector e in the rope space c-rc ; e c-rc =[e c-rc1 ...e c-rcn ] T ; Among them, e c-rci is the theoretical deviation coupling error corresponding to the i-th rope, i = 1, 2, ..., n; The theoretical deviation coupling error vector e c-rc All theoretical deviation coupling errors in form a ring topology structure, specifically: in, l di is the expected length of the i-th rope; e ci is the theoretical tracking error of the length of the i-th rope, and e ci = l i -l di ; l i is the theoretical length of the i-th rope; Step 2: Use the theoretical deviation coupling error vector e c-rc Construct an intelligent synchronization controller that takes into account the spatial synchronization of ropes; Step 3: Use deep reinforcement learning algorithm and combine theoretical deviation coupling error vector e c-rc , train the deep neural network to obtain the trained deep neural network; Step 4: Use the actual length of each rope in the rope-driven parallel system collected at the current moment to construct the actual deviation coupling error vector e′ in the rope space c-rc ; Step 5: Couple the actual deviation error vector e′ c-rc As the input of the trained deep neural network, the actual deviation coupling error vector e′ is obtained by the trained deep neural network. c-rc After recognition, output the current action a′; Step 6: Use the current action a′ to calculate the proportional parameter matrix K of the intelligent synchronous controller obtained in step 2 cp and the differential parameter matrix K cd Update and obtain the updated scale parameter matrix K′ cp and the updated differential parameter matrix K′ cd ; Step 7: The intelligent synchronous controller is based on the updated proportional parameter matrix K′ cp , the updated differential parameter matrix K′ cd , and the actual deviation coupling error vector e′ obtained in the rope space in step 4 c-rc Generate actual deviation control signal u′ crcc ; Then control the signal u′ according to the actual deviation crcc Generate synchronization control signal u tc-crcc Control the multi-rope parallel drive system; Among them, u tc-crcc =u tc -u′ crcc ,u tc It is the control signal output by the basic controller.

2. The intelligent control method for a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: In step 2, the model of the intelligent synchronous controller is constructed as follows: K cp =diag{K cp1 ,K cp2, ,...,K cpn }; K cd =diag{K cd1 ,K cd2, ,...,K cdn }; Among them, u crcc is the theoretical deviation control signal output by the intelligent synchronous controller. is the theoretical deviation coupling error vector e c-rc The derivative of K cpi is the scale parameter matrix K cp The i-th component in K cdi is the differential parameter matrix K cd The i-th component in , i = 1, 2, ..., n.

3. The intelligent control method of a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: In step 3, a deep reinforcement learning algorithm is used and combined with the theoretical deviation coupling error vector e c-rc , train the deep neural network, and obtain the implementation methods of the trained deep neural network including: Step 31: Construct a Markov decision process, specifically: First, the theoretical deviation coupling error vector e c-rc Defined as the state of the Markov decision process x(m) = e c-rc (m); where e c-rc (m) represents the theoretical deviation coupling error vector e c-rc The result of the mth step after discretization; x(m) is the state of the mth control cycle; Secondly, using K cp and K cd The vector composed of the components in is defined as the action a(m) of the Markov decision process = [K cp1 ... K cpn K cd1 ... K cdn ] T ; Among them, K cpi is the scale parameter matrix K cp The i-th component in K cdi is the differential parameter matrix K cd The i-th component in , i = 1, 2, ..., n; a(m) is the action of the m-th control cycle; Finally, based on the state x(m) and action a(m) of the Markov decision process, the Markov decision process is described as: The control cost function of the Markov decision process Discount factor γ∈[0,1) for the Markov decision process; in, represents the state space, represents the action space; represents the probability of state x(m) transferring to x(m+1) when action a(m) is taken in state x(m); W is a positive definite weight matrix; Step 32: Use reinforcement learning algorithm to train the deep neural network. During the training process, the state x(m) of the Markov decision process is used as the input of the deep neural network, and the action a(m) of the Markov decision process is used as the output of the deep neural network.

4. The intelligent control method for a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: In step 5, e′ c-rc =[e′ c-rc1 ...e′ c-rcn ] T ; Among them, e′ c-rci is the actual deviation coupling error corresponding to the i-th rope, i = 1, 2, ..., n; In the actual deviation coupling error vector e′ c-rc All actual deviation coupling errors in form a ring topology structure, specifically: in, l di is the expected length of the i-th rope; e′ ci is the actual tracking error of the length of the i-th rope, and e′ ci = l′ i -l di ; l′ i is the actual length of the i-th rope.

5. The intelligent control method for a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: In step 5, a′=[K′ cp1 …K′ cpn K′ cd1 …K′ cdn ] T ; K′ cpi is the scale parameter matrix K cp The optimal value of the i-th component in K′ cdi is the differential parameter matrix K cd The optimal value of the i-th component in .

6. The intelligent control method for a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: Step 1: Calculate the theoretical length of each rope in the rope-driven parallel system using the rope-driven parallel system dynamics model. The implementation method includes: Step 1: According to the dynamic model of the rope-driven parallel system, the speed of the end effector in the rope-driven parallel system is obtained. And the speed of the end effector Integrate to obtain the theoretical position of the end effector X = [b Τ ψ Τ ] Τ ;in, b=[x o y o z o ] Τ , b is the translation vector of the end effector, x o ,y o and z o are the positions of the end effector on the X, Y and Z axes of the world coordinate system O-XYZ, respectively; ψ=[α o β o γ o ] Τ , ψ is the rotation vector, α o , β o and γ o They are the rotation angles around the X, Y and Z axes of the world coordinate system O-XYZ; Step 1 and 2: According to the theoretical position X of the end effector, obtain the theoretical position vector l of the rope i ; l i =b+R·a i -p i ; Among them, l i is the theoretical position vector of the i-th rope, R is the rotation matrix between the world coordinate system O-XYZ and the body coordinate system B-xyz of the end effector, a i Point B i Coordinate vector in the body coordinate system B-xyz, i = 1, 2, ..., n, B i is the connection point of the i-th rope on the end effector, p i Point P i Coordinate vector in the world coordinate system O-XYZ, P i is the exit point of the i-th rope on the pulley; Step 13: According to the rope theory position vector l i , and the theoretical length of each rope is obtained i =||l i ||2; where ||·||2 is the two-norm.

7. The intelligent control method for a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: The dynamic model of the rope-driven parallel system is: in, u is the motor torque as the input of the system, M is the positive definite symmetric inertia matrix, J is the Jacobian matrix, I m and F v are the inertia matrix and viscous friction matrix of the rope winding device, R T is the transmission ratio from motor angle to rope length, is the Coriolis matrix, G is the gravity vector, and are the velocity and acceleration vectors of the end effector, (J T ) + It's J T The left pseudo-inverse matrix, (J T ) + =(JJ T ) -1 J, M eq , C eq and H eq are the first to third intermediate variable matrices respectively.

8. The intelligent control method for a rope-driven parallel system considering rope spatial synchronization according to claim 1 is characterized in that: In step seven, in, is e′ c-rc The derivative of .

Images