A robot bionic neural network control method and system

Abstract

The embodiment of the present invention discloses a robot bionic neural network control method and system, the method comprising: evaluating the robot control model to obtain multiple control modules; constructing a single neuron model corresponding to each control module according to the coupling relationship and phase relationship between each control module; generating control signals for each control module in different motion modes based on each single neuron model, so as to control the bionic robot to complete the motion in different motion modes by using the control signal. Based on the multivariate neuron model formed by the excitation and inhibition coupling relationship between each control module, it can simultaneously self-excite and oscillate to generate multiple stable control signals with adjustable amplitude, frequency, phase and waveform, so as to ensure the stability of the whole motion of the bionic robot and the flexibility of state switching, thereby achieving a motion control effect close to that of natural organisms.

Description

Technical Field

[0001] The present invention relates to the field of neural control technology, and in particular to a robot bionic neural network control method and system. Background Art

[0002] Bionic robots are robots that mimic biological physiological structures and motion functions. They simulate the physical form and motion characteristics of organisms as much as possible in design and function. In the field of dynamic modeling and control, the research on bionic robots mainly focuses on how to simulate the dynamic behavior of organisms through accurate mathematical models and realize complex motion patterns through control strategies. This process includes in-depth research on the motion mechanism and neural manipulation of organisms, as well as the transformation of these biological mechanisms into implementable mechanical design and control algorithms.

[0003] Many organisms can maintain high-frequency rhythmic stable movements (such as flapping wings, running and swimming) and can also quickly adjust their posture and movement state according to environmental feedback. Therefore, their biological neural control mechanism has extremely high research value. For example, through biological neural networks and deep learning algorithms, bionic robots can still achieve stable and adaptable action responses when facing unknown environments or unpredicted situations. These advances have greatly promoted the application potential of bionic robots in complex environments, such as disaster relief, deep-sea exploration, and off-road transportation in complex terrain.

[0004] Currently, in the motion control of bionic robots, dynamic balance needs to be maintained during walking, running, etc., especially in complex terrain and uncertain environments; and bionic robots usually have multiple joints and degrees of freedom, and achieving coordinated motion control of these joints is a difficult point. Summary of the invention

[0005] Based on this, it is necessary to propose a robot bionic neural network control method and system to address the above problems.

[0006] To achieve the above-mentioned purpose, the present application provides a robot bionic neural network control method in a first aspect, the method comprising:

[0007] Evaluate the bionic robot model and obtain multiple control modules;

[0008] According to the coupling relationship and phase relationship between various control modules, a monomer neuron model corresponding to each control module is constructed;

[0009] Based on each of the monomer neuron models, control signals of each of the control modules in different motion modes are generated, so as to utilize the control signals to control the bionic robot to complete the motion in different motion modes.

[0010] Furthermore, the monomer neuron model corresponding to each control module is constructed according to the coupling relationship and phase relationship between each control module, specifically including:

[0011] Acquire a second control module, a coupling characteristic degree parameter between the oscillators of the first control module and the second control module, a phase difference between the first control module and the second control module, and respectively acquire other characteristic parameters of the first control module and the second control module, wherein the first control module is any one of all the control modules of the bionic robot, and the second control module is any other control module in the bionic robot that is coupled with the first control module;

[0012] Based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module, a monomer neuron model corresponding to each control module is constructed.

[0013] Furthermore, constructing a monomer neuron model corresponding to each control module based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module specifically includes:

[0014] Calculate according to the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module and the linear link related to the frequency of the first control module;

[0015] Obtaining a waveform integration function of the first control module based on the asymmetric phase factor dynamic parameter and the linear link;

[0016] A monomer neuron model corresponding to each control module is constructed according to the waveform integration function and other characteristic parameters of the first control module.

[0017] Furthermore, the other characteristic parameters include at least: motion parameters and frequency;

[0018] Then, the calculation is performed according to the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module and the linear link related to the frequency of the first control module, specifically including:

[0019] Calculating according to the coupling characteristic degree parameter, the phase difference, and the motion parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module;

[0020] A linear link related to the frequency of the first control module is obtained by performing calculation according to the frequency of the first control module and a preset frequency coefficient.

