CN115470593A - SVR-PSO-based voice robot shape optimization method - Google Patents

SVR-PSO-based voice robot shape optimization method Download PDF

Info

Publication number
CN115470593A
CN115470593A CN202211197359.3A CN202211197359A CN115470593A CN 115470593 A CN115470593 A CN 115470593A CN 202211197359 A CN202211197359 A CN 202211197359A CN 115470593 A CN115470593 A CN 115470593A
Authority
CN
China
Prior art keywords
value
model
shape
optimization
axis
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
Application number
CN202211197359.3A
Other languages
Chinese (zh)
Inventor
孟晨晨
梁艳春
管仁初
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhuhai Institute Of Science And Technology
Original Assignee
Zhuhai Institute Of Science And Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhuhai Institute Of Science And Technology filed Critical Zhuhai Institute Of Science And Technology
Priority to CN202211197359.3A priority Critical patent/CN115470593A/en
Publication of CN115470593A publication Critical patent/CN115470593A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computer Hardware Design (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Optimization (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Analysis (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention discloses a voice robot shape optimization method based on SVR-PSO, which comprises the following steps: the method comprises the following steps of firstly, modeling an appearance shell and internal parts of the intelligent voice robot to obtain a shell model and a part assembly model; carrying out parametric representation on the shell model to obtain an appearance fitting function, and carrying out parametric representation on the assembly body model to obtain the projection heights of the multiple parts; step three, constructing a shape optimization model; and step four, optimizing the shape optimization model by combining a particle swarm optimization algorithm and a support vector regression algorithm, searching the optimal solution of the current shape optimization model, and adjusting the shape of the intelligent voice robot to the minimum volume. The invention has the characteristics of shortening the optimization time and improving the calculation precision.

