Bionic hexapod robot stability optimization design method based on central pattern generator model control

Abstract

The invention discloses a bionic hexapod robot stability optimization design method based on central pattern generator model control, and the method aims to improve the motion stability of a hexapod robot in a complex environment, and optimizes oscillation waveforms by constructing an improved central pattern generator model and an initial value matrix. Further designing a leg joint mapping function to control a joint movement track; the core of the method lies in optimizing the centroid stability in combination with an adaptive agent model and improving the calculation efficiency; by optimizing neuron suppression and excitation coefficients in the central pattern generator model, the uncertainty in the attitude adjustment process is reduced, good motion stability is shown in a test, the potential of expanding to a more complex scene is achieved, and efficient and reliable technical support is provided for engineering application of the hexapod robot.

Description

Technical Field

[0001] The present invention relates to the technical field of bionic hexapod robot motion control, and particularly relates to an optimization design method for the stability of a bionic hexapod robot based on a central pattern generator model control. Background Art

[0002] When using traditional methods to control the motion of bionic robots, problems such as modeling complexity and coordinated control of the leg joints in motion will be faced. These adverse factors will increase the control difficulty of the robot's motion. During the robot's motion control process, the rotation angle of its leg joints directly affects its own motion stability and safety. Especially in the case where a hexapod robot has redundant degrees of freedom, the effective control of the rotation angle of its leg joints becomes an important issue for improving the robot's motion stability. A rotary motor is installed between the leg linkages of the hexapod robot as the rotary joint of the robot, and the rotation of its joint motors is controlled through a central pattern generator model. Therefore, the rhythm period response signal of the central pattern generator model directly affects the safety and stability of the hexapod robot's operation. A large amount of research data shows that a reliable central pattern generator model can maximize the motion stability of the hexapod robot. Therefore, in order to improve the motion stability of the hexapod robot, it is particularly important to propose corresponding optimization design methods for the optimization variables in the central pattern generator model.

[0003] The existing motion methods for bionic hexapod robots based on a central pattern generator model still have the following deficiencies:

[0004] 1. For the existing motion control method of bionic hexapod robots based on a central pattern generator, the central pattern generator model network is a complex non-linear system involving a large number of parameters. These parameters play an important role in the stability, adaptability, generation and conversion of motion patterns, etc. Due to the strong non-linearity of the system and the mutual coupling between parameters, there is still no very ideal method for the parameter tuning of the central pattern generator model network. Generally, the enumeration method is used, and the efficiency is low.

[0005] 2. For the motion control design method of bionic hexapod robots based on a central pattern generator model, the model contains multiple design optimization variables, such as the rise time T of the rhythm response period oscillation signal r , the neuron interaction excitation coefficient c, the neuron inhibition coefficient a, the adaptation coefficient b, and the initial value matrix u 0Parameters such as these, where the uncertainty affects the oscillation stability of the rhythm period response signal of the central pattern generator model, making it difficult to ensure the motion stability of the bionic hexapod robot. Usually, these design optimization variables are considered as deterministic parameters, ignoring the influence of the design optimization variables in the central pattern generator model on the motion stability of the robot. As a result, there is a large difference between the rhythm signal response of the central pattern generator model and the expected waveform, leading to the failure of the robot's control and making it difficult to ensure its motion stability.

[0006] 3. Regarding the method for setting the initial value matrix of the central pattern generator model, it has an important influence on the oscillation stability of the rhythm period response signal. During the simulation and control process of the hexapod robot, the setting of the initial value matrix is crucial for the accuracy of the simulation results and the effectiveness of the control strategy. The initial value matrix contains information about the initial state of the robot, such as joint angles, speeds, accelerations, etc., and this information will directly affect the process and results of the simulation. However, the traditional method for setting the initial value matrix often relies on empirical selection and lacks a scientific and systematic method to determine the optimal initial value matrix.