[0021] Further, the motion parameters include at least: amplitude, instantaneous phase and limited amplitude;

[0022] Then, the calculation is performed according to the coupling characteristic degree parameter, the phase difference, and the motion parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module, specifically including:

[0023] Performing derivative calculation according to the coupling characteristic degree parameter, the phase difference, the frequency and instantaneous phase of the first control module, and the amplitude and instantaneous phase of the second control module to obtain the instantaneous phase derivative of the first control module;

[0024] Performing derivative calculation according to the limited amplitude of the first control module and the amplitude of the first control module to obtain the amplitude derivative of the first control module;

[0025] The asymmetric phase factor dynamic parameter is calculated according to the amplitude derivative of the first control module and the instantaneous phase derivative of the first control module.

[0026] Furthermore, constructing a monomer neuron model corresponding to each control module according to the waveform integration function and other characteristic parameters of the first control module specifically includes:

[0027] Acquire a signal actual scaling operation factor corresponding to the first control module in a preset signal actual scaling operation factor table, wherein the signal actual scaling operation factor table includes a correspondence between the first control module and the signal actual scaling operation factor;

[0028] A monomer neuron model corresponding to each control module is constructed according to the actual signal scaling operation factor corresponding to the first control module, the waveform integration function and the characteristic parameters of the first control module.

[0029] Furthermore, the monomer neuron model is expressed as:

[0030]

[0031]

[0032]

[0033]

[0034]

[0035]

[0036] In the formula, i is a first control module, j is the second control module, is the instantaneous phase derivative of the first control module, v i is the frequency of the first control module, r j is the amplitude of the second control module, w ij is the coupling characteristic degree parameter, θ j is the instantaneous phase of the second control module, θ i is the instantaneous phase of the first control module, is the phase difference between the first control module and the second control module, is the amplitude derivative of the first control module, a is the convergence speed coefficient, R i is the limiting amplitude of the first control module, r i is the amplitude of the first control module, d i is the asymmetric phase factor dynamic parameter, f ( d i ) waveform integration function, k 0 is the waveform adjustment factor, y i is a linear link related to the frequency of the first control module, x i is the control signal of the first control module, C i is the actual scaling factor of the signal, K is the proportionality coefficient.

[0037] Furthermore, the method further comprises:

[0038] When controlling the bionic robot to switch between the two motion modes, a transition state is inserted between the two motion modes to control the bionic robot to execute the transition state.

[0039] Further, the transition state is expressed as:

[0040]

[0041] in, is a transition state. is the phase relationship of the current motion mode, To select the phase relationship of the next motion mode to switch, c represents the time coefficient, Indicates the time it takes to switch the motion mode. It is a dynamic variable related to time. The maximum time it takes to switch between sport modes.

[0042] To achieve the above-mentioned purpose, the second aspect of the present application provides a robot bionic neural network control system, the system comprising: a system decomposition module, a model construction module and a signal generation module;

[0043] The system decomposition module is used to evaluate the bionic robot model to obtain multiple control modules;

[0044] The model building module is used to build a single neuron model corresponding to each control module according to the coupling relationship and phase relationship between the control modules;

[0045] The signal generating module is used to generate control signals of each control module in different motion modes based on each monomer neuron model, so as to use the control signals to control the bionic robot to complete the motion in different motion modes.

[0046] The embodiments of the present invention have the following beneficial effects:

[0047] A robot bionic neural network control method proposed in an embodiment of the present invention includes: evaluating a bionic robot model to obtain multiple control modules; constructing a monomer neuron model corresponding to each control module according to the coupling relationship and phase relationship between each control module; generating control signals for each control module in different motion modes based on each monomer neuron model, so as to control the bionic robot to complete motion in different motion modes by using the control signal. By constructing the monomer neuron model corresponding to each control module according to the coupling relationship and phase relationship between each control module of the bionic robot, a multivariate neuron model of the bionic robot is obtained, and the control signal generated by the monomer neuron model taking into account the coupling factors between each control module can make each control module output stably, thereby realizing effective control of the bionic robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] in:

[0050] Figure 1 A schematic diagram of a flow chart of a robot bionic neural network control method according to an embodiment of the present invention;