Description

SVR-PSO-based voice robot shape optimization method
Technical Field
The invention relates to the technical field of robot shape optimization, in particular to a shape optimization method of a voice robot based on SVR-PSO.
Background
The intelligent robot is small enough to be a household cleaning robot, a learning guidance robot, a service robot of a hotel or a shopping mall, large enough to be a medical rescue robot, a post-disaster ruin search robot and even an aerospace robot, and has higher requirements on space optimization and volume reduction compared with the traditional industrial robot due to the structural design of the intelligent robot. The overlarge volume can cause the internal space utilization rate of the intelligent robot to be low, the occupied area to be large and the intelligent robot to be slightly heavy, and the problems of inconvenient movement, troublesome transportation and carrying and the like can be caused in certain application occasions, so that the optimization of the appearance is particularly important in the structural design work of the intelligent robot.
However, for the optimization design of the shape of the intelligent voice robot, because some necessary design variables such as the size of internal parts, the thickness of a shell and the like belong to discrete variables, the constraint condition is complex, and the nonlinearity degree is high, how to establish a definite objective function expression becomes a big difficulty in handling the problem.
In recent years, group intelligent optimization algorithms are more applied to the robot structure optimization problem. As a random optimization algorithm, a Particle Swarm Optimization (PSO) can complete optimization only by utilizing a fitness function, and has good parallel computing capability and global searching capability, so the PSO is often applied to solving the optimization problem of a large-scale complex structure, but the PSO also has the defects of the PSO, for example, particles are easy to gather in a small area in the later iteration stage, so that the algorithm is trapped in a local optimal solution, and therefore, the search of the PSO suitable for the structural shape of the voice recognition robot has important significance for the research in the field of the shape optimization of the robot.
Disclosure of Invention
The invention aims to design and develop a shape optimization method of a voice robot based on SVR-PSO, and reduce the calculation times of fitness, shorten the optimization time and improve the calculation precision by a particle swarm optimization algorithm based on a support vector regression algorithm.
The technical scheme provided by the invention is as follows:
a shape optimization method of a voice robot based on SVR-PSO comprises the following steps:
the method comprises the following steps that firstly, modeling is carried out on an appearance shell and internal parts of the intelligent voice robot to obtain a shell model and a part assembly model;
carrying out parametric representation on the shell model to obtain a shape fitting function, and carrying out parametric representation on the assembling body model to obtain the projection heights of a plurality of parts;
step three, constructing an appearance optimization model:
taking the minimum volume of the shell model as an optimization target, taking an equation coefficient in an appearance fitting function as an optimization variable, and taking a part assembly model as a constraint condition;
optimizing the shape optimization model through a particle swarm optimization algorithm, wherein after the iteration times are reached, the minimum value of the fitness function of the particles is the optimal solution of the shape optimization model, and the shape of the intelligent voice robot is adjusted to the minimum volume;
the operation time of the particle swarm optimization algorithm is calculated through a support vector regression algorithm;
the fitness function of the particle is:
Figure BDA0003870772400000021
where fitness is the fitness function, v 0 Is the volume of the original appearance, v is the volume corresponding to the new individual, and fit is the fitness factor.
Preferably, the parameterization representation of the shell model comprises the following steps:
step 1, deriving a point cloud of a shell model;
step 2, performing centralization treatment on the point cloud:
calculating the mean values of all coordinate points in the point cloud on the X axis and the Y axis, and subtracting the mean values from all the coordinate points one by one to obtain processed coordinate points;
step 3, fitting an polynomial of the shape curved surface of the robot according to the processed coordinate points:
the method comprises the following steps of establishing a shell model coordinate system by taking intersection points of a central shaft of a shell model and upper and lower boundaries as coordinate origin points, taking an upper and lower boundary surfaces as an XOY plane, taking the central shaft of the shell model as a v-axis and taking the central shaft of the shell model as a positive direction, fitting an upper curved surface function by using coordinate points with a z value larger than zero, fitting a lower curved surface function by using coordinate points with a z value smaller than zero, and obtaining a shape fitting function:
f up (x，y)＝W 0_0 +W 0_1 x+W 0_2 y+W 0_3 x 2 +W 0_4 x*y+W 0_5 y 2 +W 0_6 x 3 +W 0_7 y*x 2 +W 0_8 x*y 2 +W 0_9 y 3 +W 0_10 x 4 +W 0_11 x 3 *y+W 0_12 x 2 *y 2 +W 0_13 x*y 3 +W 0_14 y 4 +W 0_15 x 5 +W 0_16 x 4 *y+W 0_17 x 3 *y 2 +W 0_18 x 2 *y 3 +W 0_19 x*y 4 +W 0_20 y 5 +W 0_21 x 6 +W 0_22 x 5 *y+W 0_23 x 4 *y 2 +W 0_24 x 3 *y 3 +W 0_25 x 2 *y 4 +W 0_26 x*y 5 +W 0_27 y 6
f down (x，y)＝W 1_0 +W 1_1 x+W 1_2 y+W 1_3 x 2 +W 1_4 x*y+W 1_5 y 2 +W 1_6 x 3 +W 1_7 y*x 2 +W 1_8 x*y 2 +W 1_9 y 3 +W 1_10 x 4 +W 1_11 x 3 *y+W 1_12 x 2 *y 2 +W 1_13 x*y 3 +W 1_14 y 4 +W 1_15 x 5 +W 1_16 x 4 *y+W 1_17 x 3 *y 2 +W 1_18 x 2 *y 3 +W 1_ 19 x*y 4 +W 1_20 y 5 +W 1_21 x 6 +W 1_22 x 5 *y+W 1_23 x 4 *y 2 +W 1_24 x 3 *y 3 +W 1_25 x 2 *y 4 +W 1_26 x*y 5 +W 1_27 y 6
in the formula (f) up (x, y) is the upper surface function, W 0_0 、W 0_1 、W 0_2 、W 0_3 …W 0_21 、W 0_22 、W 0_23 …W 0_27 Equation coefficients, f, which are all functions of the upper surface down (x, y) is the lower surface function, W 1_0 、W 1_1 、W 1_2 、W 1_3 …W 1_21 、…W 1_27 Are all the equation coefficients of the lower surface function.
Preferably, the value ranges of the X axis and the Y axis of the shell model in the shell model coordinate system are both [ -12,12] determined by the shape fitting function.
Preferably, the parameterization of the fitting model comprises:
projecting a plurality of parts in the assembly body to a Z axis in the shell model coordinate system to obtain the projection height of the parts on the Z axis, and if the plurality of parts are intersected in the same projection range and the projection of the parts is on the positive direction of the Z axis, taking the maximum value of the height of the parts as the parameterized representation of the parts; and if the parts intersect in the same projection range and the projection of the parts is in the negative direction of the Z axis, taking the minimum value of the height of the parts as the parameterized representation of the parts.
Preferably, the volume calculation of the shell model comprises:
