CN101953727A - Solution method of joint space parameters of artificial limb in multiple degrees of freedom - Google Patents

Abstract

The invention discloses a solution method of the joint space parameters of an artificial limb in multiple degrees of freedom, which comprises the following steps of: firstly, acquiring an input sample set and a target sample set for training the artificial neural network; then, training the artificial neural network by respectively using the acquired input sample set and the acquired target sample set as input parameters and target parameters of the artificial neural network so as to obtain a solution model of the joint space parameters of the artificial limb or a robot in multiple degrees of freedom. The solution method of a joint space of the artificial limb in multiple degrees of freedom can be used for the functional compensation of a 'shoulder breakage'type disabled person under the situation of only knowing the operating spatial position of a target but not knowing required hand gestures. The method can acquire certain feasible solution for the joint space parameters of the artificial limb or the robot in the multiple degrees of freedom under the situation, so that the artificial limb or the robot can complete designated operation tasks.

Description

The method for solving of joint of artificial limb spatial parameter on the multiple degrees of freedom

Technical field

The present invention relates to method for solving based on the multi-freedom artificial limb joint space parameter of artificial neural network, relate in particular to a kind of in the locus of only knowing target, and do not know that institute requires under the hand attitude situation, be used for the method for solving of artificial limb (referring to upper extremity prosthesis) joint space on the multiple degrees of freedom of " shoulder from disconnected " type people with disability's functional compensation.This method also can be used in finding the solution of other robotic joint space with similar operative scenario.

Background technology

For finding the solution of joint of artificial limb spatial parameter on the multiple degrees of freedom, relate to the application of robot inverse kinematics knowledge.

Learn in the equation at the positive motion of robot, the parameter value in each joint of robot is an independent variable, and robot hand is a dependent variable in attitude and the position of working place, and the positive motion equation of robot is according to the foundation of deriving of robot construction characteristic; In the inverse kinematics equation of robot, robot hand is an independent variable in the attitude and the position of working place, the parameter value in each joint of robot is a dependent variable, and the inverse kinematics equation of robot generally is oppositely to be derived by the positive motion equation of robot with analytic method to produce.In real work, people often prophet's pipeline robot hand in the attitude and the position of working place, and then go out the parameter value in each joint of robot according to the inverse kinematics equation differentiate of robot, so that each joint of robot can turn over corresponding angle according to the numerical value of obtaining, hand is put on the target with correct attitude and position.At this, every is independent variable with robot hand in the attitude and the position of working place, is the inverse kinematics solving model that the solving model of dependent variable just is called robot with the parameter value in each joint of robot, perhaps is called joint space parametric solution model.

Usually, learn in the analytic equation at the positive motion of robot, the number of independent variable is exactly the quantity in robot motion joint, and its numerical value is exactly the description of each joint rotation angle or amount of movement; The dependent variable number is 12, wherein has 9 to be hand attitude characterising parameter, has 3 to be hand center characterising parameter, its whole numerical value just target in the attitude of working place and the description of position.

Usually, in the inverse kinematics analytic equation of robot, the number of independent variable is 12, wherein has 9 to be hand attitude characterising parameter, have 3 to be hand center characterising parameter, its whole numerical value just target in the attitude of working place and the description of position; The dependent variable number is exactly the quantity in robot motion joint, and its numerical value is exactly the description of each joint rotation angle or amount of movement.

Like this, when adopting general analytical method to find the solution robot inverse kinematics equation, must know 12 parameters of target in the working place in advance, 9 attitude parameters and 3 space position parameter of comprising target, could make solving condition reach satisfied, just can obtain separating of robotic joint space.

For artificial limb or robot on the multiple degrees of freedom that is operated in the destructuring environment, obtain its more or less freely accomplishing of working place positional information of target at random, but how to describe at random the attitude information of target but is a very thing of difficulty.Only knowing 3 working place location parameters of target at random, and do not knowing under its 9 attitude parameter situations how to try to achieve separating of a kind of feasible multi-freedom artificial limb or robotic joint space, at present the research of unmatchful this technical method narration still.This is the problem that this area research worker is needed solution badly.

