A newt robot motion control method and system based on a central pattern generator

Abstract

The application discloses a salamander robot motion control method and system based on a central pattern generator, and belongs to the technical field of bionic robot motion control. The method comprises the following steps: performing configuration analysis on the salamander robot, and determining a coupling scheme of a CPG network; combining the designed coupling scheme to construct an offline learning law and an online updating law of coupling weight parameters; designing a motion control strategy based on the CPG model, considering gait characteristics and leg trajectory generation, and obtaining actual input signals; and under the driving of the actual input signals, the salamander robot generates rich gaits and smoothly switches between different gaits. The application can improve the output waveform accuracy and convergence speed of the CPG model, and the motion control strategy designed based on the same can realize more accurate robot motion gaits and faster motion initialization and gait switching.

Description

Technical Field

[0001] This invention belongs to the field of biomimetic robot motion control technology, and particularly relates to a method and system for motion control of a salamander-like robot based on a central pattern generator. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Currently, in order to understand and reproduce the movement of salamanders, domestic and foreign institutions and scholars have designed a series of salamander-like robots and carried out a lot of research. In the early stage, salamander-like robots with segmented spines and rotatable limbs were designed to analyze the switching problem of salamanders in amphibious gait and the neuromechanical principles contained therein. Later, inspired by the three-dimensional skeletal structure of real animals, salamander-like robots with multiple degrees of freedom in their legs were designed, which have stronger land movement capabilities. In order to control highly redundant salamander-like robots, the method based on the central pattern generator (CPG) is very effective. The central pattern generator is usually composed of nonlinear oscillators coupled together and is often used in the field of robotics to generate periodic smooth control signals. Among them, the asymmetric oscillation signal is very important for generating rich robot motion behavior. Reference [1] proposes an improved Hopf oscillator model, which can independently control the rising and falling phases of the oscillation. Reference [2] proposes a coupling scheme for this improved Hopf oscillator, in which the oscillators are directly coupled together through a rotation matrix. This method has been successfully applied to biomimetic robots such as quadruped robots, robotic fish, and hexapod robots. In addition, Reference [3] improves the method in Reference [2] by adding adjustable coupling weight parameters.

[0004] However, the method in reference [3] also has some limitations. First, during motion initialization and gait switching, if the coupling weight is small, the convergence speed of the CPG model is slow. Increasing the coupling weight will speed up the convergence, but the waveform of the output signal will be somewhat inaccurate. Second, the network output signal parameters, including period, amplitude and duty cycle, have not been measured and analyzed, but these parameters are crucial for the precise control of robot motion.

[0005] [1]Righetti L,Ijspeert A J.Pattern generators with sensory feedback for the control of quadruped locomotion[C] / / IEEE International Conference onRobotics and Automation.Piscataway,USA:IEEE,2008:819-824.

[0006] [2]Santos CP,Matos V.Gait transition and modulation in a quadrupedrobot:Abrainstem-like modulation approach[J].Robotics andAutonomous Systems,2011,59(9):620-634.

[0007] [3]Chen WH, Lin HS, Lin YM, et al. TurboQuad: A novel leg–wheel transformable robot with smooth and fast behavioral transitions[J]. IEEETransactions on Robotics, 2017, 33(5):1025-1040. Summary of the Invention

[0008] In view of this, the present invention provides a motion control method and system for a salamander-like robot based on a central pattern generator, which solves the technical problems existing in the background art. It will improve the accuracy of the output control signal waveform and accelerate the convergence speed of the CPG model. Under the action of the designed control strategy, it can not only achieve more accurate robot motion, but also achieve rapid motion initialization and gait switching.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] The first aspect of the present invention provides a motion control method for a salamander-like robot based on a central pattern generator, comprising: performing configuration analysis on the salamander-like robot to determine the coupling scheme of the CPG network; constructing an offline learning law and an online update law for the coupling weight parameters in combination with the designed coupling scheme; designing a motion control strategy based on the CPG model, considering gait characteristics and leg trajectory generation to obtain the actual input signal; and, driven by the actual input signal, generating rich gaits and smoothly switching between different gaits.

