Mecanum wheel AGV dynamics adaptive inversion control method

The AGV dynamics model is approximate and adaptive inversion control through ELM neural network, which solves the trajectory tracking accuracy and stability of McNum Wheel AGV in complex environments, and realizes more efficient control system stability and interference cancellation capabilities. .

CN118915439APending Publication Date: 2024-11-08SOUTH CHINA UNIV OF TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410952856.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

It is difficult to establish accurate mathematical models in complex operating environments, resulting in the impact of trajectory tracking accuracy and stability. The existing control algorithms rely on empirical parameter debugging and consume high computing resources, which is prone to overfitting problems.

Method used

The extreme learning machine (ELM) neural network is used to approximate the uncertainty and unknown interference of the AGV dynamics model, and design the adaptive inversion controller, and then decompose the system, design the intermediate virtual control quantity and Lyapunov function, and invert the inverting layer by layer. Effective control torque, improve the stability of the control system and the ability to respond to interference independently.

Benefits of technology

Accurate stable control at the AGV dynamics level is realized, training time is reduced, local optimal solution problems of gradient descent algorithms are avoided, stability and generalization capabilities of the control system are improved, and adaptive offsetting capabilities are enhanced for interference.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915439A_ABST
    Figure CN118915439A_ABST
Patent Text Reader

Abstract

The invention belongs to the field of AGV motion control, and discloses a Mecanum wheel AGV dynamics adaptive inversion control method, which comprises the steps of establishing an AGV kinematics model and a dynamics model, adopting an inversion method to control a Mecanum wheel AGV dynamics level, decomposing a system into two subsystems, and calculating a Mecanum wheel AGV kinematics model and a Mecanum wheel AGV kinematics model according to the Mecanum wheel AGV kinematics model and the Mecanum wheel AGV kinematics model. And respectively designing an intermediate virtual control quantity and a Lyapunov function for each subsystem, designing a corresponding control law according to a stability condition, and finally carrying out stack-up backstepping to obtain an equivalent control moment. And carrying out approximation on uncertainty and unknown interference of the AGV kinetic model by adopting an ELM neural network, and carrying out adaptive offset. According to the adaptive inversion controller designed by the invention, the interference is effectively simulated and adaptively eliminated in the AGV operation process by depending on the characteristics of rapid training and strong generalization of the ELM neural network, and the stability of the control system is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of AGV motion control, and in particular to a Mecanum wheel AGV dynamics adaptive inversion control method. Background Art

[0002] Automated Guided Vehicle (AGV) is not only an important part of the intelligent logistics system, but also an important embodiment of the level of automation. In the process of promoting unmanned workshops and intelligent warehouses, higher requirements are placed on the performance and application scope of AGVs. The omnidirectional AGV driven by Mecanum wheels has attracted widespread attention and application due to its flexible and stable movement, high precision, strong carrying capacity, and small space occupation.

[0003] Autonomous tracking of specific trajectories is an important part of AGV operation. In the context of intelligence, the operating environment of AGV is becoming more and more complex, and the accuracy requirements for AGV tracking process are becoming higher and higher. The omnidirectional AGV based on the Mecanum wheel chassis is a complex, strongly coupled, nonlinear system. Therefore, its control at the dynamic level is an important content. Since it is difficult to establish an accurate mathematical model for the Mecanum wheel AGV in actual operation, it is necessary to design a control algorithm that is less dependent on the mathematical model. Nowadays, such methods are common, such as neural network control, sliding mode control, adaptive control, inversion control, etc. In addition, there are inevitably unknown interferences, which affect the accuracy and stability of trajectory tracking. In addition to using conventional PID, robust control algorithms, etc. to adaptively eliminate interference, it is also possible to simulate model uncertainties and interference terms by designing interference observers, and then offset them.

