A method for a mobile phone to remotely control a robotic arm to perform grasping based on a neural network

Abstract

A method for a mobile phone to remotely control a robotic arm for grasping, which is a robotic arm motion system that uses a mobile phone to collect motion trajectories, synthesize end trajectories for trajectory planning and motion control. According to the forward and inverse kinematic models of the robotic arm, a neural network optimization method is used, and combined with the quadratic programming method, a neural network method for trajectory planning of a mobile phone remotely controlled grasping robotic arm is designed to effectively solve the robotic arm motion problem in complex scenarios and complex tasks. The present invention can efficiently calculate the motion trajectory of the mobile phone and calculate the motion state of the robotic arm in real time according to the trajectory, realizing high-precision and high-timeliness real-time tracking of the end of the robotic arm.

Description

Technical Field

[0001] The present invention relates to a method for grasping a mobile phone remote control manipulator based on a neural network, in particular to a manipulator trajectory planning method with an end pose following the motion vector of the mobile phone. Background Art

[0002] The motion of a manipulator is a complex system with high precision and multiple inputs and outputs. Due to the complexity of its system, it can be widely used in various scenarios, such as industrial assembly, dangerous goods transportation and other fields. At the same time, due to the complexity of its system, the diversity of work requirements, and the diversity of the structure of the manipulator itself, the control requirements of the manipulator are also relatively complex and variable. Existing control methods for manipulators include PID control, model predictive control, neural network trajectory planning, etc.

[0003] The neural network has a high solution efficiency in solving matrix equations, including the inverse operation of time-varying matrices. The neural network can approach the result within a limited time and converge the error to zero. In the trajectory planning of a manipulator, each joint angle changes with time and can be regarded as a fast time-varying system. The neural network solution method has high solution accuracy, fast solution speed, and certain anti-interference ability. The neural network can meet the high-precision and high-timeliness requirements of control parameters in trajectory planning, and to a certain extent, ensure the accuracy and stability of the system.

[0004] For complex and variable grasping tasks, the manipulator needs to independently compile control programs for each different task according to the requirements of each task, which will increase a lot of computational pressure on the entire manipulator motion system. In order to reduce the computational burden of data, remote control can simplify complex task scenarios and hand over part of the judgment and control to the controller of the manipulator, thereby improving the scalability of the entire system and the diversity of application scenarios. Summary of the Invention

[0005] In order to overcome the drawbacks of existing mechanical control methods and the problem of re-programming for different application scenarios, the present invention provides a high-precision and high-timeliness remote control manipulator trajectory planning method, which can control the end pose transformation of the manipulator using the motion parameters of the mobile phone.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A method for grasping a mobile phone remote control manipulator based on a neural network, comprising the following steps:

[0008] Step 1, obtain the acceleration data g(x) of the mobile phone through the app and transmit it to the control program through wifi;

[0009] Step 2, convert the acceleration to the desired end velocity and the end trajectory r(t);

[0010] Step 3, use the Denavit-Hartenberg convention to establish the inverse kinematic equation of the robotic arm by multiplying transformation matrices, and set the initial joint angle θ(0);

[0011] Step 4, calculate the error between the actual end trajectory and the desired end trajectory, and design the optimal trajectory motion route for the robotic arm to grasp;

[0012] Step 5, based on the optimal motion trajectory planning scheme in Step 4, construct a neural network model for solution;

[0013] Step 6, send the angle values of each axis obtained each time to the robotic arm through a specified protocol, and obtain the true end motion trajectory f(θ) of the robotic arm. Substitute f(θ) into the neural network model to obtain the motion joint trajectories.

[0014] In the said Step 1, the acceleration g(x) is:

[0015] g(x) = [ax; ay; az] - [0; 0; g local

[0016] where ax, ay, and az are the accelerations on the three axes respectively, and g local is the local acceleration due to gravity.

[0017] Furthermore, in the said Step 2, the desired end trajectory is:

[0018]

[0019]

[0020] where t * is the previous moment, t is the current moment, and g(x) is the acceleration at the current moment.

[0021] Still further, in the said Step 3, the transformation matrix is as follows:

[0022]

[0023] where θ is the joint angle, α is the link twist angle, a is the link length, and d is the link offset. According to the transformation matrix, the rotation matrix from the base coordinate to the coordinates of each axis of the robotic arm is obtained

[0024]

[0025] where, ​is the position of the origin of the end - effector coordinate system in the base coordinate system, is the rotation matrix from the base coordinate system to the end - effector coordinate system.