[0051] Figure 2 A physical model of the bionic butterfly robot in an embodiment of the present invention;

[0052] Figure 3 The motion model of the bionic butterfly robot in the embodiment of the present invention;

[0053] Figure 4 A comparison diagram of motion signal curves of a single neuron under different waveform adjustment factors in an embodiment of the present invention;

[0054] Figure 5 (a) is a diagram showing the coupling relationship and phase relationship between various control modules of the bionic robot in the motion mode according to an embodiment of the present invention;

[0055] Figure 5 (b) is a diagram showing the coupling relationship and phase relationship between the control modules of the bionic robot in motion mode 2 in an embodiment of the present invention;

[0056] Figure 6 A control signal diagram for switching motion modes when a transition state is added to an embodiment of the present invention;

[0057] Figure 7 This is a control signal diagram when the motion mode is switched without adding a transition state in an embodiment of the present invention;

[0058] Figure 8 This is a structural block diagram of a robot bionic neural network control system in an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the embodiments of the present invention and the accompanying drawings to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] See also Figure 1 , Figure 1 1 is a flow chart of a robot bionic neural network control method according to an embodiment of the present invention, which specifically includes:

[0061] Step 110, evaluating the bionic robot model to obtain multiple control modules.

[0062] In the embodiment of the present invention, the bionic robot can be a bionic butterfly robot. Since the goal of the bionic robot is to be able to perform complex motion tasks, such as walking, jumping or flying, etc. Therefore, when evaluating the bionic robot, the physical and mathematical relationships between its various parts and its motion state should be included, and factors such as mass distribution, joint connection, driving mode, force conditions, degrees of freedom and motion relationships of the actual system should be combined to ensure that the constructed multivariate neuron model conforms to the actual situation.

[0063] Specifically, the bionic robot model can be evaluated to obtain the modules that actually need to be controlled in the bionic robot. By controlling each module to move separately, the bionic robot can perform complex motion tasks and ensure the stability and efficiency of the motion.

[0064] Step 120: construct a single neuron model corresponding to each control module according to the coupling relationship and phase relationship between the control modules.

[0065] Since there are certain coupling relationships and phase relationships between the various control modules of the bionic robot model, the motion states of the various control modules affect each other. Therefore, the embodiment of the present invention constructs a multi-neural control model composed of multiple monomer neuron models. Specifically, according to the coupling relationship and phase relationship between the various control modules, the monomer neuron model corresponding to each control module is generated, and the multi-neural control model of the bionic robot is constructed by multiple monomer neurons, and then the control signals of each control module in different motion modes are generated based on the multi-neural control model.

[0066] Step 130 , generating control signals for each control module in different motion modes based on each monomer neuron model, so as to control the bionic robot to complete motion in different motion modes by using the control signals.

[0067] After constructing the monomer neuron model corresponding to each control module, a control signal can be generated based on the multivariate neural control model that takes into account the coupling relationship and phase relationship between the control modules, so that each control module of the bionic robot moves according to the corresponding control signal, thereby realizing the movement of the bionic robot in different motion modes.

[0068] In a feasible embodiment of the present invention, the bionic robot may be a bionic butterfly robot. Figure 2 , Figure 2 is the physical model of the bionic butterfly robot in the embodiment of the present invention, such as Figure 2 As shown in the figure, the main modules of the bionic butterfly robot include T rigid body, which represents the chest of the bionic butterfly robot in actual movement, and the other three modules are R , L , AThe rigid bodies are modules that need to be controlled, representing the right wing, left wing and abdomen of the bionic butterfly robot in actual movement.

[0069] Figure 3 The motion model of the bionic butterfly robot in the embodiment of the present invention is constructed by the degrees of freedom and motion relationship of the actual system of the bionic butterfly robot. Here, the ternary neuron control model is introduced in detail. Figure 3 As shown, R , L , A The rigid body control modules correspond to R W Control module, LW Control module, AB Control module, where the black line represents the phase relationship between the two control modules, the thick black line represents a strong coupling relationship, and the thin black line represents a weak coupling relationship. According to the coupling relationship and phase relationship between the above control modules, the monomer neuron model corresponding to each control module can be constructed, and based on each monomer neuron model, a ternary neuron control model that conforms to the bionic butterfly robot is constructed. The ternary neuron control model can be used to generate control signals in different motion modes to control the movement of each control module of the bionic butterfly robot.

