Tower crane adaptive anti-sway obstacle avoidance feedback control method and system based on neural network

Abstract

The application discloses a tower crane self-adaptive anti-swing obstacle avoidance feedback control method and system based on a neural network, relates to technical control of an electromechanical system, and comprises the following steps: a dynamics model of a four-degree-of-freedom tower crane system is established by using a Lagrange method; a self-adaptive anti-swing obstacle avoidance feedback controller is constructed with the control targets of target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion region constraint and speed constraint, and the controller comprises the following steps: the tower crane obstacle avoidance problem is converted into a state constraint problem, swing constraint terms, obstacle avoidance constraint terms, physical limit constraint terms and speed limit constraint terms and auxiliary terms for describing uncertain parameters and external disturbances in the system are constructed; wherein, a radial basis neural network is used to approximate the auxiliary terms of the tower crane, and the network weight matrix of the approximation is optimized according to an update rate; the control of the controller is used to control the cantilever rotation angle and the trolley displacement on the cantilever of the tower crane, so that safe, effective, accurate and rapid anti-swing obstacle avoidance movement control is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electromechanical system control, and particularly relates to a tower crane adaptive anti-sway obstacle avoidance feedback control method and system based on a neural network. BACKGROUND

[0002] In actual construction sites or ports, tower cranes (referred to as tower cranes) play an important role due to their high operating efficiency, large working space and strong payload capacity. In actual construction sites, obstacles such as buildings and heavy equipment inevitably appear, which poses a great challenge to the safety and efficient operation of the crane. At present, tower cranes still mainly rely on manual operation, and the operation is controlled by the experience of the operator. However, this method has the problems of low work efficiency, poor anti-swing ability and high accident risk. Therefore, it is of great significance to research an effective automatic control system for tower cranes. In fact, during the transportation process using a tower crane, the acceleration or deceleration of the trolley (or trolley) or the cantilever may cause the attached cargo to swing due to the inherent inertia, thereby reducing the transportation efficiency and causing safety hazards. In addition, due to the inherent underactuated characteristics, strong coupling and high nonlinearity of the tower crane, it is very difficult to design an anti-swing control system. Furthermore, even if the cargo swing is effectively suppressed, the cargo may still collide with the obstacles on the transportation path, thereby causing accident risks. Therefore, how to design an effective control method to ensure accurate positioning and eliminate swing while achieving automatic obstacle avoidance is a serious challenge.

[0003] In order to avoid collision with obstacles, current researches mostly rely on heuristic path planning algorithms of mobile robots to design collision-free path planning methods for cranes, such as using particle swarm optimization algorithm, RRT path planning algorithm, probabilistic roadmap algorithm and A* algorithm to generate collision-free paths for cranes. Although these heuristic path planning algorithms can provide optimal collision-free lifting paths for cranes, they do not take into account the underactuated characteristics of the crane and the actual needs of swing suppression. Further, there are also studies that mainly use trajectory planning methods to solve the problems of obstacle avoidance and swing suppression at the same time, such as a rotating crane trajectory planning algorithm based on a bisection method, an optimal controller for collision-free strategy of double-line hammerhead crane, and a tower crane adaptive input shaper with obstacle avoidance function, etc.

[0004] The current obstacle avoidance algorithms / methods for various cranes have achieved relatively good results in terms of swing suppression and obstacle avoidance. However, these methods require a large amount of time to solve the motion trajectory, and even in the same construction environment, if the initial position and target position change, the motion trajectory needs to be re-planned, resulting in time waste. In addition, due to the lack of state feedback, these methods do not have strong robustness to external disturbances and parameter uncertainties, which may cause safety hazards and accidents. SUMMARY

[0005] To solve the above problems of the prior art, the application provides a tower crane adaptive anti-swing obstacle avoidance feedback control method and system based on a neural network.

[0006] In the first aspect, the application provides a tower crane adaptive anti-swing obstacle avoidance feedback control method based on a neural network.

[0007] The tower crane adaptive anti-swing obstacle avoidance feedback control method based on a neural network comprises the following steps.

[0008] A dynamics model of a four-degree-of-freedom tower crane system is established by using the Lagrange method.

[0009] Based on the dynamics model, an adaptive anti-swing obstacle avoidance feedback controller is constructed with the control targets of target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion region constraint and speed constraint. The construction comprises the following steps: converting the tower crane obstacle avoidance problem into a state constraint problem, constructing a swing constraint term, an obstacle avoidance constraint term, a physical limit constraint term and a speed limit constraint term, and an auxiliary term for describing uncertain parameters and external disturbances in the system. The auxiliary term of the tower crane is approximated by using a radial basis function (RBF) neural network, and the network weight matrix of the approximation is optimized according to an update rate.

