CN103203746A - Method for constructing central pattern generator (CPG) control network topology structure of biped robot - Google Patents

Abstract

The invention relates to a method for constructing central pattern generator (CPG) control network topology structure of biped robot. The method comprises the steps of separating a CPG control network into a body network portion for controlling hip joints and a leg network portion for controlling leg joints to achieve symmetry of control signals for left and right leg joints and control over reasonable phase relations of left and right legs; optimizing coupling connection modes among neuron units in the CPG control network to reduce the complexity of the CPG control network; and performing parameters setting on the CPG control network to construct an optimal network topology structure. Compared with the prior art, the method has the advantages of being low in complexity, reasonable in control network structure and the like.

Description

Biped robot CPG controls network topology building method

Technical field

The present invention relates to a kind of information processing method of network topology structure, especially relate to a kind of biped robot CPG control network topology building method.

Background technology

Walking control is a key technology in two foots and anthropomorphic robot research and the application.Traditional method is the artificial planning of adopting based on model, and robot is moved by predefined movement locus.Along with robot is applied to the destructuring environment gradually, hindered the practical application of robot based on the conventional method of robot model and walking environmental modeling.Along with people to the further investigation of biped walking essence and the continuous development of Neuscience, be applied to gradually based on the control method of Neuscience in the walking control of biped robot.Control thinking based on central pattern generator (cpg) (CPG, central pattern generator) is typical case's representative of this direction.

The biologist thinks, the motion control neutral net of animal is accepted from the neural regulation and control order of high level centered by CPG, and from the feedback information of somatic sensor.CPG is the local oscillation network that is made of neuron, can produce stable phase place interlocked relationship by the mutual inhibition between the neuron, and excite the body region of interest to produce rhythmic movement by self-oscillation.High level regulation and control and the environmental feedback of brain can play regulating action to the rhythmic movement of animal, make the motion of animal have adaptability.Have interpretation on the biology based on the control method of CPG, thereby caused interest widely in engineering circle recently, and begin CPG mechanism is carried out the engineering modeling, be applied in each robotlike's the motion control.Try hard to improve the exercise performance of robot by the neural control mechanism of biological motion is combined with the bionic moving mechanism of robot, promote the practicalization of robot in various actual environments.

Basic ideas based on CPG control are: at first CPG is carried out the engineering modeling, design the function that can produce the stable oscillation stationary vibration output signal, controller as the robot free degree, the control of a plurality of frees degree generally realizes with the CPG network that a plurality of CPG unit constitutes, change topology of networks and can change the output mode of oscillator signal, thereby realize different motor patterns.At present, for the system of this complexity of biped robot, the CPG model does not have very strong practicality, and present research also more rests on the analog simulation stage, perhaps only is that the rhythmic movement in some joint of biped robot is controlled.The application of CPG mechanism in robot motion's control mainly is to adopt joint space (joint space) control method, and the control framework of its integral body as shown in Figure 1.In control system, the CPG mixed-media network modules mixed-media is the nucleus module that produces the joint control signal, and the reasonability of its design of network topology structure and validity are related to the quality of controlling effect.When research and engineering application, generally the free degree of CPG unit according to robot distributed at present, be assigned to the joint space of robot one by one, utilize intercoupling between a plurality of CPG unit to produce the motor pattern of expectation.This control method is applied to the real experiment that biped robot does not also have success at present, difficult point just is that the CPG design of network topology structure does not also have unified method for designing, the complexity of network structure has increased the difficulty of parameter tuning, is difficult to produce the joint control signal of expectation.On the other hand, because the biped robot free degree is more, it is also more relatively to constitute the needed CPG number of unit of controller.How determining that rational topology connects and model parameter between the CPG unit, is the optimizing problem of a complexity.Generally need to adopt evolution algorithm that control system is optimized, but owing to relate to topological structure optimizing and Model Parameter Optimization simultaneously, adopt traditional optimization algorithm to be difficult to obtain a good result.Simultaneously, this method is optimized consuming time longer, realizes also having certain difficulty at the robot entity.Therefore each free degree of the CPG unit simply being distributed to robot that can not be simple, rely on evolutionary computation and carry out optimizing, must network topology structure framework reasonable in design be prerequisite, be aided with connected mode and model parameter that evolutionary computation is further optimized CPG again.

Summary of the invention

Purpose of the present invention is exactly to provide in order to overcome the defective that above-mentioned prior art exists that a kind of complexity is low, the rational biped robot CPG control of control network structure network topology building method.