Summary of the invention

Purpose of the present invention is only to know the three-dimensional position parameter data of target at random, and can't obtains under its attitude information situation, for finding the solution of joint of artificial limb spatial parameter on the multiple degrees of freedom invented a kind of solution.

For reaching above-mentioned purpose, the method for solving of invention is only to know at random under 3 position parameter data situations of target in the working place, introduces a kind of solution that joint of artificial limb space on the multiple degrees of freedom is found the solution of realizing.Plan utilizes artificial neural network that the nonlinear characteristic of the parameter mapping ability of imperfect information uniqueness, fast parallel information processing capability and multiple-input and multiple-output is set up a kind of special multi-freedom artificial limb joint space parametric solution model.When the multi-freedom artificial limb system obtains D coordinates value in the working place of target at random, the compensatory attitude that these 3 coordinate figures and certain are owed to define is described the input as network model, comes to obtain fast, concurrently certain feasible value of finding the solution in artificial limb or each joint of robot with this.

Method for solving detailed step of the present invention is as follows:

The first step: the input sample set and the target sample collection that obtain the training of human artificial neural networks

At first, when the vector parameters q in each joint is learned equation as artificial limb on the known conditions substitution multiple degrees of freedom or robot positive motion, obtain the homographic solution p about the working place of artificial limb on the multiple degrees of freedom or robot.When at joint space with discretization method, scan mode is to each joint vector parameters q gradually _i(i=1,2 ... when n) carrying out value, if abundant n group data can be so that the various interblock spaces in artificial limb or each joint of robot can both be scanned by discretization, just can obtain artificial limb or robot all working position hand, that reach with various attitude the discretization working place separate p _i(i=1,2 ..., set n).

To aforesaid p _i(i=1,2 ..., the data in n) are formed set { P} will be gathered that { P} is as the input sample set of training of human artificial neural networks through the special back of handling; With aforesaid q _i(i=1,2 ..., n) data compilation is formed set { Q} will be gathered that { Q} is as the target sample collection of training of human artificial neural networks.

Second step: training of human artificial neural networks

Will { P} be as the input parameter of artificial neural network, and { Q} is as the target component of artificial neural network and incite somebody to action.When the actual of artificial neural network is output as

The time, with error e} removes the training of human artificial neural networks, wherein:

$\left\{e\right\}=\left\{Q\right\}-\left\{\hat{Q}\right\}---\left(1\right)$

Feasible { the e} → { 0} of the result of training.

Through after training up, this artificial neural network the joint space parametric solution model of artificial limb or robot on the multiple degrees of freedom.When with the position of working place target and compensatory parameters such as hand attitude description during as the input of network, the artificial neural network system who has trained will ask certain feasible solution of the joint space parameter that obtains artificial limb on the multiple degrees of freedom or robot automatically.

The invention has the advantages that:

Invented a kind ofly, and do not known that institute requires under the hand attitude situation, be used for the method for solving of the joint space of artificial limb on the multiple degrees of freedom of " shoulder from disconnected " type people with disability's functional compensation in the position, working place of only knowing target.The joint space parameter that this method can be under the said circumstances artificial limb on the multiple degrees of freedom or robot obtains certain feasible solution, so that it can finish specified operation task.

Description of drawings

1-3 illustrates one embodiment of the present of invention with reference to the accompanying drawings.

Fig. 1 is the sketch map of artificial limb on certain multiple degrees of freedom, and wherein Fig. 1 (a) is a front view, and Fig. 1 (b) is a left view.

Fig. 2 is drive mechanism principle and each coordinate system graph of a relation that artificial limb is faced on certain multiple degrees of freedom.

Fig. 3 learns hand position and the attitude artificial limb motion simulation figure in the working place that equation obtained with the spatial solving result substitution of joint of artificial limb artificial limb positive motion.