[0010] The controller is used to control the rotation angle of the cantilever and the displacement of the trolley on the cantilever of the tower crane, so as to realize anti-swing obstacle avoidance movement control.

[0011] In the second aspect, the application provides a tower crane adaptive anti-swing obstacle avoidance feedback control system based on a neural network.

[0012] The tower crane adaptive anti-swing obstacle avoidance feedback control system based on a neural network comprises the following steps.

[0013] A model building module is configured to establish a dynamics model of a four-degree-of-freedom tower crane system by using the Lagrange method.

[0014] A controller construction module is configured to construct an adaptive anti-swing obstacle avoidance feedback controller based on a dynamic model, with target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion region constraint and speed constraint as control targets, the construction comprising: converting the tower crane obstacle avoidance problem into a state constraint problem, constructing swing constraint terms, obstacle avoidance constraint terms, physical limit constraint terms and speed limit constraint terms, and auxiliary terms for describing uncertain parameters and external disturbances in the system, wherein a radial basis function (RBF) neural network is used to approximate the auxiliary terms of the tower crane, and the network weight matrix of the approximation is optimized according to an update rate.

[0015] A control module is configured to control the slewing angle of the jib and the displacement of the trolley on the jib of the tower crane by using the controller, to realize anti-swing obstacle avoidance movement control.

[0016] In a third aspect, the present application further provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and running on the processor, when the computer instructions are run by the processor, the steps of the method of the first aspect are completed.

[0017] In a fourth aspect, the present application further provides a computer readable storage medium for storing computer instructions, when the computer instructions are executed by a processor, the steps of the method of the first aspect are completed.

[0018] The above one or more technical solutions have the following beneficial effects:

[0019] 1. The present application provides a neural network-based tower crane adaptive anti-swing obstacle avoidance feedback control method and system, which utilizes the coupling characteristics of the tower crane to convert the obstacle avoidance task into a state constraint control problem, and designs multiple state constraint auxiliary terms to limit the load and trolley position within the collision-free range, realizes load obstacle avoidance, and controls the trolley and jib within the motion range and speed limit, realizes safe operation, simultaneously uses neural network approximation of the uncertain / unknown dynamics of the crane system, to construct an adaptive anti-swing obstacle avoidance feedback controller, uses the controller for tower crane operation control, simultaneously solves key and practical application-oriented control problems including obstacle avoidance, uncertain disturbance, state constraint and nonlinear dynamics, realizes precise positioning, rapid swing suppression and obstacle avoidance, the entire control process is safe and effective, has a shorter transportation time and stronger unknown disturbance robustness. On this basis, the convergence and stability of the proposed control method and system are theoretically proved by using Lyapunov stability theory, and the effectiveness of the proposed controller is verified by hardware experiments. Compared with existing methods, the proposed method can reduce the transportation time by 70% under the same external disturbance and uncertainty, greatly improving the transportation efficiency.

[0020] 2、In the application, based on the nonlinear feedback control method, the tower crane obstacle avoidance problem is converted into the state constraint problem of the tower crane, based on the constructed state constraint term, the nonlinear controller can limit the trolley and boom within the no-collision area to achieve obstacle avoidance, instead of tracking the no-collision reference trajectory, which eliminates the need for motion trajectory planning, and significantly improves the transportation efficiency.

[0021] 3、In the application, according to the coupling characteristics of the tower crane, an underactuated swing constraint term is designed to limit the load swing angle within a predetermined range, avoid collision between the load swing and obstacles, and effectively suppress the load swing during transportation.

[0022] 4、In the designed control method, the radial basis function (RBF) neural network is introduced to approximate the uncertain / unknown dynamic function, which effectively improves the positioning accuracy of the trolley / cantilever, improves the robustness of the controller to external disturbances and uncertain parameters, and improves the safety performance. BRIEF DESCRIPTION OF DRAWINGS

[0023] The drawings accompanying the specification of this application form a part thereof, serve to further provide a further understanding of the application, and together with the description of the exemplary embodiments of the application and the explanation thereof serve to explain the application, and do not constitute an improper limitation of the application.