[0007] In summary, which model to use to measure the uncertainty in the motion process of the bionic hexapod robot controlled by the central pattern generator model is the key to the current optimization design problem. Moreover, currently, in the optimization design method of the bionic hexapod robot based on central pattern generator control, there is no use of an adaptive surrogate model to describe the uncertainty in the optimization design process. Therefore, the development of using an adaptive surrogate model to describe the uncertainty in the optimization design process can not only improve the motion stability of the simulated hexapod robot but also be of great significance in the construction and design of the central pattern generator model. Summary of the Invention

[0008] To overcome the above problems, the present invention proposes a stability optimization design method for a bionic hexapod robot based on central pattern generator model control to solve the above problems.

[0009] The technical solution adopted by the present invention to solve its technical problems is: A stability optimization design method for a bionic hexapod robot based on central pattern generator model control, characterized by including the following steps:

[0010] Step 1: For the problem of optimizing the motion stability design of the bionic hexapod robot, establish its numerical model;

[0011] Step 2: Based on the improved central pattern generator model, establish the central pattern generator of the motion control model of the bionic hexapod robot to control the motion stability of the robot. The specific expression is as follows:

[0012] T r u i 'f +u i f = bv i f + ay i e + c Formula (1) T r v i ' f + v i ' f = y i f T r u i ' e + u i e = bv i e + ay i f + c T r v i ' e + v i ' e = y i e y i f,e = g(u i f,e ) g(u) = max(u, 0)

[0013] In the above formula, i is the oscillator, f is the flexor neuron, e is the extensor neuron, a is the mutual inhibition coefficient, b is the adaptation coefficient, c is the coefficient of the input excitation, T r is the rising value of the oscillation signal, is the output of the oscillator signal, g(u) is the threshold function, y i is the output of the oscillator signal of the i-th neuron, and u represents the initial value matrix of the central pattern generator model;

[0014] Step 3: Design the structure of the initial value matrix to ensure the oscillation stability of the waveform. The structure and selection of the initial value matrix include the following sub-steps:

[0015] Step 31: Let the number of columns in the initial value matrix u of the central pattern generator model represent the number of oscillators of the robot, and the rows represent the numerical values of the specific initial value parameters in each oscillator. It is represented by the matrix as shown in the following formula:

[0016]

[0017] Step 32: Construct a central pattern generator model using six single oscillators, and respectively correspond each single oscillator to the six moving legs of the bionic hexapod robot. There are 4 design optimization variables in each single oscillator. Set the design optimization variables in the six oscillator models to the same value, such as [u 1 , u 2 , u 1 , u 2 T , so as to obtain six single oscillator models U with the same parameters, as shown in the following formula:

[0018]

[0019] Step 33: Let k 1 = [u 1 , u 2 , u 1 , -u 2 T , k 2 = [u 1 , -u 2 , u 1 , u 2 T , for the leg joint motion relationship of the triangular gait of the bionic hexapod robot, the first, third, and fifth columns of the initial value matrix u 0 have the same motion relationship, and the second, fourth, and sixth columns have the same motion relationship, where the motion relationships of the first, third, and fifth columns and the second, fourth, and sixth columns are opposite;

[0020] u 0 = [k 1 , k 2 , k 1 , k 2 , k 1 , k 2 1×6 Formula (4)

[0021] Step 34: Based on the initial value matrix u 0 shown in Formula (4), adopt the enumeration method to randomly generate random initial value matrices k 1 and k 2 , as shown in the above formula, the parameter structure of the initial value matrix u 0 has good periodic properties, and the solved rhythm response signal is continuous, smooth, and stable. Moreover, through the initial value matrix and numerical analysis, it is found that the motion relationships of the first, third, and fifth legs and the second, fourth, and sixth legs are opposite and the values are the same, meeting the requirements of the rhythm signal of the bionic hexapod robot;

[0022] ​​​​Step 4: By analyzing the motion relationship between the hip joint and the knee joint, establish a bionic hexapod robot leg joint mapping function to control the rotation of the leg joints. The formula (5) is as follows:

[0023]

[0024] In the above formula, K h , X θ are the control signals of the hip joint and the knee joint of the bionic hexapod robot respectively, K′ is the derivative of the control signal of the hip joint of the bionic hexapod robot, A h , k are the amplitudes of the hip joint and the knee joint of the bionic hexapod robot respectively;

[0025] Step 5: Based on the robot motion control central pattern generator model, determine the optimized design variables in the parameters of the central pattern generator model;

[0026] Step 6: Establish a numerical simulation problem for the bionic hexapod robot, and determine the design variables, constraint functions and objective functions for the motion stability of the bionic hexapod robot, so as to establish an optimized design problem for the stability of the bionic hexapod robot. The specific expression is:

[0027] s.t.g k (X) ≤ b k k = 1, 2,..., m

[0028] In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is the design variable, and its value range is

[0029] Step 7: Use the Latin hypercube experimental design method to sample sample points in the design domain space and set the allowable error e max , and the iteration step s = 1;

[0030] Step 8: Adopt an optimization method based on an adaptive surrogate model to calculate the candidate shape parameters corresponding to the surrogate models of the objective function and the constraint function respectively, and select a radial basis function to construct the surrogate models of the objective function and the constraint function of the bionic hexapod robot system. Its basic expression is as follows:

[0031]

[0032] In the above formula, is the surrogate model function value, N is the number of sample points, h(r i ) is the radial function, also known as the kernel function, ri = ||X - X i ||, where \(i = 1, 2, \cdots, N\) is the Euclidean distance between the point \(X\) to be measured and the sample point \(X\) i and \(w\) i is the linear weighting coefficient;

[0033] Step 9: Construct the approximate optimization problem of the bionic hexapod robot system as shown in formula (8), and solve the approximate optimization problem of the bionic hexapod robot system shown in formula (8) to obtain the solution \(X\) of this approximate optimization problem (s) :

[0034] s.t. \(g\) k (X) \leq b k \(k = 1, 2, \cdots, m\)

[0035] In the above formula, \(f\) w is the objective function, \(g\) k (X) is the constraint function, \(b\) k is a constant value, \(X\) is the design variable, and its value range is

[0036] Step 10: Calculate the values of the true objective function and the constraint function at the approximate optimization design solution \(X\) (s) ;

[0037] Step 11: Calculate the error \(\delta\) max :

[0038]

[0039] If \(\delta\) max < \(\varepsilon\), then output the optimization design solution \(X\) (s) , and the iteration terminates; otherwise, resample the sample points, update the sample space, and return to Step 3, and set \(s = s + 1\).

[0040] The beneficial effects of the present invention are as follows:

[0041] 1. Aiming at the first point proposed in the background art, the present invention adopts the optimization method of the surrogate model, truly giving play to the advantages of the surrogate model method in the application of the bionic hexapod robot system, so as to essentially improve the calculation efficiency.

[0042] 2. Aiming at the second point proposed in the background art, the present invention proposes a stability optimization design method for the bionic hexapod robot based on the control of the central pattern generator model for the motion stability problem of the bionic hexapod robot.

[0043] 3. Regarding the third point proposed in the background art, the present invention ensures the oscillation stability of the rhythm cycle response signal by introducing the setting strategy of the initial value matrix, thereby controlling the motion stability of the robot.

[0044] Note: The above designs are not in any particular order, and each one makes the present invention distinct and significantly advanced compared to the prior art.

[0045] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0046] Figure 1 is the flowchart of the optimization design method of the bionic hexapod robot system based on the adaptive surrogate model of the present invention;

[0047] Figure 2 is the schematic diagram of the numerical simulation model of the bionic hexapod robot in the specific embodiment Specific Embodiments

[0048] Figure 3 is the rhythm cycle response signal of the central pattern generator model obtained by solving;

[0049] Figure 4 is the three-dimensional simulation model and simplified joint diagram of the bionic hexapod robot.