Marginal data: 1-artificial limb shoulder member, the big arm member of 2-artificial limb, 3-artificial limb forearm rotational support, 4-artificial limb forearm revolving member, the 5-member of doing evil through another person,

The specific embodiment

The present invention is an example with artificial limb shown in Figure 1, and as can be seen from the figure, this artificial limb is made up of artificial limb shoulder member 1, the big arm member 2 of artificial limb, artificial limb forearm rotational support 3, artificial limb forearm revolving member 4 and the member 5 of doing evil through another person.

To joint of artificial limb spatial attitude parameter shown in Figure 1 to find the solution concrete implementation step as follows:

STEP1. draw structural principle and each coordinate system graph of a relation of artificial limb on the multiple degrees of freedom, as shown in Figure 2.

STEP2. according to the architectural characteristic of artificial limb on the multiple degrees of freedom,, listed as table 1 with its various parameter lists.

Table 1

STEP3. by table 1, A matrix and the T matrix of listing each rod member are as follows:

$A_0=\left[\begin{array}{cccc}\cos \left({\mathrm{\&theta;}}_0\right)& -\sin \left({\mathrm{\&theta;}}_0\right)\cos \left({\mathrm{\&alpha;}}_0\right)& \sin \left({\mathrm{\&theta;}}_0\right)\sin \left({\mathrm{\&alpha;}}_0\right)& a_0\cos \left({\mathrm{\&theta;}}_0\right)\\ \sin \left({\mathrm{\&theta;}}_0\right)& \cos \left({\mathrm{\&theta;}}_0\right)\cos \left({\mathrm{\&alpha;}}_0\right)& -\cos \left({\mathrm{\&theta;}}_0\right)\sin \left({\mathrm{\&alpha;}}_0\right)& a_0\sin \left({\mathrm{\&theta;}}_0\right)\\ 0& \sin \left({\mathrm{\&alpha;}}_0\right)& \cos \left({\mathrm{\&alpha;}}_0\right)& {\mathrm{<figure-callout\; id="0"\; label="d"\; filenames="HSA00000267952000031.png"\; state="\{\{state\}\}">d</figure-callout>}}_0\\ 0& 0& 0& 1\end{array}\right]---\left(2\right)$

$A_1=\left[\begin{array}{cccc}\cos \left({\mathrm{\&theta;}}_1\right)& -\sin \left({\mathrm{\&theta;}}_1\right)\cos \left({\mathrm{\&alpha;}}_1\right)& \sin \left({\mathrm{\&theta;}}_1\right)\sin \left({\mathrm{\&alpha;}}_1\right)& a_1\cos \left({\mathrm{\&theta;}}_1\right)\\ \sin \left({\mathrm{\&theta;}}_1\right)& \cos \left({\mathrm{\&theta;}}_1\right)\cos \left({\mathrm{\&alpha;}}_1\right)& -\cos \left({\mathrm{\&theta;}}_1\right)\sin \left({\mathrm{\&alpha;}}_1\right)& a_1\sin \left({\mathrm{\&theta;}}_1\right)\\ 0& \sin \left({\mathrm{\&alpha;}}_1\right)& \cos \left({\mathrm{\&alpha;}}_1\right)& {\mathrm{<figure-callout\; id="1"\; label="d"\; filenames="HSA00000267952000031.png"\; state="\{\{state\}\}">d</figure-callout>}}_1\\ 0& 0& 0& 1\end{array}\right]---\left(3\right)$

$A_2=\left[\begin{array}{cccc}\cos \left({\mathrm{\&theta;}}_2\right)& -\sin \left({\mathrm{\&theta;}}_2\right)\cos \left({\mathrm{\&alpha;}}_2\right)& \sin \left({\mathrm{\&theta;}}_2\right)\sin \left({\mathrm{\&alpha;}}_2\right)& a_2\cos \left({\mathrm{\&theta;}}_2\right)\\ \sin \left({\mathrm{\&theta;}}_2\right)& \cos \left({\mathrm{\&theta;}}_2\right)\cos \left({\mathrm{\&alpha;}}_2\right)& -\cos \left({\mathrm{\&theta;}}_2\right)\sin \left({\mathrm{\&alpha;}}_2\right)& a_2\sin \left({\mathrm{\&theta;}}_2\right)\\ 0& \sin \left({\mathrm{\&alpha;}}_2\right)& \cos \left({\mathrm{\&alpha;}}_2\right)& d_2\\ 0& 0& 0& 1\end{array}\right]---\left(4\right)$