[0024] Figure 1 It is a structural schematic diagram of the tower crane in the embodiment of the application;

[0025] Figure 2 It is a schematic diagram of the tower crane adaptive anti-swing obstacle avoidance feedback control method in the embodiment of the application;

[0026] Figure 3 It is a comparison diagram of the anti-swing experimental results of the control method (solid line) and the sliding mode control method (dashed line) in the embodiment of the application;

[0027] Figure 4 It is an obstacle avoidance experimental result diagram of the control method (solid line) and the trajectory planning method (dashed line) in the embodiment of the application, wherein (a) is an obstacle avoidance route, and (b) is the result of the state variable of the tower crane;

[0028] Figure 5 It is an obstacle avoidance experimental result diagram of the control method in the embodiment of the application in a multi-obstacle environment, wherein (a) is an obstacle avoidance route diagram in different obstacle environments, and (b) is the change of the state variable in different obstacle environments. DETAILED DESCRIPTION

[0029] It should be noted that the following detailed description is exemplary only and is intended to provide further description of the present application in order to provide further explanation of the exemplary embodiments according to the present application and is not intended to limit the same. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application pertains. Furthermore, it should be understood that the use of the term "including", "comprising", or "having" herein is intended to indicate the presence of the features, steps, operations, devices, components, and / or combinations thereof described herein.

[0030] Embodiment One

[0031] The embodiment provides a tower crane adaptive anti-swing obstacle avoidance feedback control method based on a neural network, and the method comprises the following steps of:

[0032] A dynamics model of a four-degree-of-freedom tower crane system is established by using a Lagrange method.

[0033] Based on the dynamics model, an adaptive anti-swing obstacle avoidance feedback controller is constructed with the control targets of target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion region constraint and speed constraint. The construction comprises the following steps: converting the tower crane obstacle avoidance problem into a state constraint problem, constructing a swing constraint term, an obstacle avoidance constraint term, a physical limit constraint term and a speed limit constraint term, and an auxiliary term for describing uncertain parameters and external disturbances in the system; wherein, the auxiliary term of the tower crane is approximated by using a radial basis function (RBF) neural network, and the network weight matrix of the approximation is optimized according to an update rate.

[0034] The controller is used to control the cantilever rotation angle and the trolley displacement on the cantilever of the tower crane, so as to realize anti-swing obstacle avoidance movement control.

[0035] The tower crane adaptive anti-swing obstacle avoidance feedback control method based on a neural network provided in the embodiment is described in more detail through the following content.

[0036] In the method provided in the embodiment, an adaptive anti-swing obstacle avoidance feedback controller is first constructed, and then the controller is used to control the tower crane, so as to realize anti-swing obstacle avoidance movement. The construction and application of the controller are the key points of the embodiment, and specifically comprise the following steps:

[0037] (1) A dynamics model of a four-degree-of-freedom tower crane system is established by using a Lagrange method, and analysis is performed based on the model. The constructed dynamics model comprises two driving state variables and two non-driving state variables, wherein the driving state variables comprise a cantilever rotation angle and a trolley displacement on the cantilever, and the non-driving state variables comprise a swing angle of a load pulled by the trolley in a three-dimensional space, which comprises a load swing angle θ1 in a horizontal direction of the cantilever and a load swing angle θ2 in a vertical direction of the cantilever.

[0038] Specifically, asFigure 1 The schematic diagram of tower crane structure in three-dimensional environment is shown, and the related physical parameters and definitions of the tower crane system are shown in Table 1.

[0039] Table 1 Related physical parameters and definitions of the tower crane system

[0040]

[0041] Based on the Lagrange method, the dynamics equation of the kinematics model of the three-dimensional four-degree-of-freedom tower crane is established as follows:

[0042]

[0043]

[0044] To simplify the expression, the above dynamics equations (1) to (4) can be simplified into the following matrix form:

[0045]

[0046] wherein,

[0047]

[0048] In the above equation, the state variable is composed of the driving term vector and the underdrive vector , M(q), and G(q) are the inertia matrix, the Coriolis matrix and the gravity vector respectively; is the control input vector, wherein The specific expression form in the above matrix is as follows:

[0049]

[0050] wherein,

[0051] m 11 = ml 2 (sin 2 θ1cos 2 θ2+sin 2 θ2)+2mlβsinθ1cosθ2+(m+M)β 2 +J, m 12 =-ml2sinθ2,

[0052] m 13 =-ml 2 cosθ1sinθ2cosθ2,m 14 =ml(cosθ2β+lsinθ1),m 22 =M+m,m23 = mlcosθ1cosθ2