[0011] In this invention, the improved Hopf oscillator model constituting the basic unit of the CPG network is expressed as:

[0012]

[0013]

[0014]

[0015]

[0016] Wherein, the state variables of the Hopf oscillator are x i and z i and The positive parameter α affects the amplitude of the oscillation to converge. The speed, the frequency of oscillation is expressed as The unit is rad·s -1 The parameter ι∈{-1,1} determines the direction of the limit cycle, and the natural frequency of oscillator i is determined by ω. i This indicates that the parameter β i ∈(0,1) represents the waveform adjustment parameter, T represents the period parameter, and the positive parameter b determines the speed of frequency alternation. In this invention, if β i If β = 0.5, then the oscillator i is denoted as a symmetrical oscillator. i If ≠0.5, then the oscillator i is denoted as an asymmetric oscillator.

[0017] In this invention, the method for constructing the coupling scheme of the CPG network by performing configuration analysis on the salamander-inspired robot is as follows: the coupling between the symmetrical oscillators maintains a suitable phase relationship between the output signals, while the coupling from the symmetrical oscillator to the asymmetrical oscillator is used to adjust the waveform.

[0018] In this invention, the process of constructing the coupling scheme of the CPG network by performing configuration analysis on the salamander-like robot includes: constructing a body CPG and a limb CPG, wherein the body CPG is composed of multiple symmetrical oscillators bidirectionally coupled in a chain-like manner; for the body CPG, four symmetrical oscillators are fully connected and each of them is coupled to an asymmetrical oscillator in a unidirectional manner; the oscillators of the limb CPG and the oscillators of the body CPG are unidirectionally coupled in a radial manner.

[0019] In this invention, the construction process of the offline learning law and online update law of the coupling weight parameter by combining the designed coupling scheme includes: analyzing the dynamic process of a symmetrical oscillator coupled to an asymmetrical oscillator in a unidirectional connection; constructing an offline learning law of the coupling weight parameter between the two oscillators based on the conditions of synchronization of the two oscillators and the property of the phase angle difference between the two oscillators to obtain the critical value of the coupling weight parameter; and constructing an online update law of the actual coupling weight parameter based on the critical value.

[0020] In this invention, the expression for the offline learning law of the coupling weight parameters is:

[0021]

[0022]

[0023]

[0024]

[0025] Where w is the coupling weight parameter between the two oscillators, dt is the time interval, γ1 and γ2 are positive learning rate parameters, and φ d It is the phase angle difference between the two oscillators. According to the offline learning law of the coupling weight parameter, the coupling weight parameter between the two oscillators converges to the critical value w. * .

[0026] In this invention, the expression for the online update law of the coupling weight parameters is:

[0027]

[0028] Among them, w * k is the critical value of the coupling weight parameter between the two oscillators. w Let t be the gait switching time. s time w * The increase factor, parameter η, affects the rate at which the actual coupling weight parameter decays.

[0029] In this invention, the motion control strategy based on the CPG model, considering gait characteristics and leg trajectory generation, and the construction process of obtaining the actual input signal includes: designing the corresponding coupling weight parameters of the CPG model based on the coordination relationship between limbs and the coordination relationship between limbs and the body, constructing a leg trajectory generator based on the output signal of the CPG model, and then solving the inverse kinematics based on the leg motion trajectory generated by the leg trajectory generator to obtain the expected value of the joint angle of the salamander-like robot, which is used as the actual input signal. Under the drive of the actual input signal, the salamander-like robot generates rich gaits and smoothly switches between different gaits.

[0030] A second aspect of the invention provides a salamander-like robot motion control system based on a central pattern generator. The salamander-like robot motion control system includes a salamander-like robot, a ground station, and a router. The salamander-like robot has a segmented spine and legs with three degrees of freedom, and receives control commands from a controller configured to execute:

[0031] A configuration analysis was performed on the salamander-inspired robot to determine the coupling scheme of the CPG network. Based on the designed coupling scheme, offline learning laws and online update laws for the coupling weight parameters were constructed. A motion control strategy was designed based on the CPG model, taking into account gait characteristics and leg trajectory generation to obtain the actual input signal. Control commands were generated based on the actual input signal to enable the salamander-inspired robot to generate rich gaits and smoothly switch between different gaits.

