CN109324510A - A kind of building of quadruped robot CPG control network, parameter tuning method - Google Patents
A kind of building of quadruped robot CPG control network, parameter tuning method Download PDFInfo
- Publication number
- CN109324510A CN109324510A CN201811093718.4A CN201811093718A CN109324510A CN 109324510 A CN109324510 A CN 109324510A CN 201811093718 A CN201811093718 A CN 201811093718A CN 109324510 A CN109324510 A CN 109324510A
- Authority
- CN
- China
- Prior art keywords
- cpg
- network
- neuron
- control
- quadruped robot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 210000002569 neuron Anatomy 0.000 claims abstract description 57
- 210000004394 hip joint Anatomy 0.000 claims abstract description 27
- 238000005457 optimization Methods 0.000 claims abstract description 17
- 230000000694 effects Effects 0.000 claims description 22
- 230000003044 adaptive effect Effects 0.000 claims description 16
- 230000005284 excitation Effects 0.000 claims description 13
- 230000005021 gait Effects 0.000 claims description 13
- 210000002161 motor neuron Anatomy 0.000 claims description 12
- 230000005764 inhibitory process Effects 0.000 claims description 10
- 210000005036 nerve Anatomy 0.000 claims description 9
- 230000001537 neural effect Effects 0.000 claims description 9
- 238000012512 characterization method Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 7
- 230000002401 inhibitory effect Effects 0.000 claims description 7
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000005070 sampling Methods 0.000 claims description 7
- 230000004308 accommodation Effects 0.000 claims description 6
- 230000006978 adaptation Effects 0.000 claims description 6
- 230000010355 oscillation Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 230000006698 induction Effects 0.000 claims description 5
- 241000208340 Araliaceae Species 0.000 claims description 4
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims description 4
- 235000003140 Panax quinquefolius Nutrition 0.000 claims description 4
- 235000008434 ginseng Nutrition 0.000 claims description 4
- 238000004088 simulation Methods 0.000 claims description 3
- 239000000203 mixture Substances 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 9
- 230000033001 locomotion Effects 0.000 description 7
- 239000002245 particle Substances 0.000 description 7
- 210000002414 leg Anatomy 0.000 description 6
- 238000013461 design Methods 0.000 description 5
- 210000000629 knee joint Anatomy 0.000 description 4
- 230000001020 rhythmical effect Effects 0.000 description 4
- 241001465754 Metazoa Species 0.000 description 2
- 241000251539 Vertebrata <Metazoa> Species 0.000 description 2
- 240000000233 Melia azedarach Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000012636 effector Substances 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 210000003414 extremity Anatomy 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 210000003194 forelimb Anatomy 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000003585 interneuronal effect Effects 0.000 description 1
- 210000001503 joint Anatomy 0.000 description 1
- 210000003141 lower extremity Anatomy 0.000 description 1
- 210000000056 organ Anatomy 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000011514 reflex Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000033764 rhythmic process Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 210000000278 spinal cord Anatomy 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 210000002435 tendon Anatomy 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
- Manipulator (AREA)
Abstract
The invention discloses building, parameter tuning and the optimization methods of a kind of quadruped robot CPG control network, the following steps are included: CPG control network is divided into four oscillating units for being used to control hip joint, each oscillating unit is using two neuron mutually inhibited compositions.The present invention realizes the symmetry and the control of rotational phase relationship of the miniature quadruped robot signal of hip joint control in the process of walking;By CPG control in original differential equation group be reduced to DIFFERENCE EQUATIONS, CPG control network intrinsic parameter is optimized, the complexity for reducing CPG control network, improves parameter tuning speed, constructs output signal amplitude frequency controllably and the CPG network of fast and stable.
Description
Technical field
The invention belongs to control network reference services design method more particularly to a kind of miniature quadruped robot CPG control
The parameter tuning method of network.
Background technique
CPG (Central Pattern Generator, central pattern generator (cpg)) is to generate animal rhythmic movement behavior
Biological neural loop, it is made of a series of neural oscillators, is that neural oscillator and multipath reflection circuit system are integrated in one
Play a complicated Distributed Artificial Neural Network of composition.