[0050] The general method of the present invention will be described below in conjunction with the accompanying drawings:

[0051] As Figure 1 shown, a method for optimizing the stability of a bionic hexapod robot controlled by a central pattern generator model includes the following steps:

[0052] Step 1: Establish a numerical model for the problem of optimizing the motion stability of the bionic hexapod robot;

[0053] Step 2: Based on the improved central pattern generator model, establish a motion control central pattern generator model for the bionic hexapod robot to control the motion stability of the robot. The specific expression is as follows:

[0054] T r u i ' f +u i f =bv i f +ay i e +c Formula (1) T r v i ' f +v i ' f =y i f T r u i ' e + u i e = bv i e + ay i f + c T r v i ' e + v i ' e = y i e y i f,e = g(u i f,e ) g(u) = max(u, 0)

[0055] In the above equations, i is the oscillator, f is the flexor neuron, e is the extensor neuron, a is the mutual inhibition coefficient, b is the adaptation coefficient, c is the coefficient of the input excitation, T r is the rising value of the oscillation signal, is the output of the oscillator signal, g(u) is the threshold function, y i is the output of the oscillator signal of the i-th neuron, and u represents the initial value matrix of the central pattern generator model;

[0056] Step 3: Design the structure of the initial value matrix to ensure the oscillation stability of the waveform. The structure and selection of the initial value matrix include the following sub-steps:

[0057] Step 31: Let the number of columns in the initial value matrix u of the central pattern generator model represent the number of oscillators of the robot, and the rows represent the numerical values of the specific initial value parameters in each oscillator. It is represented by a matrix as shown in the following equation:

[0058]

[0059] Step 32: Use 6 single oscillators to construct the central pattern generator model, and correspond each single oscillator to the six moving legs of the bionic hexapod robot respectively. There are 4 design optimization variables in each single oscillator. Set the design optimization variables in the 6 oscillator models to the same value, such as [u 1 , u 2 , u 1 , u 2 T , so as to obtain 6 single oscillator models U with the same parameters, as shown in the following equation:​

[0060]

[0061] Step 33: Let k 1 = [u 1 , u 2 , u 1 , -u 2 T , k 2 = [u 1 , -u 2 , u 1 , u 2 T , for the leg joint motion relationship of the bionic hexapod robot's triangular gait, the first, third, and fifth columns of the initial value matrix u 0 have the same motion relationship, and the second, fourth, and sixth columns have the same motion relationship, where the motion relationships of the first, third, and fifth columns and the second, fourth, and sixth columns are opposite;

[0062] u 0 = [k 1 , k 2 , k 1 , k 2 , k 1 , k 2 1×6 Formula (4)

[0063] Step 34: Based on the initial value matrix u 0 shown in Formula (4), using the enumeration method, randomly generate random initial value matrices k 1 and k 2 , as shown in the above formula, the parameter structure of the initial value matrix u 0 has good periodic properties, and the solved rhythm response signal is continuous, smooth, and stable. Moreover, through the initial value matrix and numerical analysis, it is found that the motion relationships of legs 1, 3, 5 and legs 2, 4, 6 are opposite and the values are the same, meeting the requirements of the bionic hexapod robot's rhythm signal;

[0064] Step 4: By analyzing the motion relationship between the hip joint and the knee joint, establish a mapping function of the leg joints of the bionic hexapod robot to control the rotation of the leg joints. The formula (5) is as follows:

[0065]

[0066] In the above formula, K h , X θ are the control signals of the hip joint and the knee joint of the bionic hexapod robot respectively, K′ is the derivative of the control signal of the hip joint of the bionic hexapod robot, and A h , k are the amplitudes of the hip joint and the knee joint of the bionic hexapod robot respectively; ​​​

[0067] Step 5: Based on the robot motion control central pattern generator model, determine the optimized design variables among the parameters of the motion control central pattern generator model;