[0026] Furthermore, the process of step 4 is as follows:

[0027] The control problem of the robotic arm is transformed into a quadratic optimization problem, and its performance index is:

[0028]

[0029] where, is the angular velocity of each axis of the robotic arm, is the transpose of the angular velocity of each axis of the robotic arm, and the optimal motion planning equation formed is described as:

[0030]

[0031]

[0032] where, J(θ) is the Jacobian matrix of the robotic arm, r is the desired trajectory of the end - effector of the robotic arm, is the desired velocity of the end - effector of the robotic arm, k P > 0 is the parameter gain, used to adjust the rate at which the end - effector moves to the desired path, and f(θ) is the actual motion trajectory of the robotic arm;

[0033] The process of step 5 is as follows:

[0034] 5.1 Adopt the following network dynamic equation with finite - time convergence characteristics:

[0035]

[0036] where, is the derivative of the error, sign(·) is the sign function, γ > 0, b1 > 0, b2 > 0, b3 > 0, 0 < p < 1 are specified parameters;

[0037] 5.2 According to the optimal planning equation (3) in step 4, establish the Lagrangian function

[0038]

[0039] where, λ(t) is the Lagrange multiplier vector, λ T (t) is the transpose of the λ(t) vector, and the following equation is obtained by solving the Lagrangian:

[0040] w(t)y(t) = v(t) (5)

[0041] where,

[0042] To solve the above equation, the error is set as:

[0043] E(t) = w(t)y(t) - v(t) (6)

[0044] According to the definition of Equation (4), the following finite-time convergent neural network model is obtained:

[0045]

[0046] Based on the neural network algorithm, the present invention designs a method for grasping a mobile phone remote-controlled robotic arm based on a neural network, realizes the exponential convergence characteristic of the error of the motion joint angle at the end of the robotic arm, provides a trajectory planning method for manipulating the robotic arm to complete complex grasping actions, and improves the stability and scalability of the system.

[0047] The technical concept of the present invention is: for the complex motion of the robotic arm grasping, using the neural network solution method, combined with the quadratic optimal planning scheme, to design a neural network method for the trajectory planning of the mobile phone remote-controlled robotic arm grasping. The solution method of the neural network ensures that the motion joint angles at the end of the robotic arm converge to the desired joint positions within a finite time. The mobile phone remote control provides the scalability and operability of the system. Ensure that the end effector moves along with the motion vector of the mobile phone to achieve mobile phone remote control.

[0048] The beneficial effects of the present invention are: solving the problem of the end pose following of the robotic arm, and improving the operability of the moving robotic arm and the scalability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flowchart of the path planning scheme provided by the present invention.

[0050] Figure 2 It is the motion trajectory of the end of the robotic arm.

[0051] Figure 3 It is the trajectory of each joint angle of the robotic arm.

[0052] Figure 4 It is the error on each axis in three-dimensional space. DETAILED DESCRIPTION OF THE INVENTION

[0053] The present invention will be further described below with reference to the accompanying drawings.

[0054] Referring to Figures 1 - 4 , a method for grasping a mobile phone remote-controlled robotic arm based on a neural network includes the following steps:

[0055] Step 1, obtaining the acceleration data g(x) of the mobile phone through the app and transmitting it to the control program through wifi;

[0056] g(x) = [ax; ay; az] - [0; 0; g local

[0057] Step 2, convert the acceleration to the desired end velocity and the end trajectory r(t);

[0058]

[0059]

[0060] where t * is the previous moment, t is the current moment, and g(x) is the acceleration at the current moment;

[0061] Step 3, use the Denavit - Hartenberg convention to establish the inverse kinematic equation of the robotic arm by multiplying transformation matrices, set the initial joint angle θ(0), and the transformation matrices are as follows:

[0062]

[0063] where θ is the joint angle, α is the link twist angle, a is the link length, d is the link offset, and obtain the rotation matrix from the base coordinate to the coordinates of each axis of the robotic arm according to the transformation matrix

[0064]

[0065] where, is the position of the origin of the end coordinate system in the base coordinate, is the rotation matrix from the base coordinate system to the end coordinate system;

[0066] Step 4, calculate the error between the actual end trajectory and the desired end trajectory, and design the optimal trajectory motion route for the robotic arm to grasp, and the process is as follows:

[0067] The motion control problem of the robotic arm can be converted into a quadratic optimization problem, and its performance index is:

[0068]

[0069] where, is the angular velocity of each axis of the robotic arm, is the transpose of the angular velocity of each axis of the robotic arm, and the formed optimal motion planning equation is described as:

[0070]

[0071]

[0072] ​Among them, J(θ) is the Jacobian matrix of the robotic arm, r is the desired trajectory of the end effector of the robotic arm, is the desired velocity of the end effector of the robotic arm, k P >0 is the parameter gain used to adjust the rate at which the end effector moves to the desired path, and f(θ) is the actual motion trajectory of the robotic arm;

[0073] Step 5, based on the optimal motion trajectory planning scheme in Step 4, construct a neural network model for solution, and the process is as follows:

[0074] 5.1 Adopt the following network dynamic equation with finite-time convergence characteristics:

[0075]

[0076] Among them, is the derivative of the error, sign(·) is the sign function, γ>0, b1>0, b2>0, b3>0, 0<p<1 are the specified parameters;

[0077] 5.2 According to the optimal planning equation (3) in Step 4, establish the Lagrangian function

[0078]

[0079] Among them, λ(t) is the Lagrange multiplier vector, and λ T (t) is the transpose of the vector λ(t). The following equation is obtained through Lagrangian solution:

[0080] w(t)y(t) = v(t) (5)

[0081] Among them,

[0082] To solve the above equation, set the error as:

[0083] E(t) = w(t)y(t) - v(t) (6);

[0084] According to the definition of Equation (4), the following finite-time convergence neural network model can be obtained:

[0085]

[0086] Step 6, send the obtained angle values of each axis to the robotic arm through the specified protocol, and obtain the true end motion trajectory f(θ) of the robotic arm. Substitute f(θ) into the equation (7) to obtain the motion joint trajectories.