[0004] With the development of intelligence, the use of neural networks to design interference observers is an increasingly popular method. A quad-rotor drone control method and system based on RBF neural network (Zhang Zexu) involves the field of drone control technology. The technical points of the present invention include: establishing a quad-rotor drone dynamics and kinematics model; splitting the quad-rotor drone dynamics and kinematics model into a position model and an attitude model, and using an RBF neural network-based interference observer to approximate the model uncertainty and external interference in the position model and attitude model, so that the real-time observed interference estimation error approaches zero, using backstepping control theory, designing a position controller according to the position model, and designing an attitude controller according to the attitude model; using the position controller and the attitude controller to control the quad-rotor drone.

[0005] Design of Adaptive Sliding Mode Controller for Mecanum Wheel Mobile Chassis (Jiang Menglin) An adaptive sliding mode trajectory tracking controller is designed based on the dynamic model of the Mecanum wheel mobile robot. The RBF neural network is used to online approximate the uncertainties in the dynamic model. Aiming at the trajectory tracking problem of the Mecanum wheel mobile robot, firstly, considering the external disturbance and parameter changes, the kinematic and dynamic models of the Mecanum wheel mobile robot are established, and an adaptive sliding mode trajectory tracking controller is designed based on the dynamic model. Secondly, in order to achieve better motion control, the radial basis function neural network is used to online approximate the uncertainties in the dynamic model. At the same time, the adaptive learning rule is used to estimate the unknown parameters. Furthermore, the stability and robustness of the control method are analyzed using the Lyapunov theory. Finally, the simulation results verify the effectiveness of the proposed controller.

[0006] The above technologies all use RBF neural network to design interference observers. Although traditional feedforward neural networks such as BP (Back Propagation) and RBF can well achieve adaptive approximation of interference, their parameters often rely on experience for debugging, usually requiring multiple iterations, consuming a lot of computing resources and time, and easily causing overfitting problems. Summary of the invention

[0007] The purpose of the present invention is to provide a Mecanum wheel AGV dynamic adaptive inversion control method based on ELM neural network, establish AGV kinematic model and dynamic model, adopt inversion method, control the Mecanum wheel AGV dynamic level, decompose the system into two subsystems, design intermediate virtual control quantity and Lyapunov function for each subsystem, design corresponding control law according to stability condition, and finally obtain equivalent control torque by stacking inversion. The ELM neural network is used to approximate the uncertainty and unknown interference of AGV dynamic model, and adaptively offset them, so as to better realize the precise and stable control of Mecanum wheel AGV dynamic level, and improve the autonomy of AGV in dealing with interference during actual operation.

[0008] The present invention is achieved by at least one of the following technical solutions.

[0009] A Mecanum wheel AGV dynamics adaptive inversion control method, comprising:

[0010] By conducting kinematic analysis and dynamic modeling on the four-wheel Mecanum wheel chassis AGV, the dynamic relationship between the AGV wheel speed and the motor torque is obtained, and the AGV dynamic inversion control law is designed to finally obtain the equivalent control torque.

[0011] The extreme learning machine (ELM) neural network is used to approximate the uncertainty and unknown disturbance of the AGV dynamic model and adaptively offset them to improve the stability of the control system.

[0012] Furthermore, the AGV dynamics inversion control includes the following steps:

[0013] Step 1: Kinematic analysis: Establish the velocity of the AGV center of mass in its own coordinate system [V x ,V y ,ω z ] T and the angular velocity of the four wheels [ω A ,ω B ,ω C ,ω D ] T the relationship between;

[0014] Step 2: Dynamic modeling: Establish a dynamic model that represents the relationship between the AGV motor drive torque and the wheel speed;

[0015] Step 3: Design the inverse control law: Use the inverse method to control the Mecanum wheel AGV dynamics, decompose the system into two subsystems, design the intermediate virtual control quantity and Lyapunov function for each subsystem, design the corresponding control law according to the stability conditions, and finally obtain the equivalent control law by stacking and backstepping.