[0068] Step 6: Establish a numerical simulation problem for the bionic hexapod robot, and determine the design variables, constraint functions, and objective functions for the motion stability of the bionic hexapod robot, so as to establish an optimized design problem for the stability of the bionic hexapod robot. The specific expression is:

[0069] min f w (X, μ P , μ Q ) Formula (6) s.t. g k (X) ≤ b k k = 1, 2,..., m

[0070] In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is the design variable, and its value range is

[0071] Step 7: Use the Latin hypercube experimental design method to sample sample points in the design domain space and set the allowable error e max , and the number of iteration steps s = 1;

[0072] Step 8: Adopt an optimization method based on an adaptive surrogate model to calculate the candidate shape parameters corresponding to the surrogate models of the objective function and the constraint function respectively, and select a radial basis function to construct the surrogate models of the objective function and the constraint function of the bionic hexapod robot system. Its basic expression is as follows:

[0073]

[0074] In the above formula, is the surrogate model function value, N is the number of sample points, h(r i ) is the radial function, also known as the kernel function, r i = ||X - X i ||, i = 1, 2,..., N is the Euclidean distance between the point to be measured X and the sample point X i , and w i is the linear weighting coefficient;

[0075] Step 9: Construct the approximate optimization problem of the bionic hexapod robot system as shown in formula (8), and solve the approximate optimization problem of the bionic hexapod robot system shown in formula (8) to obtain the solution X of this approximate optimization problem (s) :

[0076] s.t. g k (X) ≤ b k k = 1, 2,..., m

[0077] In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is the design variable, and its value range is

[0078] Step 10: Calculate the values of the true objective function and the constraint function at the approximate optimal design solution X (s) ;

[0079] Step 11: Calculate the error δ max :

[0080]

[0081] If δ max < ε, then output the optimal design solution X (s) , and the iteration terminates; otherwise, resample the sample points, update the sample space, and return to Step 3, and set s = s + 1.

[0082] To further elaborate on the present invention in more detail, the following will further illustrate the solution of the present invention in combination with a specific embodiment. Taking the optimization design of a bionic hexapod robot as an example, and implementing it on the premise of the technical solution of the present invention, the detailed implementation manner and specific operation process are given, but the protection scope of the present invention is not limited to the following examples.

[0083] As Figure 4 shown, it is the three-dimensional simulation model and simplified joint diagram of the bionic hexapod robot targeted by the method of the present invention. Implement according to the process Figure 1 shown. A stability optimization design method for a bionic hexapod robot based on the central pattern generator model control, aiming at the bionic hexapod robot system Figure 2 shown, its specific steps are as follows:

[0084] Step 1: Establish a numerical model for the motion stability optimization design problem of the bionic hexapod robot;

[0085] Step 2: Based on the improved central pattern generator model, establish a bionic hexapod robot motion control central pattern generator model to control the motion stability of the robot. The specific expression is as follows:

[0086] T r u i ' f + u i f = bv i f + ay i e + c Formula (1) T r v i ' f + v i ' f = y i f T r u i ' e + u i e = bv i e + ay i f + c T r v i ' e + v i ' e = y i e y i f,e = g(u i f,e ) g(u) = max(u, 0)

[0087] In the above formula, i is the oscillator, f is the flexor neuron, e is the extensor neuron, a is the mutual inhibition coefficient, b is the adaptation coefficient, c is the coefficient of the input excitation, T r is the rising value of the oscillation signal, is the output of the oscillator signal, g(u) is the threshold function, y i is the output of the i-th neuron oscillator signal, and u represents the initial value matrix of the central pattern generator model;

[0088] Step 3: Design the structure of the initial value matrix to ensure the oscillation stability of the waveform. The structure and selection of the initial value matrix include the following sub-steps:

[0089] Step 31: Let the number of columns in the initial value matrix \(u\) of the central pattern generator model represent the number of oscillators of the robot, and the rows represent the numerical values of the specific initial value parameters in each oscillator. It is represented by the following matrix:

[0090]

[0091] Step 32: Use six single oscillators to construct the central pattern generator model, and respectively correspond each single oscillator to the six moving legs of the bionic hexapod robot. There are 4 design optimization variables in each single oscillator. Set the design optimization variables in the six oscillator models to the same value, such as \([u 1 ,u 2 ,u 1 ,u 2 T Thus, obtain six single oscillator models \(U\) with the same parameters, as shown in the following formula:

[0092]

[0093] Step 33: Let \(k 1 =[u 1 ,u 2 ,u 1 ,-u 2 T ,k 2 =[u 1 ,-u 2 ,u 1 ,u 2 T For the leg joint motion relationship of the triangular gait of the bionic hexapod robot, columns 1, 3, and 5 of the initial value matrix \(u 0 have the same motion relationship, and columns 2, 4, and 6 have the same motion relationship, where the motion relationships of columns 1, 3, and 5 and columns 2, 4, and 6 are opposite;

[0094] u 0 =[k 1 ,k 2 ,k 1 ,k 2 ,k 1 ,k 2 1×6 Formula (4)

[0095] Step 34: Based on the initial value matrix \(u 0 shown in Formula (4), adopt the enumeration method to randomly generate random initial value matrices \(k 1 and \(k 2 , as shown in the above initial value matrix \(u 0 ​​​​The parameter structure has good periodic properties, and the solved rhythm response signal is continuous, smooth and stable. Through the initial value matrix and numerical analysis, it is found that the motion relationships of the 1st, 3rd, and 5th legs are opposite and have the same numerical values as those of the 2nd, 4th, and 6th legs, meeting the requirements of the bionic hexapod robot's rhythm signal;

[0096] Step 4: By analyzing the motion relationship between the hip joint and the knee joint, establish a mapping function of the leg joints of the bionic hexapod robot to control the rotation of the leg joints. The formula (5) is as follows:

[0097]

[0098] In the above formula, K h , X θ are the control signals of the hip joint and the knee joint of the bionic hexapod robot respectively, K′ is the derivative of the control signal of the hip joint of the bionic hexapod robot, and A h , k are the amplitudes of the hip joint and the knee joint of the bionic hexapod robot respectively;

[0099] Step 5: Based on the robot motion control central pattern generator model, determine the optimal design variables in the parameters of the central pattern generator model;

[0100] Step 6: Establish a numerical simulation problem for the bionic hexapod robot, and determine the design variables, constraint functions, and objective functions for the motion stability of the bionic hexapod robot, so as to establish an optimal design problem for the stability of the bionic hexapod robot. The specific expression is:

[0101] minf w (X, μ P , μ Q ) Formula (6) s.t. g k (X) ≤ b k k = 1, 2,..., m

[0102] In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is the design variable, and its value range is

[0103] Step 7: Use the Latin hypercube experimental design method to sample sample points in the design domain space and set the allowable error e max , and the number of iteration steps s = 1;

[0104] Step 8: Adopt an optimization method based on an adaptive surrogate model to calculate the candidate shape parameters corresponding to the surrogate models of the objective function and the constraint function respectively. Select a radial basis function to construct the surrogate models of the objective function and the constraint function of the bionic hexapod robot system. The basic expression is as follows:

[0105]

[0106] In the above formula, is the surrogate model function value, N is the number of sample points, h(r i ) is the radial function, also known as the kernel function, r i =||X - X i ||, i = 1, 2,..., N is the Euclidean distance between the point to be measured X and the sample point X i , and w i is the linear weighting coefficient;

[0107] Step 9: Construct the approximate optimization problem of the bionic hexapod robot system shown in formula (8), and solve the approximate optimization problem of the bionic hexapod robot system shown in formula (8) to obtain the solution X (s) of this approximate optimization problem:

[0108] s.t. g k (X) ≤ b k k = 1, 2,..., m

[0109] In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is the design variable, and its value range is

[0110] Step 10: Calculate the values of the true objective function and the constraint function at the approximate optimal design solution X (s) ;

[0111] Step 11: Calculate the error δ max :

[0112]

[0113] If δ max < ε, then output the optimal design solution X (s) , and the iteration terminates; otherwise, resample the sample points, update the sample space, and return to Step 3, and set s = s + 1.