[0032] A third aspect of the present invention provides a computer-readable storage medium.

[0033] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for motion control of a salamander-like robot based on a central pattern generator.

[0034] A fourth aspect of the present invention provides a computer device.

[0035] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the above-described method for motion control of a salamander-like robot based on a central pattern generator.

[0036] Compared with the prior art, the beneficial effects of the present invention are:

[0037] The coupling scheme of the CPG network serves as the basis of the method. The offline learning law of the coupling weight parameters will improve the accuracy of the output control signal waveform, while the online update law of the coupling weight parameters can accelerate the convergence speed of the CPG model. Under the motion control strategy designed based on the CPG model, not only can more accurate robot gait be achieved, but also rapid motion initialization and gait switching can be realized.

[0038] This invention has the potential to be further applied to other biomimetic robots and has significant practical implications.

[0039] The advantages of additional aspects of the present invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0040] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0041] Figure 1 This is a schematic diagram of the overall process of the salamander robot motion control method based on a central pattern generator according to the present invention;

[0042] Figure 2 The diagram shows the CPG network structure and the physical image of the newt-like robot platform of the present invention. Items numbered 11-14 represent asymmetric oscillators, while the rest are symmetric oscillators.

[0043] Figure 3 The diagram shows the sequence of asynchronous events generated by this invention. In the diagram, the black circle represents a leg in the support phase, the white circle and the dashed circle represent a leg landing and lifting, respectively, the arrows indicate the direction of leg movement, and the lines indicate the degree of torso bending.

[0044] Figure 4 This is a schematic diagram of the leg reference coordinate system and the generated trajectory of the present invention;

[0045] Figure 5 Screenshot of the gait switching experiment conducted for this invention;

[0046] Figure 6 This is a graph showing the convergence error of the present invention and the comparative method over time during gait switching. Detailed Implementation

[0047] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Furthermore, unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0049] Example 1:

[0050] like Figure 1 As shown, this embodiment provides a motion control method for a salamander-like robot based on a central pattern generator, including the following steps:

[0051] Step S101: Perform configuration analysis on the salamander-inspired robot to determine the coupling scheme of the CPG network;

[0052] like Figure 2 (a) shows a CPG network control system consisting of mutually coupled modified Hopf oscillators, as shown in the diagram. Figure 2 (b) shows a salamander-inspired robot. This CPG model consists of a body CPG and a limb CPG. The body CPG is composed of six symmetrical oscillators bidirectionally coupled in a chain-like arrangement; for the body CPG, four symmetrical oscillators are fully connected, each unidirectionally coupled to an asymmetrical oscillator; the oscillators of the limb CPG are unidirectionally coupled to the oscillators of the body CPG in a radial arrangement. Oscillators 1 and 6 are used to control the servo motors of the robot's head and tail, respectively; oscillators 2-5 control the torso; and oscillators 11-14 control the limbs. It is important to note that the coupling between the symmetrical oscillators maintains a suitable phase relationship between the output signals, while the coupling from the symmetrical oscillators to the asymmetrical oscillators is used to adjust the waveform. Phase coordination and waveform adjustment are independent of each other, and the corresponding coupling weight parameters can be set separately. Therefore, this invention can achieve rapid convergence of the CPG model while improving the accuracy of the output waveform.