CPG is the local oscillation network being made of intrerneuron, is realized by interneuronal mutual inhibition from exciting
It swings, generates the multichannel periodic signal with stable phase angle interlocked relationship.The rhythmic movement of animal is the self-excitation of lower nervous center
Behavior is controlled by CPG, controls the rhythmic movement of limbs or trunk region of interest, it is located in the spinal cord of vertebrate.It is non-
In the neuromere of vertebrate, effector organ (such as leg, tendon forceps etc.) are directly controlled by the signal that CPG is exported, from higher nerve
The motion command that pivot passes to CPG only serves the effect of initialization, does not generate motor pattern directly.It is generated by CPG specific
Even if the order of the not no higher nervous center of motor pattern, as long as there is the excitation of external environment, CPG also can produce movement mould
Formula, equally, CPG can also work in the case where no external environment feedback signal.In general, the feedback signal root of external environment
Changing the motor pattern of CPG according to ambient conditions to obtain better adaptive capacity to environment, i.e. CPG does not need reflex activity participation,
Can also there is no external sense information input, mutually to inhibit circuit to coordinate the rhythmic movements such as walking, in musculus flexor and extensor
Rhythmicity alternating activity occurs between control neuron.
In (Lian Xiaofeng, Guo Weiping, Lian little Qin, Su Zhong, Zhao Xu " the snakelike machine based on genetic algorithm of non-patent literature 1
People CPG Model Parameter Optimization " [J] computer engineering and design, 2015,36 (07): 1859-1864.) in, author is for snakelike
The cyclic inhibition CPG model that robot uses solves the problems, such as parameter tuning low efficiency in control, proposes a kind of utilize and loses
Propagation algorithm realizes the rhythm and pace of moving things output of CPG network to the method for CPG Controlling model parameter optimization.
But the amplitude and frequency controllability for not embodying output waveform after its parameter tuning in the document, only realize oscillation
The output of waveform, and it is only applicable to snake-shaped robot, do not ensure that the CPG model of adjusting is suitable for other occasions.
Summary of the invention
The purpose of the present invention is: low efficiency, unstable disadvantage are adjusted in order to overcome in above-mentioned CPG network parameter adjusting,
And provide a kind of complexity is low, adjusting is high-efficient and amplitude frequency is controllable miniature quadruped robot CPG control network and its ginseng
Number setting method.
The purpose of the present invention can be achieved through the following technical solutions.
According to the first aspect of the invention, the present invention provides a kind of quadruped robot CPG control network structure buildings
Method comprising following steps:
1) CPG control network is divided into four oscillating units for being used to control the hip joint of the quadruped robot, it is each to shake
The neuron that unit is mutually inhibited by two is swung to form, by the control network of hip joint by being mapped to kneed control network,
Realize the rotational phase relationship control of the quadruped robot signal of hip joint control in the process of walking;
2) the corresponding differential equation group of neuron mode used by CPG network is rewritten, forms CPG and controls net
The DIFFERENCE EQUATIONS of network;The mathematic(al) representation of the DIFFERENCE EQUATIONS are as follows:
Wherein, e simulates extensor (extensor) motor neuron, and f simulates musculus flexor (flexor) motor neuron;U is mind
Through first internal state;V is neuron adaptive state;Y { e, f } and y is respectively neuron and oscillating unit output signal, as
Joint control amount;Tr is rise time coefficient, characterizes rise time of the single neuron under Stepped Impedance Resonators;Ta is adaptation time
Coefficient, characterization adapt to the lag time of effect;α is neuron accommodation coefficient, characterizes adaptive state to inside neurons state
Influence degree;β is the mutual inhibition coefficient between oscillating unit intrinsic nerve member;wijIt is oscillating unit j to the connection weight of i, positive number table
Show excitation, negative number representation inhibiting effect, between oscillating unit together by the gait matrix effect of intercoupling;S indicates mould
Quasi- high-rise neural excitation inputs, and all neurons inputs are identical in CPG network in the case of normal gait;G (u) is threshold
Value function, threshold value C are set according to demand;Tc is the sampling period.
Preferably, the neuron adaptive state v includes fatigue or inhibits certainly.
Preferably, the threshold value C is 0.
According to the second aspect of the invention, the present invention provides a kind of ginsengs of miniature quadruped robot CPG control network
The structure of the method for number adjusting, the quadruped robot CPG control network uses the structure according to any one of above technical scheme
Construction method building, which is characterized in that the method for the parameter tuning the following steps are included:
Step S11 is adjusted larger to the amplitude and frequency influence of CPG network curve of output using the method for control variable
Parameter;
Step S12 to the effect tendency of joint control signal and assists PSO optimized calculation method according to each parameter, obtains
Taking can make the amplitude of curve of output minimum at a distance from frequency and target amplitude frequency;
Step S13, according to actual curve of output Induction Peried, to its under optimal joint control signal in step S12
Yu Wei adjusts model parameter and carries out timing adjustment, makes output signal fast and stable.