[0016] Furthermore, in step 2, the Lagrange equation is combined with the total kinetic energy of the AGV during operation to obtain the nominal dynamic model of the AGV. At the same time, the rolling friction between the AGV and the ground is considered, and model uncertainty and external unknown disturbances are introduced. The final dynamic model is:

[0017]

[0018] in represents the overall disturbance, where ω is the wheel speed, is the angular acceleration of the wheel, T is the motor torque, M n is the inertia matrix, H n is the viscous friction coefficient matrix, ΔM and ΔH are the uncertain terms in the dynamic model, and d is the external unknown disturbance.

[0019] Furthermore, the nominal dynamic model is:

[0020]

[0021] in:

[0022]

[0023] Ti Indicates the torque magnitude of the four wheels. is the angular acceleration of the four wheels, i = A, B, C, D, m 车 is the AGV quality, J 车 and J 轮 are the AGV moment of inertia and wheel moment of inertia respectively, w is half of the distance between the geometric centers of the left and right wheels of the AGV, h is half of the distance between the geometric centers of the front and rear wheels of the AGV, and r is the wheel radius.

[0024] Furthermore, in step 3, the first subsystem of the inverse control is designed:

[0025] Define wheel speed ω, expected speed ω d and the speed error e, take the virtual control amount ω'=ω d -K1e, K1 is a constant and K1>0, virtual control error δ=ω-ω', design Lyapunov function:

[0026]

[0027] Push It is necessary to make δ=0.

[0028] Furthermore, in step 3, the second subsystem of the inverse control is designed:

[0029] Design Lyapunov function:

[0030]

[0031] make K2 is a constant and K2>0, so that And based on this, the motor torque output control law T is designed:

[0032]

[0033] is the interference estimate.

[0034] Furthermore, the extreme learning machine (ELM) neural network is used to approximate the uncertainty and unknown disturbance of the AGV dynamics model and adaptively offset them. The ELM neural network disturbance observer is designed as follows:

[0035]

[0036] Where h is the output layer matrix, is the current estimated output weight matrix, and the wheel speed ω is used as input:

[0037] h=g(cω+b)

[0038] Among them, c and b are the randomly generated input weight matrix and bias matrix, respectively, and g(·) is the hidden layer activation function.

[0039] Furthermore, the iterative update of the ELM neural network output weights is:

[0040] The actual total system interference is defined as:

[0041] f=hW+ε

[0042] Where ε is the network approximation error, W is the ideal output weight matrix, and the error between the true value and the estimated value of the interference error is defined as:

[0043]

[0044] in To output the weight matrix deviation, define the Lyapunov function:

[0045]

[0046] Among them, e is the speed error, δ is the virtual control error, γ is a constant and greater than 0, and finally the adaptive law is derived is the output weight change rate, which satisfies the stability of the control system.

[0047] Furthermore, the ELM neural network is a single hidden layer feedforward neural network having a structure including 4 input layer neurons, 10 hidden layer neurons and 4 output layer neurons.

[0048] Furthermore, the specific form of the activation function of the hidden layer of the ELM neural network is:

[0049]

[0050] In the formula, ω i is the input of the i-th neuron, c j and b j are the input layer weights and biases of the jth hidden layer neuron, I ij is the linear calculation of the jth node of the hidden layer for the ith input, h ij is the output of the jth node in the hidden layer for the ith input, μ and σ are the center point and width parameter of the Gaussian function respectively.

[0051] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0052] (1) The dynamic modeling of the Mecanum wheel AGV intuitively separates the model certainty and uncertainty, and introduces interference terms to facilitate the targeted design and improvement of the subsequent control algorithm.

[0053] (2) The present invention adopts inverse control to control the AGV dynamic level. Inverse control designs the Lyapunov equation of the entire system step by step and in reverse progressively according to the structural characteristics of the controlled object. The constructed controller has a clear structure and a regular system, and has obvious advantages in realizing adaptive control of uncertain nonlinear systems.