[0053] In this embodiment, the dynamics of the improved Hopf oscillator i can be represented by the following differential equation:

[0054]

[0055]

[0056]

[0057]

[0058] Wherein, the state variables of the Hopf oscillator are x i and z i and The positive parameter α affects the amplitude of the oscillation to converge. The speed, the frequency of oscillation is expressed as The unit is rad·s -1 The parameter ι∈{-1,1} determines the direction of the limit cycle; the limit cycle will rotate clockwise if ι=1, and counterclockwise otherwise. The natural frequency of oscillator i is determined by ω. i It means that the parameter β i ∈(0,1) represents the waveform adjustment parameter. If β i =0.5, then Where T represents the period parameter. But if β i ≠0.5, the natural frequency will have different values ​​in different phases. For example, β i When z < 0.5, if i If ω > 0, then i It will accelerate, if z i <0 then ω i The deceleration will be accelerated, and the positive parameter b determines the speed of frequency alternation. When acceleration and deceleration are applied at different positions of the limit cycle, the output of the CPG network will be asymmetric. In this embodiment, if β i If β = 0.5, then the oscillator i is denoted as a symmetrical oscillator. i If ≠0.5, then the oscillator i is denoted as an asymmetric oscillator.

[0059] The second part of equation (3) represents the coupling effect between oscillator i and other oscillators, where the parameter w ij and φ ij These represent the coupling weight between oscillator i and oscillator j, and the expected phase angle difference, respectively. It should be noted that the coupling term is added to the oscillation frequency. The expression indicates that the coupling term affects the oscillator's frequency but not its amplitude; therefore, the coupling effect does not change the shape of the limiting cycle. The parameter ι in the coupling term ensures that the oscillators maintain the correct phase relationship when the direction of the limiting cycle changes.

[0060] Step S102: Construct the offline learning law and online update law for the coupling weight parameters based on the designed coupling scheme;

[0061] In this embodiment, a symmetrical output signal is converted into an asymmetrical output signal by unidirectional coupling of a symmetrical oscillator to an asymmetrical oscillator. The subsystem dynamics of symmetrical oscillator 1 acting on asymmetrical oscillator 2 can then be expressed as:

[0062]

[0063]

[0064]

[0065]

[0066] Where R1 = R2 = R, ι = 1 and φ 21 =0. Transforming equations (5)-(8) to polar coordinates, they can be expressed as:

[0067]

[0068]

[0069]

[0070]

[0071] It is understandable that r1 and r2 will converge from any non-zero position to To study the synchronization phenomenon of this system, we introduce the phase angle difference and its time derivative:

[0072] φ d =φ2-φ1, (13)

[0073]

[0074] If two oscillators have the same oscillation period, then the two oscillators are defined as synchronized. An asymmetric oscillator 2 has the following properties:

[0075] When there is no coupling, i.e., when oscillator 2 is isolated, its period is less than T. The analysis is as follows: when w 21 =0 and Then, equation (12) can be rearranged into the following equation:

[0076]

[0077] Integrating equation (15) over time yields:

[0078]

[0079] Where t' and φ' represent arbitrary initial time and phase, respectively, using the properties of the integrand within the integration interval, we can obtain:

[0080]

[0081] Using Cauchy's inequality:

[0082]

[0083] Therefore, we can draw the following conclusion:

[0084]

[0085] The property that the period of isolated oscillator 2 is less than T indicates that if there is no coupling or a small coupling weight between the two oscillators, the two oscillators will not be synchronized, and the phase difference φ will be significant. d It will fluctuate upwards. If w 21 If the value is sufficiently large, then the two oscillators will synchronize, φ d It will be a periodic function with a period of T. When the two oscillators are synchronized, w 21 It will also affect φ d The trajectory and equation φ d =0 is the number of solutions within a period. Specifically, when w 21 Less than the critical value The equation will have no solution; when w 21 equal to critical value Equation φ d =0 has a unique solution within one period; when w 21 Greater than the critical value The equation has two solutions within one period.