According to the third aspect of the present invention, the present invention provides a kind of parameter of quadruped robot CPG control network is excellent
Change method, the parameter optimization method include:
CPG control network is divided into four oscillating units for being used to control hip joint by step S1, and each oscillating unit uses two
A neuron composition mutually inhibited, by the control network of hip joint by being mapped to kneed control network;
Step S2 rewrites the corresponding differential equation group of neuron mode used by CPG network, and carries out letter
Change, forms the DIFFERENCE EQUATIONS of CPG control network;
Step S3 carries out parameter tuning to the amplitude and frequency of CPG control each oscillating unit curve of output of network, constructs
Optimal network topological structure.
Preferably, the mathematic(al) representation of the DIFFERENCE EQUATIONS are as follows:
Wherein, e simulates extensor (extensor) motor neuron, and f simulates musculus flexor (flexor) motor neuron;U is mind
Through first internal state;V is neuron adaptive state;Y { e, f } and y is respectively neuron and oscillating unit output signal, as
Joint control amount;Tr is rise time coefficient, characterizes rise time of the single neuron under Stepped Impedance Resonators;Ta is adaptation time
Coefficient, characterization adapt to the lag time of effect;α is neuron accommodation coefficient, characterizes adaptive state to inside neurons state
Influence degree;β is the mutual inhibition coefficient between oscillating unit intrinsic nerve member;wijIt is oscillating unit j to the connection weight of i, positive number table
Show excitation, negative number representation inhibiting effect, between oscillating unit together by the gait matrix effect of intercoupling;S indicates mould
Quasi- high-rise neural excitation inputs, and all neurons inputs are identical in CPG network in the case of normal gait;G (u) is threshold
Value function, threshold value C are set according to demand;Tc is the sampling period.
Preferably, the step S1 and step S2 further comprises excellent to oscillating unit intrinsic nerve metamathematics expression-form
The step of change: solution form is rewritten into difference equation.
Preferably, the step S3 is further included steps of
Step S31 is adjusted larger to the amplitude and frequency influence of CPG network curve of output using the method for control variable
Parameter;
Step S32 to the effect tendency of joint control signal and assists PSO optimized calculation method according to each parameter, obtains
Taking can make the amplitude of curve of output minimum at a distance from frequency and target amplitude frequency;
Step S33, according to actual curve of output Induction Peried, to its under optimal joint control signal in step S12
Yu Wei adjusts model parameter and carries out timing adjustment, makes output signal fast and stable.
Preferably, it is adjusted in the step S31 to the amplitude of CPG network curve of output and the biggish parameter of frequency influence
Mathematical expression form are as follows:
Fitness=(x-x0) ^2+ (y-y0) ^2.
The beneficial effects of the present invention are:
Compared with prior art, the present invention is based on the control methods of the quadruped robot of CPG, propose a kind of new CPG
Network reference services method is controlled, more reasonable effective model parameter adjusting strategy is devised, reduces CPG control network
The complexity of parameter tuning keeps the output amplitude of CPG network and frequency controllable.
Detailed description of the invention
Fig. 1 is the method integral frame figure of CPG control parameter adjusting;
Fig. 2 is single leg joint freedom degree distribution map of quadruped robot;
Fig. 3 is the CPG network structure for controlling the main control network of hip joint;
Be of coupled connections mode figure of the Fig. 4 between musculus flexor neuron and extensor neuron;
Fig. 5 is the flow chart of PSO algorithm optimization CPG parameter;
Fig. 6 is the oscillator signal figure for controlling the CPG control network output of hip joint.