[0053] m 24 = -mlsinθ1sinθ2, m 33 = ml 2 cos 2 θ2, m 34 = 0, m 44 = ml 2 (8)

[0054]

[0055] Considering that the load is generally below the crane jib, the following assumptions are made:

[0056] θ i ∈(-π2,π2), i = 1, 2. (10)

[0057] (2) Based on the kinematic model of the tower crane, an adaptive anti-swing obstacle avoidance feedback controller is constructed with the control objectives of target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion region constraint and speed constraint. In this construction process, first, the obstacle avoidance problem of the tower crane is converted into a state constraint problem, that is, according to the geometric relationship among the load (also called load), the trolley (i.e. the trolley) and the obstacle, the motion constraint range of the trolley and the load is generated respectively to ensure collision-free motion of the load.

[0058] In this embodiment, an adaptive nonlinear controller is designed to simultaneously achieve positioning, anti-swing and obstacle avoidance in the presence of parameter uncertainty and external disturbance, and to ensure the safe operation of the trolley and the jib as the actuator. Specifically, the control objectives can be divided into the following six subtasks:

[0059] (2.1) Precise target positioning. The controller controls the jib to rotate to the target angular position α d and the control console to move to the target position β d within a set time, and the mathematical expression is:

[0060]

[0061] Where t represents time.

[0062] (2.2) Swing suppression. The controller suppresses the remaining swing amplitude of the effective load to 0 when the load reaches the target position within a set time, that is, when the effective load reaches the target position, the remaining swing amplitude of the effective load is suppressed, which can be described by mathematical method as:

[0063]

[0064] (2.3) Swing angle constraint. The controller controls the load (i.e., payload) swing angle within a set safety range, which is the range between 0 and the maximum swing angle, i.e.,

[0065]

[0066] where, is the maximum swing angle of θ1(t) and θ2(t).

[0067] (2.4) Obstacle avoidance. The controller controls the Euclidean distance between the load and the obstacle to be greater than or equal to the radius of the obstacle enclosing box. That is, during transportation, the payload and the trolley should avoid collision with the obstacle, which is mathematically expressed as:

[0068]

[0069] where, ω i is the radius of the obstacle enclosing box, satisfying ω i > l sin θ m + r i ; p i (x, y) is the Euclidean distance between the payload and the i th th obstacle, defined as:

[0070]

[0071] where, x = β cos a and y = β sin a are the horizontal and vertical coordinates of the trolley, (x obs,i , y obs,i ) and r i are the center coordinates and radius of the i th th obstacle.

[0072] With the above defined swing angle constraint, when the maximum swing angle of the payload is θ m , according to the geometric relationship between the payload and the trolley, the maximum distance between the payload and the trolley is l sin θ m . To ensure that the payload will not collide with the obstacle due to accidental swing, ω i as the radius of the circular enclosing box of the i th th obstacle, should satisfy the constraint condition: ω i > l sin θ m + r i . That is, when the trolley is at least ω away from the center of the obstacle and the maximum swing angle of the payload does not exceed θ m , it can be guaranteed that the payload will not collide with the obstacle.

[0073] (2.5) Motion region constraint. The controller controls the actuators (i.e., the trolley and the boom) and limits their motion within a specific interval, defined as:

[0074] α min <α<α max ,β min <β<β max (16)

[0075] wherein, α min , α max , β min , β max are the upper and lower bounds of the physical limits of the cantilever and the trolley, respectively.

[0076] (2.6) Velocity constraint. The controller controls the actuators (i.e. the trolley and the cantilever) and limits their movement speed within a safe speed range, defined as:

[0077]

[0078] wherein, α d , β d are the boundaries of the speed limits of the cantilever and the trolley, respectively.

[0079] (3) Secondly, the obstacle avoidance constraint term, the swing angle constraint term, the physical limit constraint term and the speed limit constraint term are constructed. The swing constraint term and the obstacle avoidance constraint term constructed as the nonlinear control part of the controller, respectively limit the trolley within the obstacle-free area and limit the swing angle within the preset range, for realizing the swing angle constraint and obstacle avoidance; similarly, the physical limit constraint term and the speed limit constraint term constructed as the nonlinear control part of the controller, respectively limit the trolley and the cantilever within the physical limit area (such as the track length limit, construction area limit, etc.) and the safe speed range, for realizing the safe operation of the trolley and the cantilever. By introducing the above constraint terms, the controller can automatically drive the trolley and the payload to effectively avoid obstacles in the working space, and prevent the actuators from exceeding the physical limit area, while ensuring the running speed of the actuators within the safe range, realizing automatic obstacle avoidance and safe operation.