[0086] Equation φ d The waveform adjustment effect is optimal when the value of 0 has a unique solution within one cycle. Therefore, it is necessary to determine the critical value. Based on the conditions for oscillator synchronization and the properties of phase difference, the adaptive learning law in the form of a differential equation is constructed as follows:

[0087]

[0088]

[0089]

[0090]

[0091] Where dt is the time interval, and γ1 and γ2 are positive learning rate parameters, with γ1 >> γ2. In the initial stage of the learning process, the first term of the learning law plays a decisive role, up to φ. d For w with a period of T 21 It will grow. After the two oscillators are synchronized, the first term of the learning law tends to zero, and then φ d The trajectory is updated based on the latest minimum point. Finally, w 21When the critical value is reached:

[0092]

[0093] Substituting equation (24) into equation (14), we get:

[0094]

[0095] If the value of parameter b is large, then equation (25) approaches zero, indicating that equation φ d The value of the solution = 0 approaches a fixed point. One of them. The sign of x depends on whether β² is greater than 0.5 or less than 0.5. According to the constructed learning law, w 21 It will change with the dynamic changes of the two oscillators and eventually converge to the critical value.

[0096] Based on the constructed offline learning law for the coupling weight parameters, the critical value w of the coupling weight parameters from the symmetric oscillator to the asymmetric oscillator in the CPG model can be calculated. * This improves the accuracy of the output waveform. For various gaits with different duty cycles, w * They are also different. In the initial stages of gait generation and switching, a larger actual coupling weight parameter is desired to improve the convergence speed of the CPG model. Then, it decreases exponentially over time to w. * Therefore, the online update law for the coupling weight parameters is constructed as follows:

[0097]

[0098] Among them, w * k is the critical value of the coupling weight parameter from the symmetric oscillator to the asymmetric oscillator. w t is the time when gait switching occurs s Time w * The increase factor, parameter η, affects the rate at which the actual coupling weight parameter decays.

[0099] Step S103: Design a motion control strategy based on the CPG model, taking into account gait characteristics and leg trajectory generation, to obtain the actual input signal;

[0100] In this embodiment, three gait types are designed: lateral-sequence (LS) walk, trot, and backward walk. These three gait types can be distinguished by duty cycle and relative phase relationship. Assume that the duration of the swing phase of each gait is T. sw If they are the same, then the period of motion is determined by the duty cycle:

[0101]

[0102] In this embodiment, the waveform adjustment parameters of the limb oscillators 11-14 are consistent with the duty cycle of the gait, i.e., β. i =β, i = 11, 12, 13, 14. Meanwhile, the relative phase relationship can be expressed using a parameter φ. LH It means, i.e., φ LF =0,φ RF =0.5,φ RH =φ LH -0.5. From this, we can obtain the matrix Φ containing the phase difference of oscillators 7-10, whose expression is:

[0103]

[0104] The row and column indices of the matrix are 7, 8, 9, and 10, respectively. Based on the duty cycle and the relative phase relationship between the limbs, the distribution of the gait event sequence over time can be obtained, such as... Figure 3 As shown, the events ψ represents the landing and lifting of a leg, respectively. Given β and φ... LH This allows for proper limb coordination. Detailed parameters for the three gait types are shown in Table 1:

[0105] Table 1 Gait parameters

[0106]

[0107] Furthermore, for salamander-like robots with actively flexing spines, appropriate body-limb coordination can increase stride length and improve movement speed. In this embodiment, body-limb coordination is described as follows: when moving forward, as one forelimb is about to lift, the torso bends to the same side to its maximum extent (event χ), i.e., ψ. RF =χ R ,ψ LF =χ L When moving backward, as one forelimb is about to lift up, the trunk bends to its maximum extent to the other side, i.e., ψ. RF =χ L ,ψ LF =χ R The two possible pairings can be achieved by configuring the following parameters:

[0108]

[0109] At the same time, the tail and torso are out of phase to maintain the robot's balance, from which we can conclude:

[0110]

[0111] The coupling effect from the body oscillator to the limb oscillator causes the robot's body to oscillate in an S-shape, counteracting the tendency to generate traveling waves caused by the coupling effect between the body oscillators.

[0112] φ i,i+1 =-φ i+1,i =-φ b i∈{1,...,5}, (31)

[0113] Where φ b This is the phase value of a positive constant.