Specific embodiment
A kind of embodiment 1: quadruped robot CPG control network structure construction method
A kind of quadruped robot CPG according to the present invention controls network structure construction method, comprising the following steps:
1) CPG control network is divided into four oscillating units for being used to control the hip joint of the quadruped robot, it is each to shake
The neuron that unit is mutually inhibited by two is swung to form, by the control network of hip joint by being mapped to kneed control network,
Realize the rotational phase relationship control of the quadruped robot signal of hip joint control in the process of walking;
2) the corresponding differential equation group of neuron mode used by CPG network is rewritten, forms CPG and controls net
The DIFFERENCE EQUATIONS of network;The mathematic(al) representation of the DIFFERENCE EQUATIONS are as follows:
Wherein, e simulates extensor (extensor) motor neuron, and f simulates musculus flexor (flexor) motor neuron;U is mind
Through first internal state;V is neuron adaptive state;Y { e, f } and y is respectively neuron and oscillating unit output signal, as
Joint control amount;Tr is rise time coefficient, characterizes rise time of the single neuron under Stepped Impedance Resonators;Ta is adaptation time
Coefficient, characterization adapt to the lag time of effect;α is neuron accommodation coefficient, characterizes adaptive state to inside neurons state
Influence degree;β is the mutual inhibition coefficient between oscillating unit intrinsic nerve member;wijIt is oscillating unit j to the connection weight of i, positive number table
Show excitation, negative number representation inhibiting effect, between oscillating unit together by the gait matrix effect of intercoupling;S indicates mould
Quasi- high-rise neural excitation inputs, and all neurons inputs are identical in CPG network in the case of normal gait;G (u) is threshold
Value function, threshold value C are set according to demand;Tc is the sampling period.
In a preferred embodiment, the neuron adaptive state v includes fatigue or inhibits certainly.
In a preferred embodiment, the threshold value C is 0.
A kind of embodiment 2: miniature quadruped robot CPG control network reference services method
A kind of miniature quadruped robot CPG control network reference services method, comprising the following steps:
Step 1: being the freedom degree point of the four-footed anthropomorphic robot list leg of the most common research and application at present shown in Fig. 2
Cloth schematic diagram.The control of hip joint and knee joint phase is critically important to walking, but the movement of the two joint freedom degrees when walking
Process can regard an oscillator signal as, and the frequency of the two motor message can be regarded as unanimously substantially.Therefore the design is by four
A oscillating unit CPG control network output is used to control kneed pair for controlling the main control network portion of hip joint
The output of network portion is controlled by the output signal of hip joint, realizes that the symmetry of quadruped robot control signal and robot exist
Phase relation in walking process.
As shown in figure 3, connecting and composing the control network of hip joint by the bidirectional couple of four neuron elements to control
Four freedom degrees of quadruped robot hip joint.By hip joint control network portion four neuron elements between using etc. power
The mutual inhibition connection (weight is inhibited to be set as -1) of weight, the phase difference of forelimb or so leg and hind leg or so leg is 180 °, diagonal leg
Motor message phase it is consistent, meet left and right hip joint front-rear direction freedom degree control signal inverted relationship.For knee joint
Control, the transformation of translation and amplitude of the curve of output that we are obtained by hip joint network Jing Guo time domain can be obtained by its control
Koji-making line.T is the time shaft sequence of hip joint curve of output, and Y is the amplitude of hip joint curve of output, and T1 represents knee joint output
The time shaft sequence of curve, Y1 are the amplitude of knee joint curve of output.A represents the proportionality coefficient of amplitude, and t0 is the delay of time domain
Time.Its mathematical expression form are as follows:
Specific optimization method in step 2) are as follows: CPG oscillating unit is as shown in figure 4, the CPG of Matsuoka model is vibrated
Original differential equation group operation, is rewritten using calculus of differences, is reduced to DIFFERENCE EQUATIONS used by unit, is accelerated into one
The CPG parameter optimization of step.Sliding-model control is carried out to the CPG oscillating unit equation based on Matsuoka model, is obtained following
CPG nervous centralis discrete equation.In actual operation, accurate for this spline equation is solved in the past using quadravalence Long Ge-
Ku Tafa is solved, but we are used in parameter tuning, and extremely accurate evaluation is not needed, thus using discrete equation come
Carry out the emulation of oscillating curve.But it is 0.01s for the sampling period of discrete point, had not only guaranteed its precision, but also accelerate adjusting
Speed.