[0070] In the embodiment of the present invention, step 120, based on the coupling relationship and phase relationship between the control modules, constructing a monomer neuron model corresponding to each control module, specifically includes:

[0071] Step 121, obtain the second control module, and the coupling characteristic degree parameters between the oscillators of the first control module and the second control module, the phase difference between the first control module and the second control module, and respectively obtain other characteristic parameters of the first control module and the second control module, wherein the first control module is any one of all the control modules of the bionic robot, and the second control module is any other control module in the bionic robot that has a coupling relationship with the first control module.

[0072] Specifically, other control modules that are coupled with the first control module are obtained. Taking the bionic butterfly robot as an example, assuming that the first control module is the RW control module corresponding to the R rigid body, other control modules that are coupled with the RW control module are found in all control modules of the bionic butterfly robot. In the bionic butterfly robot, the AW control module and the LW control module are both coupled with the RW control module, so the AW control module and the LW control module can be used as the second control module.

[0073] After acquiring the second control module, the coupling characteristic degree parameter between the oscillator values ​​of the first control module and the second control module, the phase difference between the first control module and the second control module, and other characteristic parameters of the first control module and the second control module can be acquired. Other characteristic parameters may include: instantaneous phase, amplitude, limited amplitude, frequency and other parameters.

[0074] Step 122, based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module, construct a monomer neuron model corresponding to each control module.

[0075] Specifically, based on other characteristic parameters obtained from the actual motion characteristics of the bionic robot parts corresponding to the first control module and the second control module, as well as the coupling characteristic degree parameters and phase difference between the two control modules, the monomer neuron model corresponding to each control module is constructed. Taking the bionic butterfly robot as an example, after obtaining the coupling characteristic parameters and phase difference between the oscillators of the RW control module and the AW control module and the LW control module, as well as other characteristic parameters of the RW control module, the AW control module and the LW control module, the monomer neuron model corresponding to the RW control module can be constructed based on the above parameters.

[0076] In one embodiment of the present invention, Step 122, based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module, constructing a monomer neuron model corresponding to each control module specifically includes:

[0077] A. Calculate according to the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module and the linear link related to the frequency of the first control module.

[0078] In the embodiment of the present invention, other characteristic parameters at least include: motion parameters and frequency, then the asymmetric phase factor dynamic parameters of the first control module and the linear link related to the frequency of the first control module are obtained by the following method:

[0079] A1. Calculate the asymmetric phase factor dynamic parameters of the first control module according to the coupling characteristic degree parameters, phase difference, and motion parameters of the first control module and the second control module.

[0080] Specifically, the motion parameters at least include: amplitude, instantaneous phase and limited amplitude. Then, the asymmetric phase factor dynamic parameters of the first control module are calculated based on the motion parameters of the first control module and the second control module, and specifically include:

[0081] A11. Perform derivative calculation according to the coupling characteristic degree parameter, the phase difference, the frequency and instantaneous phase of the first control module, and the amplitude and instantaneous phase of the second control module to obtain the instantaneous phase derivative of the first control module.

[0082] In the embodiment of the present invention, the instantaneous phase derivative of the first control module is calculated by the following formula:

[0083]

[0084] In the formula, is the instantaneous phase derivative of the first control module, v i is the frequency of the first control module, r j is the amplitude of the second control module, w ij is the coupling characteristic degree parameter, θ j is the instantaneous phase of the second control module, is the instantaneous phase of the first control module, is the phase difference between the first control module and the second control module.

[0085] A12. Perform derivative calculation according to the limited amplitude of the first control module and the amplitude of the first control module to obtain the amplitude derivative of the first control module.

[0086] In the embodiment of the present invention, the amplitude derivative of the first control module is calculated by the following formula:

[0087]

[0088] In the formula, is the amplitude derivative of the first control module, a is the convergence speed coefficient, R i is the limiting amplitude of the first control module, r i is the amplitude of the first control module.