[0114] In this embodiment, the state variables of oscillators 11-14 are used to generate the foot end-effector trajectory. For example... Figure 4 As shown, the origin of the reference coordinate system for the left foreleg is located at the center of the first joint. Its y-axis is parallel to the forearm, and its z-axis points vertically downwards. The trajectory of the foot's end in Cartesian space is shown. The expression is:

[0115]

[0116]

[0117]

[0118] Where d w and d h d represents the distance from point T0 along the x-axis and z-axis, where T0 corresponds to the initial position of the foot's end. c It is the swing distance between T0 and T1, d p It is the support distance between T0 and T3, d l It is the stride length between T0 and T2, equivalent to the distance between T0 and T4. T1, T2, T3, and T4 represent points on the stable leg trajectory. θ represents the angle between the plane containing the foot's end trajectory and the yz plane, used to adjust the distance of the trajectory along the x-axis so that the trajectory is within the reachable space of the leg.

[0119] Step S104: Driven by the actual input signal, the salamander-like robot generates rich gaits and smoothly switches between different gaits.

[0120] In this embodiment, the state variable x of oscillators 1-6 i The desired angles are directly used as the servo motors on the body joints, while the desired angles on the leg joints are obtained by solving the inverse kinematics. All desired angles are simultaneously sent to the proportional-integral-derivative (PID) controller of the servo motors for execution.

[0121] To produce a smooth gait transition, gait parameters β and φ LHWith first-order dynamic change:

[0122]

[0123] Where k p It is a time constant, u and These are the actual values ​​and the set expected values ​​of the gait parameters, respectively.

[0124] In this embodiment, following the steps described above, testing was conducted on a self-built platform to verify the effectiveness of the salamander-inspired robot motion control method based on a central pattern generator described in this invention. The main parameters of the controller are given in Table 2, where R is given for all oscillators. i =R=μ 2 Furthermore, the amplitude μ depends on the robot platform.

[0125] Table 2 Controller Parameters

[0126]

[0127] In this embodiment, the w values ​​of the LS walk, trot, and backward walk gaits obtained according to the offline learning law are... * The values ​​are 0.273, 0.162, and 0.280, respectively. The remaining parameters are given in Table 2. It should be noted that the coupling weight parameters between the symmetric oscillators are all set to constant values. During the test, the salamander-inspired robot was initialized in trot gait, then switched to LS walk gait starting at 10s, and then switched to backward walk gait starting at 26s. Screenshots of the salamander-inspired robot's movement are shown below. Figure 5 As shown, it can be clearly seen that the robot successfully executed the three designed gaits while maintaining appropriate body-limb coordination throughout the process. In summary, the gait transitions were rapid, smooth, and natural.

[0128] In this embodiment, the method proposed in this invention is compared with the prior art to verify the superiority of the salamander robot motion control method based on a central pattern generator described in this invention.

[0129] This embodiment compares the proposed method with the method in reference [3], and improves upon the method in reference [2] by adding an adjustable coupling weight parameter:

[0130]

[0131] Where R(φ) ij ) is a rotation matrix, and its expression is:

[0132]

[0133] In the comparative method, four pairs of asymmetric oscillators are directly coupled together to form the body CPG. The coupling weights are set to a smaller value K = 0.04 and a larger value K = 0.28, while other parameter settings are the same as in this embodiment. On one hand, this embodiment compares the synchronization error during gait switching, calculated as follows:

[0134]

[0135] Where Φ ij The elements of the matrix in equation (28). Figure 6 As shown, the salamander-inspired robot switches from a stable trot gait to an LS walk gait in 40 seconds. When K is 0.04 in the comparison method, the magnitude of the error E(t) decreases slowly, indicating a relatively slow gait switching. When K is 0.28, although the magnitude of the error E(t) decreases more quickly, its steady-state error is larger, which is caused by the inaccuracy of the output waveform. Under the coupling scheme and online parameter update law proposed in this invention, the magnitude of E(t) decreases more quickly and eventually remains within a minimum range.