[0114] According to the Figure 2 centroid change diagram, during the entire movement process, the centroid change is relatively stable, and there are no sudden increases or decreases. At the same time, the relative error is calculated by comparing the centroid target function value of the robot, 3.5073 mm, with the centroid amplitude of 3.6 mm measured in the simulation experiment. The error range is 2.5%, meeting the requirement of a minimum error of 10%. The results show that the response value obtained from the approximate model of the target function has good accuracy, and the optimization result obtained using this approximate model meets the design requirements.

[0115] Based on the initial value matrix u 0 as shown in the formula, the enumeration method is used to randomly generate the initial value matrix u 1 and u 2 , where the specific parameters are shown in Table 1 below. The rhythm period response signal of the central pattern generator model is solved as follows Figure 3 shown.

[0116] Table 1 Parameters of the motion control model of the bionic hexapod robot

[0117]

[0118] The above detailed description is a specific description of the feasible embodiments of the present invention. This embodiment is not intended to limit the patent scope of the present invention. Any equivalent implementation or modification without departing from the present invention shall be included in the patent scope of this case.

Claims

1. A stability optimization design method for a bionic hexapod robot based on central pattern generator model control, characterized in that: The steps include: Step 1: Establish a numerical model for the optimization design of the motion stability of a bionic hexapod robot; Step 2: Based on the improved central pattern generator model, a bionic hexapod robot motion control central pattern generator model is established to control the robot's motion stability. The specific expression is as follows: In the above formula, i is the oscillator, f is the flexor neuron, e is the extensor neuron, a is the mutual inhibition coefficient, b is the adaptation coefficient, c is the coefficient of input excitation, T r is the rising value of the oscillation signal, is the output of the oscillator signal, g(u) is the threshold function, y i is the output signal of the ith neuron oscillator, and u represents the initial value matrix of the central pattern generator model; Step 3: Design the structure of the initial value matrix to ensure the oscillation stability of the waveform. The structure and selection of the initial value matrix include the following steps: Step 31: Let the column number in the initial value matrix u of the central pattern generator model represent the number of oscillators of the robot, and the row represent the value of the specific initial value parameter in each oscillator, that is, the matrix is ​​represented as shown in the following formula: Step 32: Use 6 single oscillators to build a central pattern generator model, and each single oscillator corresponds to the six moving legs of the bionic hexapod robot. There are 4 design optimization variables in each single oscillator. Set the design optimization variables in the 6 oscillator models to the same value, such as [u1, u2, u1, u2] T , thus obtaining 6 single oscillator models U with the same parameters, as shown in the following formula: Step 33: Let k1 = [u1, u2, u1, -u2] T , k2=[u1,-u2,u1,u2] T ,For the motion relationship of the leg joints of the bionic hexapod robot in triangular gait, the columns 1, 3 and 5 of the initial value matrix u0 have the same motion relationship, and the columns 2, 4 and 6 have the same motion relationship, among which the motion relationship of columns 1, 3 and 5 is opposite to that of columns 2, 4 and 6; u0=[k1, k2, k1, k2, k1, k2] 1×6 Formula(4) Step 34: Based on the initial value matrix u0 shown in formula (4), the enumeration method is used to randomly generate random initial value matrices k1 and k2. As shown in the above formula, the parameter structure of the initial value matrix u0 has good periodic properties, and the solved rhythmic response signal has the characteristics of continuity, smoothness and stability. In addition, through the initial value matrix and numerical analysis, it is found that the movement relationship between 1, 3, 5 feet and 2, 4, 6 feet is opposite and the values ​​are the same, which meets the requirements of the rhythmic signal of the bionic hexapod robot; Step 4: By analyzing the motion relationship between the hip joint and the knee joint, the leg joint mapping function of the bionic hexapod robot is established to control the rotation of the leg joint. Formula (5) is as follows: In the above formula, K h , X θ are the control signals of the hip joint and knee joint of the bionic hexapod robot, K′ is the derivative of the control signal of the hip joint of the bionic hexapod robot, and A h , k are the amplitudes of the hip joint and knee joint of the bionic hexapod robot respectively; Step 5: Based on the robot motion control central pattern generator model, determine the optimal design variables in the motion control central pattern generator model parameters; Step 6: Establish a numerical simulation problem of the bionic hexapod robot, and determine the design variables, constraint functions and objective functions for the motion stability of the bionic hexapod robot, so as to establish an optimization design problem for the stability of the bionic hexapod robot. The specific expression is: In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is a design variable, and its value range is Step 7: Use Latin hypercube experimental design method in the design domain space Sample points are sampled within and the allowable error e is set max , iterative steps s = 1; Step 8: Adopt the optimization method based on adaptive proxy model to calculate the shape parameters corresponding to the proxy models of the objective function and constraint function, and select radial basis function to construct the proxy model of the objective function and constraint function of the bionic hexapod robot system. The basic expression is as follows: In the above formula, is the function value of the proxy model, N is the number of sample points, h(r i ) is a radial function, also known as a kernel function, r i =||XX i ||, i=1,2,...,N is the measured point X and the sample point X i The Euclidean distance between i is the linear weighting coefficient; Step 9: Construct the approximate optimization problem of the bionic hexapod robot system as shown in formula (8), and solve the approximate optimization problem of the bionic hexapod robot system as shown in formula (8) to obtain the solution X of this approximate optimization problem. (s) : In the above formula, f w is the objective function, g k (X) is the constraint function, b k is a constant value, X is a design variable, and its value range is Step 10: Calculate the true objective function and constraint functions In the approximate optimization design solution X (s) The value at Step 11: Calculate the error δ max : If δ max <ε, then output the optimal design solution X (s) , the iteration terminates; otherwise, resample the sample points, update the sample space, and return to step 3, and set s=s+1.