[0054] (3) The present invention adopts ELM neural network to design interference observer. Its fast and unique training method not only greatly reduces the training time, but also avoids the local optimal solution problem of gradient descent algorithm, thereby improving the performance of the entire network, and has strong generalization ability, so that it can ensure high approximation accuracy while training quickly. During the operation of AGV, interference is effectively simulated and adaptively eliminated, improving the stability of the control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention can be better understood by referring to the description given below in conjunction with the accompanying drawings, which together with the following detailed description are included in this specification and form a part of this specification, and are used to further illustrate the preferred embodiments of the present invention and explain the principles and advantages of the present invention.

[0056] Figure 1 This is a kinematic analysis diagram of a Mecanum wheel driven AGV in an embodiment of the present invention;

[0057] Figure 2 This is a flow chart of the ELM-based Mecanum wheel AGV dynamic adaptive inversion control method in an embodiment of the present invention;

[0058] Figure 3 Schematic diagram of the ELM neural network structure in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the scheme of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0060] like Figure 1 As shown, Figure 1 This is the kinematic analysis diagram of the Mecanum wheel driven AGV in this embodiment. Only the planar motion of the Mecanum wheel AGV is considered. XOY represents the global coordinate system, and X q O q Y qRepresents the AGV's own coordinate system, w is half of the distance between the geometric centers of the left and right wheels of the AGV, h is half of the distance between the geometric centers of the front and rear wheels of the AGV, [x, y, θ] T is the position and yaw angle of the AGV in the global coordinate system, β is the angle between the wheel axle and the roller axis, and the angles of all four wheels are 45°. Ax ,V Bx ,V Cx ,V Dx ] T is the speed of the wheel along the coordinate axis Xq, [V Ay ,V By ,V Cy ,V Dy ] T is the speed of the wheel along the coordinate axis Yq, [V A辊 ,V B辊 ,V C辊 ,V D辊 ] T is the linear velocity of the roller on the wheel in contact with the ground, which is caused by the relative sliding between the wheel and the ground; [V A轮 ,V B轮 ,V C轮 ,V D轮 ] T is the linear velocity of the wheel, which is generated by the rotation of the motor driving the wheel hub; [V x ,V y ,ω z ] T is the velocity of the AGV’s center of mass in its own coordinate system.

[0061] like Figure 2 As shown, a Mecanum wheel AGV dynamic adaptive inversion control method of this embodiment includes the following steps:

[0062] Step S100: Assume that the AGV as a whole can be considered as a rigid body, and the body and the four wheels are rigidly connected, so the speed of the AGV body is related to the speed of the wheels. From the above, it can be seen that the center of mass speed of wheels A, B, C, and D is fundamentally determined by [V A轮 ,V B轮 ,V C轮 ,V D轮 ] T and [V A辊 ,V B辊 ,V C辊 ,V D辊 ] T The combined result is [V A辊 ,V B辊 ,V C辊 ,V D辊 ] TThe size and direction of is determined by the differential speed and mechanical structure of the four wheels, and its power source is also the motor. Therefore, the center of mass speed of each wheel can be calculated as:

[0063]

[0064] get

[0065]

[0066] Assume [ω A ,ω B ,ω C ,ω D ] T is the angular velocity of the four wheels, r is the wheel radius, combined with V 轮 = rω, we can get the kinematic model of Mecanum wheel AGV:

[0067]

[0068] Step S200: Set the mass of the AGV to m 车 , the AGV moment of inertia and wheel moment of inertia are J 车 and J 轮 , then the total kinetic energy of the Mecanum wheel AGV during motion is:

[0069]

[0070] The Lagrange equation is introduced to solve the inertia moment of each wheel, and considering the rolling friction between the AGV and the ground, we get

[0071]

[0072] Where T i (i=A,B,C,D) represents the torque on the four wheels, q i and Respectively represent the rotation angle and speed of the four wheels, λ i (i=A,B,C,D) is the viscous friction coefficient between the wheel and the ground, and the solution is:

[0073]

[0074] in

[0075]

[0076] Rewrite the above formula as

[0077]

[0078] This is the nominal dynamic model of AGV, where ω is the wheel speed, T is the motor torque, and M n is the inertia matrix, H n is the viscous friction coefficient matrix. Considering the influence of model uncertainty and external unknown disturbances, it can be rewritten as

[0079]

[0080] Where ΔM and ΔH are uncertainties in the dynamic model, and d is an external unknown disturbance.