[0089] A13. Calculate the asymmetric phase factor dynamic parameter according to the amplitude derivative of the first control module and the instantaneous phase derivative of the first control module.

[0090] In the embodiment of the present invention, the asymmetric phase factor dynamic parameter is calculated by the following formula:

[0091]

[0092] In the formula, d i is the asymmetric phase factor dynamic parameter, is the amplitude derivative of the first control module, is the instantaneous phase of the first control module, r i is the amplitude of the first control module, is the instantaneous phase derivative of the first control module.

[0093] A2. Calculate according to the frequency of the first control module and a preset frequency coefficient to obtain a linear link related to the frequency of the first control module.

[0094] In the embodiment of the present invention, the linear link related to the frequency of the first control module is calculated by the following formula:

[0095]

[0096] In the formula, y i is a linear link related to the frequency of the first control module, k is the proportionality coefficient, c i is a constant.

[0097] In the embodiment of the present invention, an activation function is used to obtain a waveform integration function, and the activation function expression is as follows:

[0098]

[0099] The activation function has the characteristic of taking a value range of (0,1), which can map a real number to the interval of (0,1), and if x If the function is periodic, the value of the function will also show a periodic property within (0,1), which is beneficial for subsequent analysis. x is the dynamic parameter of the asymmetric phase factor d i Therefore, the asymmetric phase factor dynamic parameter d i It should also be periodic.

[0100] Based on this, the embodiment of the present invention performs interval scaling processing through a linear link, and the asymmetric phase factor dynamic parameter can be d i Through creative labor, it is found that when other static parameters of the single neuron model are set well, the dynamic parameter of the asymmetric phase factor d i The amplitude and input frequency vi Therefore, a linear fitting is performed according to the frequency of the first control module to adjust the asymmetric phase factor dynamic parameter d i The value range of .

[0101] B. The waveform integration function of the first control module is obtained based on the asymmetric phase factor dynamic parameters and the linear link.

[0102] Combining the formulas in steps A1 and A2 above, the waveform integration function of the first control module in the embodiment of the present invention is expressed by the following formula:

[0103]

[0104] In the formula, d i is the asymmetric phase factor dynamic parameter, f ( d i ) waveform integration function, k 0 is the waveform adjustment factor, y i is a linear link related to the frequency of the first control module.

[0105] In the embodiment of the present invention, the waveform adjustment factor k 0 can change the ratio of the up and down cycles in a single cycle of the waveform within a certain range, while maintaining the smoothness of the waveform within a certain range. Figure 4 , Figure 4 : is a comparison diagram of the motion signal curves of a single neuron under different waveform adjustment factors in an embodiment of the present invention, where B, C, D, E, and F represent the waveforms when the waveform adjustment factors are 0.1, 0.3, 0.6, 0.8, and 1.0, respectively. It can be seen that as the waveform adjustment factor k 0, the ratio of the up cycle to the down cycle in a single cycle of the waveform is also changing. Without changing the overall cycle, the waveform adjustment factor is adjusted according to the actual application scenario, thereby changing the ratio of the up cycle to the down cycle in a single cycle of the waveform.

[0106] Taking the bionic butterfly robot as an example, in order to make the bionic butterfly robot fit the actual butterfly flight motion characteristics, the waveform adjustment factor needs to be adjusted. Since the butterfly's flight is a low-frequency flight motion, the ratio of the uplink to downlink cycles in a single cycle observed during the flight process is close to 0.5 to 0.8. Therefore, combined with the waveform adjustment factor, the waveform integration function is used to adjust the monomer neuron model so that the output control signal is in line with the actual situation.

[0107] C. Construct the monomer neuron model corresponding to each control module according to the waveform integration function and other characteristic parameters of the first control module.

[0108] In the embodiment of the present invention, the monomer neuron model is represented by the following formula:

[0109]

[0110] In the formula, x i is the control signal of the first module, C i is the actual scaling factor of the signal, f ( d i ) waveform integration function, r i is the amplitude of the first control module, θ i is the instantaneous phase of the first control module, K is the proportionality coefficient.