[0087] To verify the feasibility of the scheme, the present invention gives the simulation results of this scheme on MATLAB:

[0088] The parameters are given as follows:

[0089] Set the initial joint angles Assume that the received acceleration g(x) is:

[0090]

[0091] where t is time, set the total running time T = 10s. In Equation (3), k P = 1. In Equation (4), γ = 10, b1 = 1, b2 = 2, b3 = 3, p = 0.5.

[0092] The execution process of the entire scheme is as Figure 1 shown, and the running effect is as Figure 2 shown. The changes in each joint angle during operation are as Figure 3 shown.

[0093] From Figure 2 and Figure 4 it can be seen that the error between the manipulator and the desired trajectory during the grasping process is small, and the end of the manipulator can reach the specified position within a limited time to perform corresponding actions.

[0094] In summary, the mobile phone remote control manipulator grasping method based on neural network can effectively solve the problem of end - pose following of the manipulator, ensure that the system error converges within a limited time, and make the system have high real - time performance.

Claims

1. A method for a mobile phone to remotely control a robotic arm to perform grasping based on a neural network, characterized in that, The method includes the following steps: Step 1: Obtain the acceleration data g(x) of the mobile phone through the app and transmit it to the control program via Wi-Fi; Step 2, convert the acceleration to the desired end velocity and the end trajectory r(t); Step 3: Use the Denavit-Hartenberg convention to establish the inverse kinematic equation of the robotic arm by multiplying transformation matrices, and set the initial joint angle θ(0); Step 4: Calculate the error between the actual end trajectory and the desired end trajectory, and design the optimal trajectory motion route for the robotic arm to grasp; Step 5: Based on the optimal motion trajectory planning scheme in Step 4, construct a neural network model for solution; Step 6: Send the angle values of each axis obtained each time to the robotic arm through the specified protocol, and obtain the actual end motion trajectory f(θ) of the robotic arm. Substitute f(θ) into the neural network model to obtain the motion joint trajectories; The process of Step 4 is as follows: The control problem of the robotic arm is converted into a quadratic optimization problem, and its performance index is: Among them, is the angular velocity of each axis of the robotic arm, is the transpose of the angular velocity of each axis of the robotic arm, and the optimal motion planning equation is described as: Among them, J(θ) is the Jacobian matrix of the robotic arm, r is the desired trajectory of the end effector of the robotic arm, is the desired velocity of the end effector of the robotic arm, and k P > 0 is the parameter gain used to adjust the rate at which the end effector moves to the desired path, and f(θ) is the actual motion trajectory of the robotic arm; The process of Step 5 is as follows: 5.1 Adopt the following network dynamic equation with finite-time convergence characteristics: Among them, is the derivative of the error, sign(·) is the sign function, γ>0, b1>0, b2>0, b3>0, 0<p<1 are specified parameters; 5.2 According to the optimal planning equation (3) in Step 4, establish the Lagrangian function where λ(t) is the Lagrange multiplier vector, and λ T (t) is the transpose of the vector λ(t), and the following equation is obtained by solving the Lagrange equation: w(t)y(t) = v(t) (5) Among them, To solve the above equation, set the error as: E(t) = w(t)y(t) - v(t) (6) According to the definition of formula (4), obtain the following finite-time convergence neural network model:

2. The method for grasping with a mobile phone remote control robotic arm based on a neural network according to claim 1, wherein In Step 1, the acceleration g(x) is: g(x) = [ax; ay; az] - [0; 0; g local ​ where ax, ay, and az are the accelerations on the three axes respectively, and g local is the local acceleration due to gravity.

3. A method for a mobile phone to remotely control a robotic arm to perform grasping based on a neural network according to claim 1 or 2, characterized in that, In the said step 2, the desired end trajectory is as follows: where t * is the previous moment, t is the current moment, and g(x) is the acceleration at the current moment.

4. A method for a mobile phone to remotely control a robotic arm to perform grasping based on a neural network according to claim 1 or 2, characterized in that, In Step 3, the transformation matrix is as follows: Among them, θ is the joint angle, α is the link twist angle, a is the link length, and d is the link offset. According to the transformation matrix, the rotation matrix from the base coordinate to the coordinates of each axis of the robotic arm is obtained Among them, is the position of the origin of the end coordinate system in the base coordinate system, is the rotation matrix from the base coordinate system to the end coordinate system.

Images