[0136] On the other hand, this embodiment evaluates the waveform accuracy in a stable gait mode by comparing the actual and nominal values ​​of the period T, duty cycle β, and amplitude μ. This embodiment measures the average over one hundred periods and summarizes the results in Table 3. It is worth noting that the amplitude μ is normalized by dividing by π / 12. From Table 3, the following conclusions can be drawn: 1. When asymmetric oscillators are directly coupled, a larger coupling weight will shorten the motion period, causing the duty cycle to change towards 0.5 and increasing the amplitude. 2. Compared to isolated oscillators, the method proposed in this invention can not only adjust the period to equal its nominal value and adjust the duty cycle to be close to its nominal value, but also maintain a constant amplitude. 3. The data in Table 3 and... Figure 6 The error curves in the figures together verify the performance improvement brought about by the method proposed in this invention.

[0137] Table 3 Quantitative data on waveform accuracy

[0138]

[0139] In summary, the salamander-inspired robot motion control method proposed in this invention, based on a central pattern generator, can improve the accuracy of the output control signal waveform and accelerate the convergence speed of the CPG model. Under the action of the designed control strategy, it can not only achieve more accurate robot gait but also achieve faster motion initialization and gait switching.

[0140] Example 2:

[0141] This embodiment provides a salamander-like robot motion control system based on a central pattern generator. The system comprises a salamander-like robot, a ground station, and a router. The salamander-like robot has a segmented spine and legs with three degrees of freedom, and receives control commands from a controller configured to execute:

[0142] A configuration analysis was performed on the salamander-inspired robot to determine the coupling scheme of the CPG network. Based on the designed coupling scheme, offline learning laws and online update laws for the coupling weight parameters were constructed. A motion control strategy was designed based on the CPG model, taking into account gait characteristics and leg trajectory generation to obtain the actual input signal. Driven by the actual input signal, the gait of the salamander-inspired robot was generated and switched, and control commands were generated.

[0143] In this embodiment, the actuator of the self-built salamander-inspired robot uses a Dynamixel XW-540-T260 servo motor, while the remaining body connecting parts are 3D printed and some are machined from aluminum alloy. The head of the salamander-inspired robot is equipped with a small, low-power, and high-performance onboard computing unit, the NVIDIA Jetson Xavier NX, which sends control commands to the Dynamixel servo motor via the RS-485 communication protocol.

[0144] In this embodiment, the ground station remotely controls the onboard computing unit NVIDIA Jetson Xavier NX via a WIFI signal emitted by the router, causing it to execute the controller program.

[0145] Example 3:

[0146] A third aspect of the present invention provides a computer-readable storage medium.

[0147] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for motion control of a salamander-like robot based on a central pattern generator.

[0148] A fourth aspect of the present invention provides a computer device.

[0149] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the above-described method for motion control of a salamander-like robot based on a central pattern generator.

[0150] Those skilled in the art will understand that each module or step of the present invention can be implemented using a general-purpose computer device. On the one hand, they can be implemented using computer-executable program code, which can then be stored in a storage device for execution by a computer. On the other hand, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0151] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A motion control method for a salamander-like robot based on a central pattern generator, characterized in that, Includes the following steps: A configuration analysis was performed on the salamander-inspired robot to determine the coupling scheme of the CPG network; Construct offline learning and online update laws for the coupling weight parameters based on the designed coupling scheme; A motion control strategy is designed based on the CPG model, taking into account gait characteristics and leg trajectory generation to obtain the actual input signal; Driven by actual input signals, the salamander-like robot generates a variety of gaits and smoothly switches between different gaits; The improved Hopf oscillator model constituting the basic unit of the CPG network is expressed as: The state variables of the Hopf oscillator are respectively and and positive parameter The amplitude of the oscillation converges to The speed, the frequency of oscillation is expressed as The unit is ,parameter Determines the direction of the limit cycle, oscillator The natural frequency is determined by Indicates that the parameters Represents waveform adjustment parameters. Represents a periodic parameter, a positive parameter. Determine the speed of frequency alternation, if The oscillator It is denoted as a symmetrical oscillator, if The oscillator This is referred to as an asymmetric oscillator; The offline learning law expression for the coupling weight parameters is: in It is the coupling weight parameter between the two oscillators. It is a time interval. and It is a positive learning rate parameter. It is the phase angle difference between the two oscillators. According to the offline learning law of the coupling weight parameter, the coupling weight parameter between the two oscillators converges to a critical value. Based on this critical value, the expression for the online update law of the coupling weight parameters is: in, This is the critical value of the coupling weight parameter between the two oscillators. For gait switching time The increase factor, parameter It affects the rate at which the actual coupling weight parameters decay.