Purpose of the present invention can be achieved through the following technical solutions:

A kind of biped robot CPG control network topology building method may further comprise the steps:

1) CPG is controlled network and be divided into for health control network (bodynetwork) part of control robot hip joint and be used for shank control network (leg network) part in control shank joint, be convenient to control the symmetry of biped robot left and right sides leg control signal and the robot phase relation of left and right sides leg in the process of walking;

2) coupled modes between the neuron elements in the CPG network are optimized, have simplified the connected mode between the neuron;

3) to the parameter tuning that carries out of CPG control network, construct the optimal network topological structure.

CPG control network using neuron oscillator model in the step 1), the mathematic(al) representation of this model is:

$T_r{\stackrel{\·}{u}}_i^{\{e,f\}}={-u}_i^{\{e,f\}}+w_{\mathrm{fe}}{r}_i^{\{f,e\}}-\mathrm{\β}{v}_i^{\{e,f\}}+s_0+{\mathrm{Feed}}_i^{\{e,f\}}+\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{r}_j^{\{e,f\}}$

$T_a{\stackrel{\·}{v}}_i^{\{e,f\}}=-v_i^{\{e,f\}}+r_i^{\{e,f\}}$

$r_i^{\{e,f\}}=\max (u_i^{\{e,f\}},0)$

$r_i={-u}_i^{\left\{e\right\}}+u_i^{\left\{f\right\}}$

Wherein, i represents i CPG unit, and e represents the musculus flexor neuron, and f represents extensor neuron, u _iBe neuronic internal state, v _iFor neuron from holddown,

Be neuronic output, T _rAnd T _aBe respectively rise time and adaptation time constant, w _FeBe neuronic mutual rejection coefficient, β is neuronic from rejection coefficient, s ₀Represent the periodic swinging signal of motion control network output, Feed _iBe feedback input signal, w _IjBe the connection weight between neuron j and neuron i, r _iBe the output of i CPG unit, for oscillation-damped device output signal is the influence of null part, the design adopts musculus flexor neuron, the neuronic status items of extensor

(the Kimura model utilizes neuronic output item in the output of linear synthetic oscillator

With

The output of linear synthetic oscillator).

In the step 1) when making up the CPG network topology structure, other joints of the hip joint of biped robot and shank are designed health control network (body network) and shank respectively to be controlled network (leg network) and controls, purpose is in the symmetry that guarantees biped robot left and right sides leg control signal and robot in the process of walking in the phase relation of left and right sides leg, the complexity of attenuating system reduces the network parameter that needs optimization.

Step 2) optimization method concrete in is: only consider between the musculus flexor neuron and the mutual rejection coefficient between the extensor neuron obtains phase relation between the joint freedom degrees.

Step 3) specifically may further comprise the steps:

A) adopt the single-parameter analysis method to obtain the single model parameter of CPG control network for the trend that influences of joint control signal;

B) according to each model parameter to the trend that influences of joint control signal and auxiliary multi-target evolution computational methods, obtain and can make robot in the optimum joint of level walking control signal;

C) according to the walking effect of reality, to step B) in model parameter under the control signal of optimum joint finely tune.

Compared with prior art, the present invention is based on the space control method of the biped robot of CPG, proposed a kind of new CPG control network establishing method, and designed more the rational and effective model parameter strategy of adjusting.The design has improved traditional control network establishing method to a certain extent, makes its complexity reduce, and the control network structure is more reasonable.

Description of drawings

Fig. 1 is the control system integral frame of joint space control method;

Fig. 2 is the joint freedom degrees distribution map of biped robot;

Fig. 3 is for being used for health control network (body network) the partial C PG network structure of control hip joint;

Fig. 4 is the oscillator signal of health control network (body network) part output;

Fig. 5 is the CPG control network topology structure that is proposed by the present invention;

Fig. 6 is the mode that is of coupled connections between musculus flexor neuron and the extensor neuron.

The specific embodiment

The present invention is described in detail below in conjunction with the drawings and specific embodiments.