[0080] Specifically, the construction of the above swing constraint term, obstacle avoidance constraint term, physical limit constraint term and speed limit constraint term is: according to the target value and the actual value of the swing angle of the load in the three-dimensional space, an error term is constructed; based on the error term, the swing constraint term, the obstacle avoidance constraint term, the physical limit constraint term and the speed limit constraint term are constructed, as shown in Figure 2 , including:

[0081] Firstly, the error term and its derivative are defined, and the specific expression forms are:

[0082]

[0083] In the above formula, α d , βd respectively represent the target position of the jib and the trolley.

[0084] To solve the problem of obstacle avoidance and load swing of the tower crane, the following auxiliary constraint terms are constructed to realize obstacle avoidance and load swing angle constraint:

[0085] (3.1) Construct swing constraint term The load swing angle is limited within a certain range, defined as:

[0086]

[0087] (3.2) Construct obstacle avoidance constraint term defined as:

[0088]

[0089] In the above formula, k c is the tower crane obstacle avoidance parameter, which can be adjusted according to the actual use; for the convenience of representation, the obstacles in the embodiment are represented by the surrounding circle of the obstacle, that is, an obstacle is represented by a circle surrounding the obstacle, ω represents the radius length of the surrounding circle, which can ensure that the trolley will not enter the surrounding circle area.

[0090] (3.3) Construct physical restriction constraint term Limit the trolley and jib within the range limited by the physical constraint, defined as:

[0091]

[0092] In the above formula, k wα and k wβ are physical restriction parameters, which determine the size of the restriction force, and can be adjusted according to the actual use.

[0093] (3.4) Construct speed limit constraint term of actuator Limit the speed of the actuator within a certain range, defined as:

[0094]

[0095] In the above formula, k rα and k rβ are speed limit parameters, which can be adjusted according to the actual use.

[0096] (4) Construct an auxiliary term for describing the uncertain parameters and external disturbances in the system, and use a radial basis function (RBF) neural network to approximate the auxiliary term of the tower crane, and optimize the weight matrix of the approximate network according to the update rate. In this embodiment, the radial basis function (RBF) neural network is introduced into the control to process the parameter uncertainty and external disturbance, especially for eliminating the influence of uncertain parameters on the obstacle avoidance constraint term and the swing angle constraint term, ensuring the accuracy of obstacle avoidance, effectively improving the positioning accuracy of the trolley and the cantilever, and improving the robustness of the control system.

[0097] Specifically, for stability analysis, an auxiliary term Φ is proposed, which is defined as:

[0098]

[0099] wherein, is a unit matrix, Φ0 is a known nominal value, and Φ Δ is an unknown part of the crane system. The unknown / unavailable parameters / structures of the tower crane are approximated by a radial basis function (RBF) neural network, which is represented as:

[0100]

[0101] wherein, is the input of the neural network, is the approximation error satisfying , W * is the optimal weight matrix, which can be represented as:

[0102]

[0103] In addition, φ(x) represents an activation function vector, and in this embodiment, the following Gaussian function is selected, which is represented as:

[0104]

[0105] wherein, c represents a center vector, and b represents a standard deviation vector.

[0106] Further, the controller constructed in this embodiment can be represented as:

[0107]

[0108] wherein, is defined as K p , K z , K d , K A , K I , is a diagonal matrix with positive control gain, is W *the estimation matrix, whose update rate is

[0109]

[0110] where Γ represents a diagonal matrix with positive parameters.

[0111] In addition, part of the control gain in the controller also satisfies the following conditions:

[0112]

[0113] Finally, in order to prove the stability of the above-mentioned constructed controller, the following Lyapunov function is designed, and the stability and convergence of the proposed controller are verified by using the Lyapunov function. The function is:

[0114]

[0115] Taking the derivative of the above Lyapunov function, we get:

[0116]

[0117] where It can be arranged in the following matrix form:

[0118]

[0119] According to the dynamic equation, we can get:

[0120]

[0121] Therefore, It can be derived that:

[0122]

[0123] After that, the designed controller and update rate are brought in, and we can get:

[0124]

[0125] Therefore, it can be deduced that:

[0126]

[0127] Based on the above calculation results, it can be seen that as long as the initial swing angles θ1(0) and θ2(0) are within the preset range (-θ m ,θ m ), there must be a certain moment T If the swing angle tends to exceed the preset range, there must be driving variables θ1(T) and θ2(T) close to the preset boundary, which makes or As time t tends to infinity, we can conclude that V(t)→+∞, which contradicts the result Thus, during the whole control process, the load displacement θ1and θ2will not exceed the preset range, i.e.