Wherein, e simulates extensor (extensor) motor neuron, and f simulates musculus flexor (flexor) motor neuron;U is mind
Through first internal state;V is neuron adaptive (fatigue or inhibiting certainly) state;Y { e, f } and y is respectively that neuron and oscillation are single
First output signal can be used as joint control amount;Tr is rise time coefficient, characterizes rising of the single neuron under Stepped Impedance Resonators
Time;Ta is adaptation time coefficient, and characterization adapts to the lag time of effect;α is neuron accommodation coefficient, characterizes adaptive state pair
The influence degree β of inside neurons state is the mutual inhibition coefficient between oscillating unit intrinsic nerve member;Wij is oscillating unit j to i's
Connection weight, positive number indicate excitation, and negative number representation inhibiting effect passes through the gait matrix effect of intercoupling between oscillating unit
Together;S indicates the high-rise neural excitation input of simulation, all neuron inputs in CPG network in the case of normal gait
It is identical;G (u) is threshold function table, and threshold value C is set according to demand, is usually chosen to 0.Tc is the week for resolving first difference equation
Phase, i.e. sampling period.
Specific optimization method in step 3) are as follows:
First using control variable method, first adjust on the amplitude of CPG network curve of output and frequency variation influence compared with
Big parameter.It is first analyzed with the result of trial and error procedure and single-parameter analysis method, finds to control curve of output to whole CPG
Amplitude and frequency influence it is bigger be five variables of Tr, Ta, a, b and s.And initial value Ue, Ve, Uf and Vf are to the defeated of entirety
The amplitude and frequency resultant of curve influence less out, but have large effect to the time of starting of oscillation.Gait matrix is to whole width
Value has large effect, but the influence to the period is little.It follows that the adjusting of CPG univers parameter is one extremely cumbersome
Process, most parameter have an impact to the amplitude of curve of output with frequency.
PSO optimized calculation method to the effect tendency of joint control signal and is assisted according to each parameter later, acquisition can
So that the amplitude of curve of output is minimum at a distance from frequency and target amplitude frequency, specific algorithm structure is as shown in Figure 5.PSO
Algorithm is a good optimization algorithm for the parameter tuning of CPG, it not only needs to adjust without many parameters, but also parallel
Algorithm let us entirety realization it is simple and quick.
PSO is initialized as a group random particles (RANDOM SOLUTION).Then optimal solution is found by iterating.It is iterated each time
In, particle updates oneself by tracking two " extreme value ".First is exactly optimal solution that particle itself is found.This solution
Doing individual another extreme value of extreme value pBest. is the optimal solution that entire population is found at present.This extreme value is global extremum gBest.
In addition it can also not have to entire population and be the neighbours with a portion particle the most, then the extreme value in all neighbours
It is exactly local extremum.When finding the two optimal values, particle updated according to following formula oneself speed and new position
It sets.
V []=w × v []+c1×rand()×(pbest[]-ppresent[])+c2×rand()×(gbest[]-ppresent
[])
ppresent[]=ppresent[]+v[]
V [] is the speed of particle, and persent [] is the position of current particle.Pbest [] and gbest [] as previously defined,
Rand () is the random number between (0,1).C1, c2 are Studying factors, and w is inertia weight.These parameters in PSO are big
More reliable empirical settings influence it in specific programming smaller although CPG parameter tuning has its particularity.
The output signal of CPG control needs frequency and amplitude to reach more excellent, is related to the parameter tuning of multiple target.So
It is real that present invention employs the amplitude x of reality output curve and frequency y with L squares of the distance between target amplitude x0 and frequency y0
The fitness function on border realizes the Optimization Solution simultaneously to two targets of amplitude and frequency.Specific mathematical expression form are as follows:
Fitness=(x-x0) ^2+ (y-y0) ^2
Secondly not whole to remaining under optimal joint control signal in back according to actual curve of output Induction Peried
Rational method carries out timing adjustment.Due to being adjusted to five parametric variables, having ignored initial value to it at adjusting initial stage
The influence of CPG output signal, and cause output signal can not fast start-up, reach desired frequency and amplitude.For initial value
It chooses, the present invention recycles its stabilization after five parameters of b, s determine using the Tr for optimizing back, Ta, a
Ue after output, uf, the value of ve, vf substitute into CPG control, and as the initial value of CPG control, stable output can be obtained
And the output signal of quick oscillation, output signal are shown in Fig. 6.
The parameter tuning method sufficiently combines intelligent algorithm, and designer can be exported according to the target of design using excellent
Change algorithm and obtain the comprehensive output after parameter tuning as a result, deepening the understanding to controller and object parameters, and obtains new
Design philosophy and inspiration.
A kind of parameter tuning method of quadruped robot CPG control network of the present invention, proposes a kind of new CPG control
Network parameter setting method, and more reasonable effective model parameter adjusting strategy is devised, biography is improved to a certain extent
The CPG of system controls network trial and error procedure setting parameter, so that CPG control complexity reduces, control network parameter selection is more reasonable.