$A_3=\left[\begin{array}{cccc}\cos \left({\mathrm{\&theta;}}_3\right)& -\sin \left({\mathrm{\&theta;}}_3\right)\cos \left({\mathrm{\&alpha;}}_3\right)& \sin \left({\mathrm{\&theta;}}_3\right)\sin \left({\mathrm{\&alpha;}}_3\right)& a_3\cos \left({\mathrm{\&theta;}}_3\right)\\ \sin \left({\mathrm{\&theta;}}_3\right)& \cos \left({\mathrm{\&theta;}}_3\right)\cos \left({\mathrm{\&alpha;}}_3\right)& -\cos \left({\mathrm{\&theta;}}_3\right)\sin \left({\mathrm{\&alpha;}}_3\right)& a_3\sin \left({\mathrm{\&theta;}}_3\right)\\ 0& \sin \left({\mathrm{\&alpha;}}_3\right)& \cos \left({\mathrm{\&alpha;}}_3\right)& d_3\\ 0& 0& 0& 1\end{array}\right]---\left(5\right)$

$A_4=\left[\begin{array}{cccc}\cos \left({\mathrm{\&theta;}}_4\right)& -\sin \left({\mathrm{\&theta;}}_4\right)\cos \left({\mathrm{\&alpha;}}_4\right)& \sin \left({\mathrm{\&theta;}}_4\right)\sin \left({\mathrm{\&alpha;}}_4\right)& a_4\cos \left({\mathrm{\&theta;}}_4\right)\\ \sin \left({\mathrm{\&theta;}}_4\right)& \cos \left({\mathrm{\&theta;}}_4\right)\cos \left({\mathrm{\&alpha;}}_4\right)& -\cos \left({\mathrm{\&theta;}}_4\right)\sin \left({\mathrm{\&alpha;}}_4\right)& a_4\sin \left({\mathrm{\&theta;}}_4\right)\\ 0& \sin \left({\mathrm{\&alpha;}}_4\right)& \cos \left({\mathrm{\&alpha;}}_4\right)& d_4\\ 0& 0& 0& 1\end{array}\right]---\left(6\right)$

$A_5=\left[\begin{array}{cccc}\cos \left({\mathrm{\&theta;}}_5\right)& -\sin \left({\mathrm{\&theta;}}_5\right)\cos \left({\mathrm{\&alpha;}}_5\right)& \sin \left({\mathrm{\&theta;}}_5\right)\sin \left({\mathrm{\&alpha;}}_5\right)& a_5\cos \left({\mathrm{\&theta;}}_5\right)\\ \sin \left({\mathrm{\&theta;}}_5\right)& \cos \left({\mathrm{\&theta;}}_5\right)\cos \left({\mathrm{\&alpha;}}_5\right)& -\cos \left({\mathrm{\&theta;}}_5\right)\sin \left({\mathrm{\&alpha;}}_5\right)& a_5\sin \left({\mathrm{\&theta;}}_5\right)\\ 0& \sin \left({\mathrm{\&alpha;}}_5\right)& \cos \left({\mathrm{\&alpha;}}_5\right)& d_5\\ 0& 0& 0& 1\end{array}\right]---\left(7\right)$

Order:

${}_{0}T=A_0=\left[\begin{array}{cccc}{}_{0}n_x& {}_{0}o_x& {}_{0}a_x& {}_{0}p_x\\ {}_{0}n_y& {}_{0}o_y& {}_{0}a_y& {}_{0}p_y\\ {}_{0}n_z& {}_{0}o_z& {}_{0}a_z& {}_{0}p_z\\ 0& 0& 0& 1\end{array}\right]---\left(8\right)$

Each rod member coordinate system that formula (2)-(7) are described is with respect to the position and the attitude of its front one coordinate system, what formula (8)-(12) were described is position and the attitude of the coordinate system of each rod member end of artificial limb in the working place, and what formula (13) was described is position and the attitude of the hand of artificial limb in the working place.