[0111] In an embodiment of the present invention, the actual signal scaling operation factor is related to different control systems. In different bionic robots, the actual signal scaling operation factors corresponding to each control module may also be different. The actual signal scaling operation factor can be set according to the actual required signal amplitude. Then step C, based on the waveform integration function and the characteristic parameters of the first control module, constructs the monomer neuron model corresponding to each control module, specifically including: obtaining the actual signal scaling operation factor corresponding to the first control module in the preset actual signal scaling operation factor table, the actual signal scaling operation factor table contains the correspondence between the first control module and the actual signal scaling operation factor; constructing the monomer neuron model corresponding to each control module according to the actual signal scaling operation factor corresponding to the first control module, the waveform integration function and the characteristic parameters of the first control module.

[0112] In a feasible embodiment of the present invention, the bionic robot is a bionic butterfly robot, and the single neuron model of the RW control module is fully expressed as follows:

[0113]

[0114]

[0115]

[0116]

[0117]

[0118]

[0119] In the formula, i For the RW control module, j For AW or LW control modules, is the instantaneous phase derivative of the RW control module, v i The frequency of the RW control module,r j The amplitude of the AW or LW control module, w ij is the characteristic degree parameter of the coupling between the RW control module and the oscillator of the AW or LW control module, θ j is the instantaneous phase of the AW or LW control module, θ i is the instantaneous phase of the RW control module, is the phase difference between the RW control module and the AW or LW control module, is the amplitude derivative of the RW control module, a is the convergence speed coefficient, R i is the limit amplitude of the RW control module, r i is the amplitude of the RW control module, d i is the asymmetric phase factor dynamic parameter, f ( d i ) waveform integration function, k 0 is the waveform adjustment factor, y i is a linear link related to the frequency of the RW control module, x i It is the control signal of RW control module. C i is the actual scaling factor of the signal, K is the proportionality coefficient.

[0120] The present invention also proposes a transition state, which is used to insert the transition state between the two motion modes when the bionic robot switches between the two motion modes, and control the bionic robot to execute the transition state after executing the previous motion mode and before executing the next motion mode, so that the bionic robot can switch motion modes smoothly.

[0121] According to actual biological observations, biological motion is divided into different modes. For example, during the flight of a butterfly, there are basic flight modes such as climbing flight and hovering flight. The corresponding phase relationship matrix and coupling relationship matrix in different motion states will be different, and the generated control signals will also be different. Here we will explain the two motion modes of the butterfly's flight motion.

[0122] See also Figure 5 , Figure 5 (a) is a diagram showing the coupling relationship and phase relationship between various control modules of the bionic robot in the motion mode according to an embodiment of the present invention. Figure 5(b) is a diagram showing the coupling relationship and phase relationship between the control modules of the bionic robot in motion mode 2 according to an embodiment of the present invention. L rigid body, R rigid body, A The rigid body corresponds to the left wing, right wing, and abdomen of the bionic butterfly robot. , and , They represent the coupling relationship matrix and phase relationship matrix under motion mode 1 and motion mode 2 respectively, where motion mode 1:

[0123]

[0124]

[0125] Sports Mode 2:

[0126]

[0127]

[0128] In order to achieve smooth conversion between different motion states in the switching of different motion modes, the embodiment of the present invention introduces a transition state. When the bionic robot switches the motion mode, a transition state is inserted, that is, a period of time is given to respond to the switching process changes between the two modes, and time is exchanged for space, so as to eliminate the abnormal state generated in the switching process between the two states.

[0129] Specifically, since the phase relationship and coupling relationship between different control modules are different in different motion modes, the phase relationship and coupling relationship between different modules can be determined by adjusting the phase difference and coupling characteristic degree parameters when switching different motion states.

[0130] In an embodiment of the present invention, the transition state is represented by the following formula:

[0131]

[0132] in, is a transition state. is the phase relationship of the current motion mode, To select the phase relationship of the next motion mode to switch, c represents the time coefficient, Indicates the time it takes to switch the motion mode. It is a dynamic variable related to time. The maximum time it takes to switch between sport modes.

[0133] By switching the corresponding phase relationship matrix in different motion modes, precise control can be achieved. The larger the coefficient, the greater the weight, and the coupling relationship will be dominated by the one with a large coefficient weight value.