Claims (9)
1. a kind of quadruped robot CPG controls network structure construction method, which comprises the following steps:
1) CPG control network is divided into four oscillating units for being used to control the hip joint of the quadruped robot, each oscillation is single
The neuron that member is mutually inhibited by two forms, and by the control network of hip joint by being mapped to kneed control network, realizes
The rotational phase relationship control of the quadruped robot signal of hip joint control in the process of walking;
2) the corresponding differential equation group of neuron mode used by CPG network is rewritten, forms the difference of CPG control network
Divide equation group;The mathematic(al) representation of the DIFFERENCE EQUATIONS are as follows:
Wherein, e simulates extensor (extensor) motor neuron, and f simulates musculus flexor (flexor) motor neuron;U is neuron
Internal state;V is neuron adaptive state;Y { e, f } and y is respectively neuron and oscillating unit output signal, as joint
Control amount;Tr is rise time coefficient, characterizes rise time of the single neuron under Stepped Impedance Resonators;Ta is adaptation time system
Number, characterization adapt to the lag time of effect;α is neuron accommodation coefficient, shadow of the characterization adaptive state to inside neurons state
The degree of sound;β is the mutual inhibition coefficient between oscillating unit intrinsic nerve member;wijIt is oscillating unit j to the connection weight of i, positive number indicates
Excitation, negative number representation inhibiting effect, between oscillating unit together by the gait matrix effect of intercoupling;S indicates simulation
High-rise neural excitation inputs, and all neurons inputs are identical in CPG network in the case of normal gait;G (u) is threshold value letter
Number, threshold value C are set according to demand;Tc is the sampling period.
2. a kind of quadruped robot CPG according to claim 1 controls network structure construction method, which is characterized in that institute
Neuron adaptive state v is stated to include fatigue or inhibit certainly.
3. a kind of quadruped robot CPG according to claim 1 controls network structure construction method, which is characterized in that institute
Stating threshold value C is 0.
4. a kind of method of the parameter tuning of miniature quadruped robot CPG control network, the quadruped robot CPG control network
Structure using construction method as claimed in one of claims 1-3 construct, which is characterized in that the method for the parameter tuning
The following steps are included:
Step S11 adjusts amplitude and the biggish ginseng of frequency influence to CPG network curve of output using the method for control variable
Number;
Step S12 to the effect tendency of joint control signal and assists PSO optimized calculation method according to each parameter, and acquisition can
So that the amplitude of curve of output is minimum at a distance from frequency and target amplitude frequency;
Step S13, according to actual curve of output Induction Peried, not to remaining under optimal joint control signal in step S12
It adjusts model parameter and carries out timing adjustment, make output signal fast and stable.
5. a kind of parameter optimization method of quadruped robot CPG control network, the parameter optimization method include:
CPG control network is divided into four oscillating units for being used to control hip joint by step S1, and each oscillating unit is mutual using two
The neuron of inhibition forms, by the control network of hip joint by being mapped to kneed control network;
Step S2 rewrites the corresponding differential equation group of neuron mode used by CPG network, and is simplified, shape
At the DIFFERENCE EQUATIONS of CPG control network;
Step S3 carries out parameter tuning to the amplitude and frequency of CPG control each oscillating unit curve of output of network, constructs optimal
Network topology structure.
6. a kind of parameter optimization method of miniature quadruped robot CPG control network according to claim 5, feature exist
In the mathematic(al) representation of the DIFFERENCE EQUATIONS are as follows:
Wherein, e simulates extensor (extensor) motor neuron, and f simulates musculus flexor (flexor) motor neuron;U is neuron
Internal state;V is neuron adaptive state;Y { e, f } and y is respectively neuron and oscillating unit output signal, as joint
Control amount;Tr is rise time coefficient, characterizes rise time of the single neuron under Stepped Impedance Resonators;Ta is adaptation time system
Number, characterization adapt to the lag time of effect;α is neuron accommodation coefficient, shadow of the characterization adaptive state to inside neurons state
The degree of sound;β is the mutual inhibition coefficient between oscillating unit intrinsic nerve member;wijIt is oscillating unit j to the connection weight of i, positive number indicates
Excitation, negative number representation inhibiting effect, between oscillating unit together by the gait matrix effect of intercoupling;S indicates simulation
High-rise neural excitation inputs, and all neurons inputs are identical in CPG network in the case of normal gait;G (u) is threshold value letter
Number, threshold value C are set according to demand;Tc is the sampling period.