STEP4. according to the span and formula (2)-(13) in each joint in the table 1 joint space of artificial limb is scanned.In scanning joint space process, with i (i=1,2 ..., n) vector value of the joint space of inferior scanning is write as q _i(i=1,2 ..., n):

q _i(i，1)＝θ _1i

q _i(i，2)＝θ _2i

q _i(i，3)＝θ _3i (i＝1，2…，n) (14)

q _i(i，4)＝θ _4i

q _i(i，5)＝θ _5i

STEP5. will use the operation by human hand locus of formula (13) acquisition and the description simplification arrangement of attitude mutually and be write as p _i(i=1,2 ..., n):

p _i(i，1)＝ ₅T(1，3)＝ ₅a _xi

p _i(i，2)＝ ₅T(2，3)＝ ₅a _yi

p _i(i，3)＝ ₅T(3，3)＝ ₅a _zi

p _i(i，4)＝ ₅T(1，4)＝ ₅p _xi (i＝1，2…，n)(15)

p _i(i，5)＝ ₅T(2，4)＝ ₅p _yi

p _i(i，6)＝ ₅T(3，4)＝ ₅p _zi

That is: at p _i(i=1,2 ..., only kept hand Z in n) ₅The description of axle attitude and palm of the hand position.

STEP6. through after the n time abundant scanning, make the various interblock spaces in artificial limb or each joint of robot to be scanned, that is the position and the attitude of artificial limb in the working place also can both discretization obtain by discretization.With aforesaid q _i(i=1,2 ..., n) { Q}, { Q} will be used as the target sample collection of training of human artificial neural networks to the set of data compilation composition in set.With aforesaid p _i(i=1,2 ..., n) { P}, { P} will be used as the input sample set of training of human artificial neural networks in set in the data compilation composition set in.

STEP7. { { Q} is used for the training of human artificial neural networks for P} and set with set.A kind of major part of MATLAB program of the BP of training neutral net (wherein the subnetwork parameter is adjustable) as follows:

Xmin=min (P (:, 4)); % calculates the minima of X

Xmax=max (P (:, 4)); % calculates the maximum of X

Ymin=min (P (:, 5)); % calculates the minima of Y

Ymax=max (P (:, 5)); % calculates the maximum of Y

Zmin=min (P (:, 6)); % calculates the minima of Z

Zmax=max (P (:, 6)); % calculates the maximum of Z

Net=newff ([1 1;-11;-11; Xmin xmax; Ymin ymax; Zminzmax], [30,50,5], { ' tansig ', ' purelin ' }, ' traingdx '); % sets up the BP network

Net.trainParam.goal=0.001; The iteration precision of % project training

Net.trainParam.epochs=500000; % designs maximum iteration operation step-lengths.

[net, tr]=train (net, P ', Q '); The % training network

The weights and the threshold value of each layer of % output training back:

iwl＝net.IW{1}；

bl＝net.b{1}；

lw2＝net.LW{2}；

b2＝net.b{2}；

The neutral net that the % storage trains:

Save net net % is stored as the MATLAB file of name for " net " (being meant front one " net " at this) with the artificial nerve network model " net " (being meant next " net " at this) that trains.

The artificial nerve network model of above-mentioned " net " by name is exactly the multi-freedom artificial limb joint space parametric solution model of requirement.