[0081]

[0082] This is the final dynamic model that represents the relationship between the AGV motor driving torque and the wheel speed. represents the overall disturbance.

[0083] Step S300: define the wheel speed error as:

[0084]

[0085] in and are the wheel angle and the desired angle, ω and ω d is the wheel speed and the expected speed, e is the speed error, and the virtual control amount ω'=ω d -K1e, K1 is a constant and K1>0, the virtual control amount error δ=ω-ω'.

[0086] Step S301: for the first subsystem, the Lyapunov function is designed as:

[0087]

[0088] but To make It is necessary to satisfy δ=0, so the next step of design is required.

[0089] Step S302: Design a Lyapunov function for the second subsystem

[0090]

[0091] but:

[0092]

[0093] make but

[0094]

[0095] To make Design control law

[0096]

[0097] in, is the interference estimate. It can be seen that if f can approach but The entire control system is stable.

[0098] Step S400, the ELM neural network structure is as follows Figure 3 As shown, the number of neurons in the input layer is D, the number of neurons in the hidden layer is L, and the number of neurons in the output layer is m. i (i=1,2,…,D) is the i-th network input, c j , b j (j=1,2,…,L) is the input layer weight and bias of the jth hidden layer neuron, y k (k=1,2,…m) is the kth network output, h j (x) is the output of the jth node in the hidden layer, that is,

[0099] h j (x) = g(c j x+b j )

[0100] Where g(·) is the activation function, i.e., a nonlinear piecewise continuous function that satisfies the ELM universal approximation ability theorem. The ELM output is

[0101]

[0102] Where W j (j=1,2,…,L) is the weight between the jth hidden layer node and the output node.

[0103] As an embodiment, the ELM neural network structure adopts a network structure including 4 input layer neurons, 10 hidden layer neurons and 4 output layer neurons for the next step of design.

[0104] Step S401, ELM neural network observer design: ELM is used to approximate the interference, and the interference observer is defined as:

[0105]

[0106] Where h is the output layer matrix, is the current estimated output weight matrix, and the wheel speed ω is used as input:

[0107] h=g(cω+b)

[0108] Wherein, c and b are the randomly generated input weight matrix and bias matrix respectively, g(·) is the hidden layer activation function, which plays a limiting role. As an embodiment, the hidden layer activation function adopts the Gaussian function to improve the ability of ELM to handle nonlinear tasks. is the output weight matrix of the current estimate, which is solved by the adaptive law and needs to satisfy Lyapunov stability at the same time. Therefore, the whole observation process is to adjust iteratively. Make it close to the ideal output weight matrix W, and finally make Approximate f.

[0109] Step S402, iterative update of ELM neural network output weights: define the actual system overall interference as:

[0110] f=hW+ε

[0111] Where ε is the network approximation error, W is the ideal output weight matrix, and the error between the true value and the estimated value of the interference error is defined as:

[0112]

[0113] in To output the weight matrix deviation, define the Lyapunov function:

[0114]

[0115] γ is a constant and greater than 0, and its derivative is

[0116]

[0117] Adaptive Law is the output weight change rate, then That is, the Lyapunov stability condition is finally satisfied.

[0118] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well.

Claims

1. A Mecanum wheel AGV dynamic adaptive inversion control method, characterized in that: include: By conducting kinematic analysis and dynamic modeling on the four-wheel Mecanum wheel chassis AGV, the dynamic relationship between the AGV wheel speed and the motor torque is obtained, and the AGV dynamic inversion control law is designed to finally obtain the equivalent control torque. The extreme learning machine (ELM) neural network is used to approximate the uncertainty and unknown disturbance of the AGV dynamic model and adaptively offset them to improve the stability of the control system.

2. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 1, characterized in that: AGV dynamic inversion control includes the following steps: Step 1: Kinematic analysis: Establish the velocity of the AGV center of mass in its own coordinate system [V x ,V y ,ω z ] T and the angular velocity of the four wheels [ω A ,ω B ,ω C ,ω D ] T the relationship between; Step 2: Dynamic modeling: Establish a dynamic model that represents the relationship between the AGV motor drive torque and the wheel speed; Step 3: Design the inverse control law: Use the inverse method to control the Mecanum wheel AGV dynamics, decompose the system into two subsystems, design the intermediate virtual control quantity and Lyapunov function for each subsystem, design the corresponding control law according to the stability conditions, and finally obtain the equivalent control law by stacking and backstepping.

3. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 2, characterized in that: In step 2, the Lagrange equation is combined with the total kinetic energy of the AGV during operation to obtain the nominal dynamic model of the AGV. At the same time, the rolling friction between the AGV and the ground is considered, and model uncertainty and external unknown disturbances are introduced. The final dynamic model is: in represents the overall disturbance, where ω is the wheel speed, is the angular acceleration of the wheel, T is the motor torque, M n is the inertia matrix, H n is the viscous friction coefficient matrix, ΔM and ΔH are the uncertain terms in the dynamic model, and d is the external unknown disturbance.

4. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 3, characterized in that: The nominal kinetic model is: in T i Indicates the torque magnitude of the four wheels. is the angular acceleration of the four wheels, i = A, B, C, D, m 车 is the AGV quality, J 车 and J 轮 are the AGV moment of inertia and wheel moment of inertia respectively, w is half of the distance between the geometric centers of the left and right wheels of the AGV, h is half of the distance between the geometric centers of the front and rear wheels of the AGV, and r is the wheel radius.

5. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 2, characterized in that: In step 3, design the first subsystem of the inverse control: Define wheel speed ω, expected speed ω d and the speed error e, take the virtual control amount ω'=ω d -K1e, K1 is a constant and K1>0, virtual control error δ=ω-ω', design Lyapunov function: Push It is necessary to make δ=0.

6. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 2, characterized in that: In step 3, the second subsystem of the backstepping control is designed: Design Lyapunov function: make K2 is a constant and K2>0, so that And based on this, the motor torque output control law T is designed: is the interference estimate.

7. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 1, characterized in that: The extreme learning machine (ELM) neural network is used to approximate the uncertainty and unknown disturbance of the AGV dynamic model and adaptively offset them. The ELM neural network disturbance observer is designed as follows: Where h is the output layer matrix, is the current estimated output weight matrix, and the wheel speed ω is used as input: h=g(cω+b) Among them, c and b are the randomly generated input weight matrix and bias matrix, respectively, and g(·) is the hidden layer activation function.

8. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 1, characterized in that: Iterative update of the output weights of the ELM neural network: The actual total system interference is defined as: f=hW+ε Where ε is the network approximation error, W is the ideal output weight matrix, and the error between the true value and the estimated value of the interference error is defined as: in To output the weight matrix deviation, define the Lyapunov function: Among them, e is the speed error, δ is the virtual control error, γ is a constant and greater than 0, and finally the adaptive law is derived is the output weight change rate, which satisfies the stability of the control system.

9. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 7, characterized in that: The ELM neural network is a single hidden layer feedforward neural network with a structure consisting of 4 input layer neurons, 10 hidden layer neurons and 4 output layer neurons.

10. The method for adaptive inversion control of Mecanum wheel AGV dynamics according to claim 7, characterized in that: The specific form of the activation function of the hidden layer of the ELM neural network is: In the formula, ω i is the input of the i-th neuron, c j and b j are the input layer weights and biases of the jth hidden layer neuron, I ij is the linear calculation of the jth node of the hidden layer for the ith input, h ij is the output of the jth node in the hidden layer for the ith input, μ and σ are the center point and width parameter of the Gaussian function respectively.

Citation Information

Cited By

  • Intelligent logistics transportation stacking AGV trolley and control method and system thereof

    CN120308114A