7. a kind of parameter optimization method of miniature quadruped robot CPG control network according to claim 5, feature exist
In the step S1 and step S2 further comprise the step of optimization to oscillating unit intrinsic nerve metamathematics expression-form: will
Solution form is rewritten into difference equation.
8. a kind of parameter optimization side of miniature quadruped robot CPG control network according to any one of claim 5-7
Method, which is characterized in that the step S3 is further included steps of
Step S31 adjusts amplitude and the biggish ginseng of frequency influence to CPG network curve of output using the method for control variable
Number;
Step S32 to the effect tendency of joint control signal and assists PSO optimized calculation method according to each parameter, and acquisition can
So that the amplitude of curve of output is minimum at a distance from frequency and target amplitude frequency;
Step S33, according to actual curve of output Induction Peried, not to remaining under optimal joint control signal in step S12
It adjusts model parameter and carries out timing adjustment, make output signal fast and stable.
9. a kind of parameter optimization method of miniature quadruped robot CPG control network according to claim 8, feature exist
In adjusting is to the amplitude of CPG network curve of output and the mathematical expression form of the biggish parameter of frequency influence in the step S31
Are as follows:
Fitness=(x-x0) ^2+ (y-y0) ^2.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811093718.4A CN109324510A (en) | 2018-09-19 | 2018-09-19 | A kind of building of quadruped robot CPG control network, parameter tuning method |
PCT/CN2018/115481 WO2020056895A1 (en) | 2018-09-19 | 2018-11-14 | Method for constructing cpg control network of quadruped robot, and methods for tuning and optimizing parameter of cpg control network of quadruped robot |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811093718.4A CN109324510A (en) | 2018-09-19 | 2018-09-19 | A kind of building of quadruped robot CPG control network, parameter tuning method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109324510A true CN109324510A (en) | 2019-02-12 |
Family
ID=65265904
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811093718.4A Pending CN109324510A (en) | 2018-09-19 | 2018-09-19 | A kind of building of quadruped robot CPG control network, parameter tuning method |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN109324510A (en) |
WO (1) | WO2020056895A1 (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111208822A (en) * | 2020-02-17 | 2020-05-29 | 清华大学深圳国际研究生院 | Quadruped robot gait control method based on reinforcement learning and CPG controller |
CN111438694A (en) * | 2020-05-21 | 2020-07-24 | 中国计量大学 | Biped robot diagonal gait planning method based on double generation functions |
CN117930663A (en) * | 2024-03-20 | 2024-04-26 | 浙江大学 | Motion control system of four-foot robot based on eight-element neural network |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1490134A (en) * | 2003-09-19 | 2004-04-21 | 清华大学 | Method and device for controlling robot simulating animal foot movement |
CN102759923A (en) * | 2012-04-13 | 2012-10-31 | 中国科学院合肥物质科学研究院 | Control method for bionic dual-feet robot walking on water |
CN103092197A (en) * | 2011-10-28 | 2013-05-08 | 同济大学 | Four-foot robot working space track generating method based on certified program generator (CPG) mechanism |
CN107088307A (en) * | 2017-07-03 | 2017-08-25 | 山东建筑大学 | A kind of bionic machine fish and its Optimization about control parameter method |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5224506B2 (en) * | 2008-03-04 | 2013-07-03 | 国立大学法人九州工業大学 | Control system, vibration control device, and control signal generation method |
CN102637036A (en) * | 2012-05-08 | 2012-08-15 | 北京理工大学 | Combined type bionic quadruped robot controller |
CN103203746B (en) * | 2012-09-29 | 2015-10-28 | 同济大学 | Biped robot CPG net control topological structure construction method |
-
2018
- 2018-09-19 CN CN201811093718.4A patent/CN109324510A/en active Pending
- 2018-11-14 WO PCT/CN2018/115481 patent/WO2020056895A1/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1490134A (en) * | 2003-09-19 | 2004-04-21 | 清华大学 | Method and device for controlling robot simulating animal foot movement |
CN103092197A (en) * | 2011-10-28 | 2013-05-08 | 同济大学 | Four-foot robot working space track generating method based on certified program generator (CPG) mechanism |