STEP8. when needs use this model, with the Z of hand coordinate system ₅Direction of principal axis always makes progress, and is used as a compensatory description of hand attitude.Like this, as position, the working place p that obtains certain target _sHomogeneous being described below the time:

$p_s=\left[\begin{array}{c}{x}_s\\ y_s\\ z_s\\ 1\end{array}\right]---\left(16\right)$

Always can be with the input vector p of joint of artificial limb spatial parameter solving model on the multiple degrees of freedom _nBe expressed as:

$p_n=\left[\begin{array}{c}0\\ 0\\ 1\\ x_s\\ y_s\\ z_s\end{array}\right]---\left(17\right)$

STEP9. with p _nImport aforesaid artificial neural network solving model as input vector, just can obtain rapidly that joint of artificial limb is spatial on the multiple degrees of freedom separates.

Below with specific embodiment, and with reference to the application of the above artificial neural network solving model about joint of artificial limb spatial parameter on the multiple degrees of freedom of accompanying drawing 1-3 explanation.

Embodiment one:

When we obtain homogeneous the be described as p of target about the position, working place _S1The time:

${\mathrm{<figure-callout\; id="1"\; label="p\; s"\; filenames="HSA00000267952000031.png"\; state="\{\{state\}\}">p</figure-callout>}}_s=\left[\begin{array}{c}400\\ -100\\ -300\\ 1\end{array}\right]$

Input p with network model _N1Write as following form:

${\mathrm{<figure-callout\; id="1"\; label="p\; n"\; filenames="HSA00000267952000031.png"\; state="\{\{state\}\}">p</figure-callout>}}_n=\left[\begin{array}{c}0\\ 0\\ 1\\ 400\\ -100\\ -300\end{array}\right]$

With p _N1Import in the above-mentioned network model and can export, be i.e. the spatial q that separates of joint of artificial limb on the multiple degrees of freedom in the hope of network _N1For:

With above solving result again this artificial limb positive motion of substitution learn in equation-formula (13), can be in the hope of the physical location and attitude description of prosthetic hand portion in the working place ₅T ₁Will for:

${}_5\mathrm{<figure-callout\; id="1"\; label="T"\; filenames="HSA00000267952000031.png"\; state="\{\{state\}\}">T</figure-callout>}_1=\left[\begin{array}{cccc}0.4109& -0.5462& -0.7300& 399.2204\\ 0.8362& -0.0931& 0.5404& -100.1796\\ -0.3631& -0.8325& 0.4184& -297.7855\\ 0& 0& 0& 1.0000\end{array}\right]$

That is, under not strict consideration hand attitude situation, the homogeneous description p ' of the position, working place of the actual arrival of artificial limb hand _S1For:

${\mathrm{<figure-callout\; id="1"\; label="p\; s"\; filenames="HSA00000267952000031.png"\; state="\{\{state\}\}">p</figure-callout>}}_s^1=\left[\begin{array}{c}399.2204\\ -100.1796\\ -297.7855\\ 1.0000\end{array}\right]$

The demonstration of a broken line in the middle of seeing among Fig. 3.

Embodiment two:

When we obtain homogeneous the be described as p of target about the position, working place _S2The time:

$p_{s2}=\left[\begin{array}{c}400\\ -300\\ -300\\ 1\end{array}\right]$

Network model is imported p _N2Write as following form:

$p_{n2}=\left[\begin{array}{c}0\\ 0\\ 1\\ 400\\ -300\\ -300\end{array}\right]$

With p _N2Import in the above-mentioned network model and can export, be i.e. the spatial q that separates of joint of artificial limb on the multiple degrees of freedom in the hope of network _N2For:

With above solving result again this artificial limb positive motion of substitution learn in equation-formula (13), can be in the hope of the physical location and attitude description of prosthetic hand portion in the working place ₅T ₂Will for:

${}_{5}T_2=\left[\begin{array}{cccc}0.7469& -0.5329& -0.3978& 400.1136\\ 0.2671& -0.3074& 0.9133& -301.4309\\ -0.6089& -0.7884& -0.0872& -298.6426\\ 0& 0& 0& 1.0000\end{array}\right]$

That is, under not strict consideration hand attitude situation, the homogeneous description p ' of the position, working place of the actual arrival of artificial limb hand _S2For:

$p_{s2}^{,}=\left[\begin{array}{c}400.1136\\ -301.4309\\ -298.6426\\ 1.0000\end{array}\right]$

See the demonstration of the broken line in the left side among Fig. 3.