2. The method for motion control of a salamander-like robot based on a central pattern generator according to claim 1, characterized in that, The method for constructing the coupling scheme of the CPG network by performing configuration analysis on the salamander-inspired robot is as follows: The coupling between the symmetrical oscillators maintains a suitable phase relationship between the output signals, while the coupling from the symmetrical oscillator to the asymmetrical oscillator is used to adjust the waveform.

3. The method for motion control of a salamander-like robot based on a central pattern generator according to claim 2, characterized in that, The process of constructing the coupling scheme of the CPG network by performing configuration analysis on the salamander-inspired robot includes: Construct a body CPG and a limb CPG. The body CPG consists of multiple symmetrical oscillators bidirectionally coupled in a chain-like manner. For the body CPG, four symmetrical oscillators are fully connected, and each of them is coupled to an asymmetrical oscillator in a unidirectional manner. The oscillators of the limb CPG and the oscillators of the body CPG are unidirectionally coupled in a radial manner.

4. The method for motion control of a salamander-like robot based on a central pattern generator according to claim 1, characterized in that, The process of constructing the offline learning law and online update law of the coupled weight parameters using the combined design coupling scheme includes: The dynamic process of a symmetrical oscillator coupled to an asymmetrical oscillator in a unidirectional connection is analyzed. Based on the synchronization condition of the two oscillators and the property of the phase angle difference between the two oscillators, an offline learning law for the coupling weight parameter between the two oscillators is constructed to obtain the critical value of the coupling weight parameter. Based on the critical value, an online update law for the actual coupling weight parameter is constructed.

5. The method for motion control of a salamander-like robot based on a central pattern generator according to claim 1, characterized in that, The motion control strategy based on the CPG model, considering gait characteristics and leg trajectory generation, is constructed as follows: The corresponding coupling weight parameters of the CPG model are designed based on the coordination relationships between limbs and between limbs and the body. A leg trajectory generator is constructed based on the output signal of the CPG model. Then, the inverse kinematics are solved based on the leg motion trajectory generated by the leg trajectory generator to obtain the expected values ​​of the joint angles of the salamander-like robot, which serve as the actual input signal.

6. A motion control system for a salamander-like robot based on a central pattern generator, characterized in that, The system includes a salamander-like robot, a ground station, and a router. The salamander-like robot has a segmented spine and three-degree-of-freedom legs, and receives control commands from a controller configured to execute: A configuration analysis of a salamander-inspired robot was conducted to determine the coupling scheme of the CPG network. Based on the designed coupling scheme, offline learning laws and online update laws for coupling weight parameters were constructed. A motion control strategy was designed based on the CPG model, taking into account gait characteristics and leg trajectory generation to obtain the actual input signal. Control commands were generated based on the actual input signal to enable the salamander-inspired robot to generate rich gaits and smoothly switch between different gaits. The offline learning law expression for the coupling weight parameters is: in It is the coupling weight parameter between the two oscillators. It is a time interval. and It is a positive learning rate parameter. It is the phase angle difference between the two oscillators. According to the offline learning law of the coupling weight parameter, the coupling weight parameter between the two oscillators converges to a critical value. Based on this critical value, the expression for the online update law of the coupling weight parameters is: in, This is the critical value of the coupling weight parameter between the two oscillators. For gait switching time The increase factor, parameter It affects the rate at which the actual coupling weight parameters decay.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of a salamander robot motion control method based on a central pattern generator according to any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the motion control method for a salamander robot based on a central pattern generator according to any one of claims 1-5.

Images