CN102759923A (en) * | 2012-04-13 | 2012-10-31 | 中国科学院合肥物质科学研究院 | Control method for bionic dual-feet robot walking on water |
CN107088307A (en) * | 2017-07-03 | 2017-08-25 | 山东建筑大学 | A kind of bionic machine fish and its Optimization about control parameter method |
Non-Patent Citations (5)
Title |
---|
ALEXANDER SPRÖWITZ 等: "Towards Dynamic Trot Gait Locomotion: Design, Control, and Experiments with Cheetah-cub, a Compliant Quadruped Robot", 《THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH》 * |
KIYOTOSHI MATSUOKA: "Sustained Oscillations Generated by Mutually Inhibiting Neurons with Adaptation", 《BIOLOGICAL CYBERNETICS》 * |
Z.G. ZHANG 等: "Adaptive Running of a Quadruped Robot on Irregular Terrain based on Biological Concepts", 《LNTERNATIONAL CONFERENCE ON ROBOTICS & AUTOMATION》 * |
张秀丽: "四足机器人节律运动及环境适应性的生物控制研究", 《中国优秀博硕士学位论文全文数据库(博士) 信息科技辑》 * |
李鑫: "中枢模式发生器在六足机器人运动控制中的应用", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111208822A (en) * | 2020-02-17 | 2020-05-29 | 清华大学深圳国际研究生院 | Quadruped robot gait control method based on reinforcement learning and CPG controller |
CN111438694A (en) * | 2020-05-21 | 2020-07-24 | 中国计量大学 | Biped robot diagonal gait planning method based on double generation functions |
CN117930663A (en) * | 2024-03-20 | 2024-04-26 | 浙江大学 | Motion control system of four-foot robot based on eight-element neural network |
CN117930663B (en) * | 2024-03-20 | 2024-06-04 | 浙江大学 | Motion control system of four-foot robot based on eight-element neural network |
Also Published As
Publication number | Publication date |
---|---|
WO2020056895A1 (en) | 2020-03-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109324510A (en) | A kind of building of quadruped robot CPG control network, parameter tuning method | |
Ijspeert | A connectionist central pattern generator for the aquatic and terrestrial gaits of a simulated salamander | |
JP6227683B2 (en) | A way to develop and control a robot to get a movement that matches an animated character. | |
US20040148268A1 (en) | Artificial multiped and motion controller therefor | |
Wadden et al. | A neuro-mechanical model of legged locomotion: single leg control | |
Gao et al. | Reinforcement learning control of robotic knee with human-in-the-loop by flexible policy iteration | |
Filho et al. | Modeling of a bipedal robot using mutually coupled Rayleigh oscillators | |
CN105182754B (en) | The construction method of bio-robot motion control neutral net | |
CN103203746A (en) | Method for constructing central pattern generator (CPG) control network topology structure of biped robot | |
Previdi et al. | Closed-loop control of FES supported standing up and sitting down using Virtual Reference Feedback Tuning | |
Inada et al. | Behavior generation of bipedal robot using central pattern generator (CPG)(1st report: CPG parameters searching method by genetic algorithm) | |
Cruz Ruiz et al. | Muscle‐based control for character animation | |
Anand et al. | A deep reinforcement learning based approach towards generating human walking behavior with a neuromuscular model | |
Zhang et al. | Modeling biological motor control for human locomotion with functional electrical stimulation | |
CN106826830B (en) | A kind of robot control method and device | |
Hein et al. | Evolution of biped walking using neural oscillators and physical simulation | |
CN105589994A (en) | Topological optimization design method for porous material unit grid structure | |
CN115230162A (en) | Lightweight implementation method of human wrist external fixation brace based on topology optimization | |
Schmitz et al. | No need for a body model: positive velocity feedback for the control of an 18-DOF robot walker | |
Torres et al. | Automated generation of cpg-based locomotion for robot nao | |
CN107991868B (en) | Coordination control method for left and right pectoral fins of pharyngeal-jaw mode bionic system | |
Larpin et al. | Co-evolution of morphology and control of virtual legged robots for a steering task | |
Boeing et al. | Evolving Splines: An alternative locomotion controller for a bipedal robot | |
Oller et al. | A stress driven growth model for soft tissue considering biological availability | |
Arcolezi et al. | A RISE-based controller fine-tuned by an improved genetic algorithm for human lower limb rehabilitation via neuromuscular electrical stimulation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190212 |
|
RJ01 | Rejection of invention patent application after publication |