Embodiment three:

When we obtain homogeneous the be described as p of target about the position, working place _S3The time:

$p_{s3}=\left[\begin{array}{c}400\\ 200\\ -300\\ 1\end{array}\right]$

Network model is imported p _N3Write as following form:

$p_{n3}=\left[\begin{array}{c}0\\ 0\\ 1\\ 400\\ 200\\ -300\end{array}\right]$

With p _N3Import in the above-mentioned network model and can export in the hope of network, promptly the multi-freedom artificial limb joint space separates q _N3For:

With above solving result again this artificial limb positive motion of substitution learn in equation-formula (13), can be in the hope of the physical location and attitude description of prosthetic hand portion in the working place ₅T ₃Will for:

${}_{5}T_3=\left[\begin{array}{cccc}0.4452& -0.8771& 0.1803& 401.0491\\ 0.8913& 0.4147& -0.1833& 198.9831\\ 0.0860& 0.2423& 0.9664& -302.6534\\ 0& 0& 0& 1.0000\end{array}\right]$

That is, under not strict consideration hand attitude situation, the homogeneous description p ' of the position, working place of the actual arrival of artificial limb hand _S3For:

$p_{s3}^{,}=\left[\begin{array}{c}401.0491\\ 198.9831\\ -302.6534\\ 1.0000\end{array}\right]$

See the demonstration of the broken line in the right side among Fig. 3.

By above three embodiment as can be seen, method of the present invention can be separated and joint of artificial limb is spatial on a kind of feasible multiple degrees of freedom of acquisition under the targeted attitude loss of learning situation in the position, working place of only knowing target.Though the attitude that this method requires hand can't accurate description and is realized it, but can be so that hand reach the target position with relative error less than 1% precision, the problem that can't find the solution for common analytic method finds a kind of feasible solution.

The content of not describing in detail in the description of the present invention belongs to this area professional and technical personnel's known prior art or theory.

Claims (1)

1. the method for solving of joint of artificial limb spatial parameter on the multiple degrees of freedom is characterized in that detailed step is as follows:

The first step: the input sample set and the target sample collection that obtain the training of human artificial neural networks

At first, when the vector parameters q in each joint is learned equation as artificial limb on the known conditions substitution multiple degrees of freedom or robot positive motion, obtain the homographic solution p about the working place of artificial limb on the multiple degrees of freedom or robot; When at joint space with discretization method, scan mode is to each joint vector parameters q gradually _i(i=1,2 ... when n) carrying out value, if abundant n group data can be so that the various interblock spaces in artificial limb or each joint of robot can both be scanned by discretization, just can obtain artificial limb or robot all working position hand, that reach with various attitude the discretization working place separate p _i(i=1,2 ..., set n);

To aforesaid p _i(i=1,2 ..., the data in n) are formed set { P} will be gathered that { P} is as the input sample set of training of human artificial neural networks through the special back of handling; With aforesaid q _i(i=1,2 ..., n) data compilation is formed set { Q} will be gathered that { Q} is as the target sample collection of training of human artificial neural networks;

Second step: training of human artificial neural networks

Will { P} be as the input parameter of artificial neural network, and { Q} is as the target component of artificial neural network and incite somebody to action; When the actual of artificial neural network is output as

The time, with error e} removes the training of human artificial neural networks, wherein:

$\left\{e\right\}=\left\{Q\right\}-\left\{\hat{Q}\right\}---\left(1\right)$

Feasible { the e} → { 0} of the result of training;

Through after training up, this artificial neural network the joint space parametric solution model of artificial limb or robot on the multiple degrees of freedom; When with the position of working place target and compensatory hand attitude characterising parameter during as the input of network, the artificial neural network system who has trained will ask certain feasible solution that obtains artificial limb on the multiple degrees of freedom or robotic joint space parameter automatically.

Images