[0134] Based on the two situations of the inserted transition state and the non-inserted transition state, the present invention conducts experimental analysis on the control signals of each control module of the bionic butterfly robot. The experimental results are as follows: Figure 6 and 7 , Figure 6 This is a control signal diagram for switching the motion mode when adding a transition state in an embodiment of the present invention. Figure 7 FIG. 1 is a control signal diagram of switching motion modes when the transition state is not added in the embodiment of the present invention. Assume that the bionic robot switches from motion state 1 to motion state 2. Figure 6 In the process of switching to the second motion mode, a transition state is given, which exchanges time for space dimension, making the switching process relatively smooth. Figure 7 There is no transition state introduced in Figure 6 The whole process is much smoother, so it can be seen that the added transition state effectively improves the quality of the control signal.

[0135] The monomer neuron model proposed in the embodiment of the present invention designs a multi-neuron control module composed of multiple monomer neuron models and having polymorphic stable output, which can maintain stable output and continuously generate control signals that meet the actual motion effect. At the same time, a transition state is introduced to exchange time for space, thereby achieving the effect of smooth switching between different states.

[0136] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

[0137] The present invention also provides a robot bionic neural network control system, which can be found in Figure 8 , Figure 8 80 is a structural block diagram of a neural network model control system of a bionic robot in an embodiment of the present invention. The system includes: a system decomposition module 801, a model construction module 802 and a signal generation module 803.

[0138] The system decomposition module 801 is used to evaluate the bionic robot model to obtain multiple control modules.

[0139] The model building module 802 is used to build a single neuron model corresponding to each control module according to the coupling relationship and phase relationship between each control module.

[0140] The signal generating module 803 is used to generate control signals for each control module in different motion modes based on each monomer neuron model, so as to control the bionic robot to complete motion in different motion modes by using the control signals.

[0141] A robot bionic neural network control system proposed in an embodiment of the present invention obtains a multi-neuron model of the bionic robot by constructing a monomer neuron model corresponding to each control module according to the coupling relationship and phase relationship between the various control modules of the bionic robot model. The control signal generated by the monomer neuron model based on the coupling factors between the various control modules can make each control module output stably, thereby realizing effective control of the bionic robot.

[0142] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0143] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A robot bionic neural network control method, characterized in that: The method comprises: Evaluate the bionic robot model and obtain multiple control modules; According to the coupling relationship and phase relationship between various control modules, a monomer neuron model corresponding to each control module is constructed; Generate control signals for each of the control modules in different motion modes based on each of the monomer neuron models, so as to control the bionic robot to complete motion in different motion modes by using the control signals; The step of constructing a monomer neuron model corresponding to each control module according to the coupling relationship and phase relationship between the control modules specifically includes: Acquire a second control module, a coupling characteristic degree parameter between the oscillators of the first control module and the second control module, a phase difference between the first control module and the second control module, and respectively acquire other characteristic parameters of the first control module and the second control module, wherein the first control module is any one of all the control modules of the bionic robot, and the second control module is any other control module in the bionic robot that is coupled with the first control module; Based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module, constructing a monomer neuron model corresponding to each control module; Wherein, constructing a monomer neuron model corresponding to each control module based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module specifically includes: Calculate according to the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module and the linear link related to the frequency of the first control module; Obtaining a waveform integration function of the first control module based on the asymmetric phase factor dynamic parameter and the linear link; Constructing a monomer neuron model corresponding to each control module according to the waveform integration function and other characteristic parameters of the first control module; The method further comprises: When controlling the bionic robot to switch between the two motion modes, inserting a transition state between the two motion modes to control the bionic robot to execute the transition state; Wherein, the transition state is expressed as: in, is a transition state. is the phase relationship of the current motion mode, To select the phase relationship of the next motion mode to switch, c represents the time coefficient, Indicates the time it takes to switch the motion mode. It is a dynamic variable related to time. The maximum time it takes to switch between sport modes.