2. The stability optimization design method of a bionic hexapod robot based on central pattern generator model control according to claim 1, characterized in that: The method based on adaptive agent model is adopted, and its basic steps are as follows: Step 1: From the sample point Ω X ={X1,...,X j ,...,X N }, select X from j=1,2,...,N j As the key point, select the distance key point X j The nearest M sample points p=1,2,...,M and To build a micro-agent model as shown below: In the above formula, the shape parameter ε and the weight coefficient w=(w1,...,w M ) T is the unknown quantity to be determined, and the weight coefficient w=(w1,...,w M ) T The expanded form is shown below: Step 2: Move the key point X j As the prediction point, it is brought into formula (10), and let j The predicted value at and the true value f(X j ) are equal, as shown in the following formula: Step 3: Substitute formula (11) into formula (12) to obtain the equation for solving the shape parameters as shown in the following formula. Solve the following equation to obtain the key point X j And the shape parameter ε corresponding to the M sample points nearby: Step 4: Repeat steps 2 and 3, select the remaining N-1 sample points as key points in turn, and solve the shape parameters corresponding to this sample point and the M sample points near it, and finally obtain N shape parameters to be selected; Step 5: Set the sample point Ω X ={X1,...,X N } is divided into construction groups {X1,...,X N / 2 } and test group There are two groups of sample points, where the sample points of the construction group are combined with each shape parameter to be selected to construct the corresponding proxy model, and the sample points of the test group are used as test points to verify the accuracy of the proxy model, and the error between the proxy model function value of the test point and the true value of the function is calculated by the following formula: In the above formula, f(X i ) is the true value of the function at the test point, is the proxy model function value of the test point.

Images