[0128]

[0129] Similarly, if the initial position of the trolley is within the preset range, A α and A β will tend to infinity as p i (x,y)→ω i , i.e., V(t)→+∞, which contradicts the conclusion Therefore, we have

[0130] p i (α(t),y(t))≥ω i (42)

[0131] Further, it can be determined that the swing angle and the trolley will only be in a specific area, which can ensure that the load will not collide with the obstacle.

[0132] Similarly, if the trolley and the cantilever approach the physical limit area α max and β max , there must exist V(t)→+∞, which contradicts the conclusion Therefore, we have

[0133] α min <α<α max ,β min <β<β max (43)

[0134] Next, integrating (39) gives:

[0135]

[0136] Here, we discuss two cases:

[0137] The first case: when , suppose there exists a time t∈[T1,T2], if there is a point approaching the speed limit boundary value and , then there must be:

[0138]

[0139] This contradicts the conclusion of equation (39), so the speed of the actuator must not exceed the set value, i.e., the speed must be within the set range ​

[0140] The second case: when The velocity of the actuator must be within the set region

[0141] Therefore, the above two cases can be combined to get the velocity of the actuator will not exceed the set region, that is:

[0142]

[0143] Further, to prove that the state variables converge to the target value at the equilibrium point, define an invariant set, which is:

[0144]

[0145] The invariant set has a maximum invariant subset In There is:

[0146]

[0147] In addition, it also includes:

[0148]

[0149] Similarly, the following results can be obtained:

[0150]

[0151] According to the dynamics expression and the controller, the following results can be obtained:

[0152] T = 0, F = 0 . (51)

[0153] Therefore, the underactuated part of the dynamics equation (i.e. the non-driven part) can be written as:

[0154]

[0155] Bring the above two equations into the driven part of the dynamics equation, we can get:

[0156]

[0157] Further, we can get:

[0158]

[0159] Then, according to and cosθ2>0,cosθ1>0, we can get:

[0160]

[0161] Similarly, we can get:​

[0162]

[0163] Therefore, according to the above conclusion, it can be concluded that only the equilibrium point of the closed-loop system is in the subset S Therefore, it can be ensured that the tower crane can complete the transportation task, avoid obstacles, and the swing angle is 0 after the load reaches the target position.

[0164] Further, in order to ensure the effectiveness of the method proposed in the embodiment, it is compared with the existing sliding mode control method. As shown in Figure 3 , both methods can achieve fast and accurate rotation / translation positioning and effective swing suppression, and it takes about 10 seconds for the rotating arm and trolley to reach the target and remain stable. However, the controller proposed in the embodiment can achieve more accurate positioning and smaller residual swing; in addition, the proposed method performs superior performance in terms of energy consumption, especially in terms of rotating arm movement energy consumption; and the method proposed in the embodiment has no obvious residual swing, and the safety is more, while the comparative sliding mode control method has obvious residual swing, which means that when reaching the target point, the comparative method still has swing, which may cause the occurrence of unexpected collision.

[0165] In order to verify the obstacle avoidance performance of the method proposed in the embodiment, the effectiveness of the method is verified by comparison with the trajectory planning method. As shown in Figure 4 , in (a), the two obstacles are represented by a circle, it can be seen that both methods can effectively avoid obstacles and accurately reach the target position, but the trajectory planning method needs a lot of offline solving time, combined with (b) shown, the method significantly reduces the transportation time by 70% in the calculation platform of Intel Core-i5 CPU 3.6GHzx4and RAM of 16GB, which includes offline calculation time.

[0166] Finally, the obstacle avoidance experiment in the multi-obstacle environment is also carried out, as shown in Figure 5 (a), (b), it can be seen that the method can effectively avoid multiple obstacles while transporting the load to the target point, and effectively suppresses the swing, and the swing of the load does not exceed the limit value during the entire transportation process. After reaching the target point, the method proposed in the embodiment hardly participates in the swing, which ensures the stability and safety of the tower crane transportation.