Embodiment

A kind of biped robot CPG control network topology building method may further comprise the steps:

Step 1: shown in Figure 2 is the free degree distribution schematic diagram of two sufficient anthropomorphic robot shanks of present modal research and application.For the biped robot of the type walking control, the hip joint free degree to about the control of two leg phase places the most important.Therefore the design controls network with CPG and is divided into for health control network (body network) part of control hip joint and is used for shank control network (leg network) part in control shank joint, realizes the symmetry of biped robot left and right sides leg control signal and the robot phase relation of left and right sides leg in the process of walking.Be example with health control network (body network) part, as shown in Figure 3, the bidirectional couple by four neuron elements connects and composes health and controls four frees degree that network (body network) is controlled the biped robot hip joint.Health is controlled the mutual inhibition of adopting equal weight between network (body network) four neuron elements partly connect (suppress weight and be made as-1), can obtain phase relation as shown in Figure 4, the HipPitch phase difference of left and right sides leg is 180 °, satisfy the anti-phase relation of left and right sides hip joint fore-and-aft direction free degree control signal, can obtain the 90 ° desirable phase difference relation of two frees degree in same joint simultaneously.

Herein, CPG control network using neuron oscillator model, the mathematic(al) representation of this model is:

$T_r{\stackrel{\·}{u}}_i^{\{e,f\}}={-u}_i^{\{e,f\}}+w_{\mathrm{fe}}{r}_i^{\{f,e\}}-\mathrm{\β}{v}_i^{\{e,f\}}+s_0+{\mathrm{Feed}}_i^{\{e,f\}}+\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{r}_j^{\{e,f\}}$

$T_a{\stackrel{\·}{v}}_i^{\{e,f\}}=-v_i^{\{e,f\}}+r_i^{\{e,f\}}$

$r_i^{\{e,f\}}=\max (u_i^{\{e,f\}},0)$

$r_i={-u}_i^{\left\{e\right\}}+u_i^{\left\{f\right\}}$

Wherein, i represents i CPG unit, and e represents the musculus flexor neuron, and f represents extensor neuron, u _iBe neuronic internal state, v _iFor neuron from holddown,

Be neuronic output, T _rAnd T _aBe respectively rise time and adaptation time constant, w _FeBe neuronic mutual rejection coefficient, β is neuronic from rejection coefficient, s ₀Represent the periodic swinging signal of motion control network output, Feed _iBe feedback input signal, w _IjBe the connection weight between neuron j and neuron i, r _iBe the output of i CPG unit, for oscillation-damped device output signal is the influence of null part, the design adopts musculus flexor neuron, the neuronic status items of extensor

(the Kimura model utilizes neuronic output item in the output of linear synthetic oscillator With

The output of linear synthetic oscillator).

Step 2: the mode of being of coupled connections between the neuron in the CPG control network is optimized, reduces the complexity of CPG network.Connected mode between traditional neuron is musculus flexor neuron and the coupling of extensor neuron omnidirectional, for two neuron elements, needs to consider the modulation of four connection weights.For the biped robot of a plurality of frees degree, if according to traditional connected mode, need the connection number of parameters of optimization huge, be difficult to obtain the proper optimization result.Adopt the method that is of coupled connections as shown in Figure 6 among the present invention: consider between the musculus flexor neuron and phase relation that the mutual rejection coefficient between the extensor neuron just can obtain expecting, therefore in order to reduce the complexity of whole topological structure.

Step 3: based on the carry out parameter tuning of evolutionary computation to CPG control network, optimize model parameter and the topological connected mode of CPG control network, specifically can be subdivided into following a few step:

At first, adopt the single-parameter analysis method to obtain the single model parameter (T of CPG control network _r, T _a, β, s ₀And w _Fe) for the trend that influences of joint control signal.Owing to be to intercouple between the parameter, there is not linear relationship in parameter to the influence of output, and selected one of them model parameter only changes the value of this measured parameter as measured parameter, by judging the whether stable oscillation stationary vibration of CPG output signal, estimate the span of measured parameter; In the scope of estimating, regulate measured parameter continuously, measure the quantized sign amount (amplitude, frequency, phase place etc.) of output and the quantitative relationship of measured parameter;

Then, obtain to make the joint control signal of robot level walking based on the evolutionary computation method.Based on the joint space control method of CPG owing to do not calculate based on kinematics model and the inverse kinematics of biped robot, so the stability of robot ambulation can only obtain realizing the joint control signal of each free degree coordination of robot by optimizing the parameter of CPG network.In order to improve the walking stability of robot, the stability margin when the design considers to utilize robot ambulation in the parameter optimization target (with ZMP to the stability margin of the beeline that supports the convex polygon border as pedestrian system) improves the stability of robot ambulation.The design utilizes the multi-target evolution algorithm, based on the NSGA-II(Non-Dominated Sorting Genetic Algorithm-II of Kalyanmoy Deb) carry out the design of multi-objective genetic algorithm.The target of optimizing is to obtain realizing the optimum topological connected mode of robot level walking and the model parameter of CPG, and feasibility and the stability of robot ambulation is considered in the design of fitness function, is designed to:

${\mathrm{<figure-callout\; id="1"\; label="fitness"\; filenames="DEST\_PATH\_HDA00003146800300021.png"\; state="\{\{state\}\}">fitness</figure-callout>}}_1=1/\sqrt{{(x_{\mathrm{end}}-x_0)}^2}$

fitness ₂＝1/D _s

Wherein, x ₀Be the initial position (x direction) of robot, x _EndPosition (x direction) when representing robot and stopping to walk.First object function is in order to realize the straight line moving of robot, therefore only considers robot (x direction) the direction travel distance that advances when optimizing.Second optimization aim considered the stability of walking, D _sStability margin for robot ambulation.

At last, according to the walking effect of biped robot reality, to step B) in model parameter under the control signal of optimum joint finely tune, it is optimized.At last, according to the robot ambulation experiment effect of reality, at completing steps A) and step B) the basis on comprehensively adjust parameter in conjunction with designer's experience.This parameter tuning method fully combines computational intelligence and human intelligence, allows the designer progressively deepen the understanding to controller and controlled device, and to obtain new design philosophy and inspiration according to the comprehensive output result who optimizes algorithm.

The present invention is based on the space control method of the biped robot of CPG, a kind of new CPG control network establishing method has been proposed, and designed more the rational and effective model parameter strategy of adjusting, improved traditional control network establishing method to a certain extent, make its complexity reduce, the control network structure is more reasonable.

Claims (5)

1. a biped robot CPG control network topology building method is characterized in that, may further comprise the steps:

1) CPG is controlled network and be divided into for health control network (body network) part of control hip joint and be used for shank control network (leg network) part in control shank joint, guarantee the symmetry of biped robot left and right sides leg control signal and the robot phase relation of left and right sides leg in the process of walking;

2) mode of being of coupled connections between CPG network introflexion muscular nerve unit and the extensor neuron is optimized, reduces the complexity of CPG control network;

3) CPG control network is carried out parameter tuning, construct the optimal network topological structure.

2. a kind of biped robot CPG control network topology building method according to claim 1 is characterized in that, the CPG network using neuron oscillator model in the step 1), and the mathematic(al) representation of this model is:

Wherein, i represents i CPG unit, and e represents the musculus flexor neuron, and f represents extensor neuron, u _iBe neuronic internal state, v _iFor neuron from holddown,

Be neuronic output, T _rAnd T _aBe respectively rise time and adaptation time constant, w _FeBe neuronic mutual rejection coefficient, β is neuronic from rejection coefficient, s ₀Represent the periodic swinging signal of motion control network output, Feed _iBe feedback input signal, w _IjBe the connection weight between neuron j and neuron i, r _iBe the output of i CPG unit, for oscillation-damped device output signal is the influence of null part, adopt musculus flexor neuron, the neuronic status items of extensor The output of linear synthetic oscillator.

3. a kind of biped robot CPG according to claim 1 controls network topology building method, it is characterized in that: with the hip joint of biped robot with health control network (body network) part is designed in the shank joint respectively and shank control network (leg network) is partly controlled, in the symmetry that guarantees biped robot left and right sides leg control signal and robot in the process of walking in the phase relation of left and right sides leg, the complexity of attenuating system reduces the network parameter that needs optimization.

4. a kind of biped robot CPG according to claim 1 controls network topology building method, the method of attachment of oscillating unit intrinsic nerve unit is optimized, it is characterized in that step 2) in concrete optimization method be: only consider between the musculus flexor neuron and the mutual rejection coefficient between the extensor neuron obtains phase relation between the joint freedom degrees.

5. a kind of biped robot CPG according to claim 1 controls network topology building method, it is characterized in that step 3) specifically may further comprise the steps:

A) adopt the single-parameter analysis method to obtain the single model parameter of CPG control network for the trend that influences of joint control signal;

B) according to each model parameter to the trend that influences of joint control signal and auxiliary multi-target evolution computational methods, obtain and can make robot in the optimum joint of level walking control signal;

C) according to the walking effect of reality, to step B) in model parameter under the control signal of optimum joint finely tune.

Images