2. The method according to claim 1, characterized in that The other characteristic parameters at least include: motion parameters and frequency; Then, the calculation is performed according to the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module and the linear link related to the frequency of the first control module, specifically including: Calculating according to the coupling characteristic degree parameter, the phase difference, and the motion parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module; A linear link related to the frequency of the first control module is obtained by performing calculation according to the frequency of the first control module and a preset frequency coefficient.

3. The method according to claim 2, characterized in that The motion parameters include at least: amplitude, instantaneous phase and limited amplitude; Then the calculation is performed according to the coupling characteristic degree parameter, the phase difference, and the motion parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module, specifically including: Performing derivative calculation according to the coupling characteristic degree parameter, the phase difference, the frequency and instantaneous phase of the first control module, and the amplitude and instantaneous phase of the second control module to obtain the instantaneous phase derivative of the first control module; Performing derivative calculation according to the limited amplitude of the first control module and the amplitude of the first control module to obtain the amplitude derivative of the first control module; The asymmetric phase factor dynamic parameter is calculated according to the amplitude derivative of the first control module and the instantaneous phase derivative of the first control module.

4. The method according to claim 3, characterized in that The step of constructing a monomer neuron model corresponding to each control module according to the waveform integration function and other characteristic parameters of the first control module specifically includes: Acquire a signal actual scaling operation factor corresponding to the first control module in a preset signal actual scaling operation factor table, wherein the signal actual scaling operation factor table includes a correspondence between the first control module and the signal actual scaling operation factor; A monomer neuron model corresponding to each control module is constructed according to the actual signal scaling operation factor corresponding to the first control module, the waveform integration function and the characteristic parameters of the first control module.

5. The method according to claim 4, characterized in that The monomer neuron model is expressed as: In the formula, i is a first control module, j is the second control module, is the instantaneous phase derivative of the first control module, v i is the frequency of the first control module, r j is the amplitude of the second control module, w ij is the coupling characteristic degree parameter, θ j is the instantaneous phase of the second control module, θ i is the instantaneous phase of the first control module, is the phase difference between the first control module and the second control module, is the amplitude derivative of the first control module, a is the convergence speed coefficient, R i is the limiting amplitude of the first control module, r i is the amplitude of the first control module, d i is the asymmetric phase factor dynamic parameter, f ( d i ) waveform integration function, k 0 is the waveform adjustment factor, y i is a linear link related to the frequency of the first control module, x i is the control signal of the first control module, C i is the actual scaling factor of the signal, K is the proportionality coefficient.

6. A robot bionic neural network control system, characterized in that: The system comprises: a system decomposition module, a model construction module and a signal generation module; The system decomposition module is used to evaluate the bionic robot model to obtain multiple control modules; The model building module is used to build a single neuron model corresponding to each control module according to the coupling relationship and phase relationship between the control modules; The signal generating module is used to generate control signals of each control module in different motion modes based on each monomer neuron model, so as to control the bionic robot to complete motion in different motion modes by using the control signals; The model building module is further used to obtain the second control module, and the coupling characteristic degree parameter between the oscillators of the first control module and the second control module, the phase difference between the first control module and the second control module, and respectively obtain other characteristic parameters of the first control module and the second control module, wherein the first control module is any one of all the control modules of the bionic robot, and the second control module is any other control module in the bionic robot that has a coupling relationship with the first control module; Based on the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module, constructing a monomer neuron model corresponding to each control module; The model building module is further used to calculate according to the coupling characteristic degree parameter, the phase difference, and other characteristic parameters of the first control module and the second control module to obtain the asymmetric phase factor dynamic parameter of the first control module and the linear link related to the frequency of the first control module; Obtaining a waveform integration function of the first control module based on the asymmetric phase factor dynamic parameter and the linear link; Constructing a monomer neuron model corresponding to each control module according to the waveform integration function and other characteristic parameters of the first control module; The system is also used to: When controlling the bionic robot to switch between the two motion modes, inserting a transition state between the two motion modes to control the bionic robot to execute the transition state; Wherein, the transition state is expressed as: in, is a transition state. is the phase relationship of the current motion mode, To select the phase relationship of the next motion mode to switch, c represents the time coefficient, Indicates the time it takes to switch the motion mode. It is a dynamic variable related to time. The maximum time it takes to switch between sport modes.

Images