[0167] That is, compared with the prior art sliding mode controller and the trajectory planning method, the controller and control method proposed in the embodiment have higher positioning accuracy and more effective swing suppression performance.

[0168] Embodiment Two

[0169] The embodiment provides a tower crane adaptive anti-swing obstacle avoidance feedback control system based on a neural network, which comprises:

[0170] A model building module is configured to build a dynamic model of a four-degree-of-freedom tower crane system by using a Lagrange method.

[0171] A controller building module is configured to build an adaptive anti-swing obstacle avoidance feedback controller based on the dynamic model, with target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion area constraint and speed constraint as control targets. The building comprises: converting the tower crane obstacle avoidance problem into a state constraint problem, building swing constraint terms, obstacle avoidance constraint terms, physical limit constraint terms and speed limit constraint terms and auxiliary terms for describing uncertain parameters and external disturbances in the system. The auxiliary terms of the tower crane are approximated by using a radial basis function (RBF) neural network, and the network weight matrix of the approximation is optimized according to an update rate.

[0172] A control module is configured to control the rotation angle of the cantilever and the displacement of the trolley on the cantilever of the tower crane by using the controller, so as to realize anti-swing obstacle avoidance movement control.

[0173] Embodiment three

[0174] The embodiment provides an electronic device, which comprises a memory and a processor, and computer instructions stored in the memory and running on the processor. When the computer instructions are run by the processor, the steps of the method for adaptive anti-swing obstacle avoidance feedback control of a tower crane based on a neural network are completed.

[0175] Embodiment four

[0176] The embodiment also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the method for adaptive anti-swing obstacle avoidance feedback control of a tower crane based on a neural network are completed.

[0177] The steps and method of the above embodiments two to four correspond to the method of embodiment one, and the specific embodiments can be referred to the related description part of embodiment one. The term "computer-readable storage medium" should be understood as including a single medium or multiple media of one or more instruction sets; and should also be understood as including any medium capable of storing, encoding or carrying instruction sets for execution by a processor and causing the processor to execute any method in the present application.

[0178] Those skilled in the art should understand that the modules or steps of the present application described above can be realized by a general computer device, alternatively, they can be realized by program codes executable by a computing device, so that they can be stored in a storage device and executed by a computing device, or they can be respectively made into individual integrated circuit modules, or a plurality of modules or steps among them can be made into a single integrated circuit module to realize. The present application is not limited to any specific combination of hardware and software.

[0179] The above only describes the preferred embodiments of the present application, and the specific embodiments of the present application are described in conjunction with the drawings, but are not limited to the protection scope of the present application. Those skilled in the art should understand that various modifications or changes made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. A neural network-based adaptive anti-sway obstacle avoidance feedback control method for a tower crane, characterized in that, include: The dynamic model of the four-degree-of-freedom tower crane system is established using the Lagrangian method; Based on a dynamic model, an adaptive anti-sway and obstacle avoidance feedback controller is constructed with target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion range constraint, and speed constraint as control objectives. The construction includes: converting the tower crane obstacle avoidance problem into a state constraint problem, constructing an error term based on the target value and actual value of the load's swing angle in three-dimensional space; constructing a swing constraint term, an obstacle avoidance constraint term, a physical restriction constraint term, and a speed restriction constraint term based on the error term; and constructing auxiliary terms for describing uncertain parameters and external interference in the system. The auxiliary terms of the tower crane are approximated using a radial basis function (RBF) neural network, and the approximate network weight matrix is ​​optimized based on the update rate. The proposed controller is used to control the cantilever rotation angle and the displacement of the cantilever trolley of a tower crane to achieve anti-sway and obstacle avoidance motion control. The stability and convergence of the proposed controller are verified using the Lyapunov function. The controller is represented as: In the above formula, , defined as , , , , , , is a diagonal matrix with positive control gain, is the estimated matrix of the optimal weight matrix, and the update rate of the matrix is: , represents a diagonal matrix with positive parameters, represents an activation function vector, is a known nominal value, and the error term and its derivative , α and β are the cantilever angle, trolley displacement, swing constraint term , obstacle avoidance constraint term , state variable , is the control input vector.

2. The neural network-based adaptive anti-sway obstacle avoidance feedback control method for tower cranes according to claim 1, characterized in that, The constructed dynamic model includes driven state quantities and non-driven state quantities, among which the driven state quantities include the cantilever rotation angle and the trolley displacement on the cantilever, and the non-driven state quantities include the swing angle of the load pulled by the trolley in three-dimensional space, which includes the swing angle of the load in the horizontal direction of the cantilever and the swing angle of the load in the vertical direction of the cantilever.

3. The neural network-based adaptive anti-sway obstacle avoidance feedback control method for tower cranes according to claim 1, characterized in that, Target positioning control means controlling the cantilever to rotate to the target angular position and the control console area to move to the target position within the set time; The swing suppression control target is to suppress the remaining swing of the effective load to 0 when controlling the load to reach the target position within the set time; The swing angle constraint control goal is to control the load swing angle within a set safety range, which is the range between 0 and the maximum amplitude swing angle; The obstacle avoidance control goal is to control the Euclidean distance between the payload and the obstacle to be greater than or equal to the radius of the obstacle enclosing box; The motion range constraint control objective is to restrict the movement of the trolley and cantilever as the actuator to a specific range; The speed constraint control objective is to limit the movement speed of the trolley and cantilever as the actuator to a safe speed range.

4. The neural network-based adaptive anti-sway obstacle avoidance feedback control method for tower cranes according to claim 1, characterized in that, The swing constraint term and obstacle avoidance constraint term serve as the nonlinear control parts of the controller, respectively limiting the car to an obstacle-free area and limiting the swing angle to a preset range, so as to achieve swing angle constraint and obstacle avoidance; The physical restriction constraint item and the speed restriction constraint item serve as the nonlinear control part of the controller, which respectively restrict the trolley and the cantilever to operate within the physical restriction area and the safe speed range, so as to achieve safe operation of the trolley and the cantilever.

5. The neural network-based adaptive anti-swing obstacle avoidance feedback control method for tower cranes according to claim 1, characterized in that, The radial basis function (RBF) neural network is used to approximate the auxiliary items of the tower crane, which can be expressed as follows: In the above formula, is the input of the neural network, to meet approximation error, is the optimal weight matrix, , denotes the activation function vector; is the unknown part of the crane system, the velocity limit constraint term of the actuator , , and are the inertia matrix, the Coriolis matrix and the gravity vector respectively, , the driving term vector , α and β are the cantilever angle, the trolley displacement, the underactuated vector , θ 1 ,θ 2 for load three-dimensional space swing angle, is the control input vector, F is the trolley control force.

6. A neural network-based adaptive anti-sway obstacle avoidance feedback control system for a tower crane, characterized in that, include: Model building module, used to establish the dynamic model of the four-degree-of-freedom tower crane system using the Lagrangian method; A controller construction module is configured to construct an adaptive anti-swing obstacle avoidance feedback controller based on a dynamic model, with target positioning, swing suppression, swing angle constraint, obstacle avoidance, motion area constraint and speed constraint as control targets. The construction includes: converting the tower crane obstacle avoidance problem into a state constraint problem, constructing an error term according to the target value and actual value of the swing angle of the load in the three-dimensional space; constructing a swing constraint term, an obstacle avoidance constraint term, a physical limit constraint term and a speed limit constraint term based on the error term; constructing an auxiliary term for describing the uncertain parameters and external disturbances in the system; wherein the auxiliary term of the tower crane is approximated by a radial basis function (RBF) neural network, and the network weight matrix of the approximation is optimized according to an update rate; A control module is configured to control the rotation angle of the jib and the displacement of the trolley on the jib of the tower crane by using the controller, so as to realize anti-swing obstacle avoidance movement control; and the stability and convergence of the proposed controller are verified by using a Lyapunov function. The controller is represented as: In the above formula, , defined as , , , , , , is a diagonal matrix with positive control gain, is the estimation matrix of the optimal weight matrix, and the update rate of the matrix is: , is a diagonal matrix with positive parameters, is an activation function vector, is a known nominal value, and the error term and its derivative , α and β are the cantilever angle, trolley displacement, swing constraint term , obstacle avoidance constraint term , state variable , is the control input vector.

7. An electronic device, comprising: A computer program product is provided, which comprises a memory and a processor, and computer instructions stored in the memory and running on the processor, and when the computer instructions are run by the processor, the steps of the neural network-based adaptive anti-swing obstacle avoidance feedback control method for a tower crane according to any one of claims 1-5 are completed.

8. A computer-readable storage medium, characterized in that, A computer program product is provided, which is configured to store computer instructions, and when the computer instructions are executed by a processor, the steps of the neural network-based adaptive anti-swing obstacle avoidance feedback control method for a tower crane according to any one of claims 1-5 are completed.

Images