dividing the bottom surface of the shell model into small squares of 1mm x 1mm, obtaining the coordinate of the central point of each small square in a shell model coordinate system, respectively substituting the coordinate of the central point of each small square into an appearance fitting function to obtain an upper curved surface function value and a lower curved surface function value, if the difference value of subtracting the lower curved surface function value from the upper curved surface function value is greater than 0, the difference value is the height value of the corresponding cube, and then calculating the volumes of all cubes to further obtain the volume of the shell model; and if the difference value obtained by subtracting the lower curved surface function value from the upper curved surface function value is not more than 0, the volume of the corresponding cube is 0, and then the volumes of all the small squares are calculated to further obtain the volume of the shell model.
Preferably, the constraint condition includes:
when the x value of the part assembling body model is within the value range of the x value of the shell model, and the y value of the part assembling body model is within the value range of the y value of the shell model:
if the shape-optimized coordinate point is in the Z-axis positive direction in the shell model coordinate system, calculating a curved surface function value, and if the difference value between the shape-optimized coordinate point and the Z-axis positive part parameterization is greater than 0, the shape-optimized coordinate point is a feasible solution;
and if the coordinate point after the shape optimization is in the Z-axis negative direction in the shell model coordinate system, calculating a lower curved surface function value, and if the difference value of the Z-axis negative direction part parameterization and the Z-axis negative direction part parameterization is more than 0, determining the coordinate point after the shape optimization as a feasible solution.
Preferably, the value of the optimization variable is { -8e-10,30}.
Preferably, the particle swarm optimization algorithm has an initial size of 200 particles, an iteration number of 900 generations, and a particle dimension of 56.
Preferably, the support vector regression algorithm specifically includes the following steps:
step 1, establishing a local training set:
the scale of the local training set is 200, the local training set is empty initially, when the particle swarm optimization algorithm is iterated for the first time, the particle swarm is added into the local training set until the local training set reaches the scale, after the particle swarm optimization algorithm is iterated for the second time, 1/4 particles of the particle swarm are selected and added into the local training set each time according to the mode that 1 particle is randomly selected from every 4 particles, and the FIFO principle is executed to maintain the scale of the training set;
step 2, constructing a support vector regression algorithm and updating along with iteration of the particle swarm optimization algorithm;
the penalty factor is set to be 300, the gamma value is set to be 0.003, the updated local training set is input into the support vector regression algorithm during each iteration of the particle swarm optimization algorithm, and the fitness function calculation is carried out on the particles added into the local training set from the second iteration of the particle swarm optimization algorithm, so that the support vector regression algorithm is trained;
and 3, predicting the fitness of the remaining 3/4 particles of the particle swarm in each iteration by using a support vector regression algorithm.
Preferably, the fitness factor satisfies the following values:
when the parameterization representation of any part exceeds the shape fitting function value under a certain shape optimization scheme, fit =0;
fit =1 when, under a certain profile optimization scheme, the parameterized representations of all parts are within the profile fitting function values.
The invention has the following beneficial effects:
(1) Compared with a standard particle swarm algorithm, the method greatly reduces the calculation times of the fitness, can shorten the time of optimization iteration, improves the calculation precision, and provides a referable thought for the processing of high-dimensional structure optimization problems.
(2) The SVR-PSO-based voice robot appearance optimization method is designed and developed by the invention, a design framework for overall appearance parameter optimization of the voice recognition robot is established, 3D modeling is carried out on an appearance shell and internal parts of the voice recognition robot by utilizing Solidworks software, parameterized representation of the appearance shell and parameterized representation of a part assembly body are realized, and the problem that a complex constraint function is difficult to design in robot structure optimization can be solved.
Drawings
FIG. 1 is a schematic flow chart of the shape optimization method of the SVR-PSO-based voice robot according to the present invention.
Fig. 2 is a schematic structural diagram of the shell model according to the present invention.
Fig. 3 is a schematic structural diagram of the part assembly model according to the present invention.
Detailed Description
The present invention is described in further detail below to enable those skilled in the art to practice the invention with reference to the description.
As shown in fig. 1, the shape optimization method of the speech robot based on SVR-PSO provided by the present invention is based on a Particle Swarm Optimization (PSO), a new fitness evaluation strategy is established by using a Support Vector Regression (SVR), and fitness values of some individuals are predicted by using an SVR model, so as to reduce the calculation times of the fitness, shorten the time of optimization iteration, improve the calculation accuracy, and provide an identifiable idea for processing a high-dimensional structure optimization problem.
The method specifically comprises the following steps:
step one, modeling an appearance shell and internal parts of the intelligent voice robot, and specifically comprising:
(1) As shown in fig. 2, a picture of an existing intelligent voice robot is imported into Solidworks software (2016 edition), an outline curve corresponding to the intelligent voice robot can be obtained by drawing the outline edge and adjusting the size of the outline edge, and a shell model of the intelligent voice robot can be obtained by rotating the outline curve;
(2) As shown in fig. 3, in Solidworks, the design of each part is completed first, the size of the part is modified to meet the requirement, and then all parts are combined together to obtain a part assembly model;
wherein, the part assembly model specifically includes: first sound pickup microphone 110, laser radar 120, power amplifier 130, main board and hub140, middle fixing plate 150, first sound box 161, second sound box 162, battery 170, processor 180 and chassis 190.
Step two, carrying out parametric representation on the shell model and the part assembling body model of the intelligent voice robot, and specifically comprising the following steps:
1. carrying out parameterized representation on the shell model of the intelligent voice robot:
(1) Deriving a Point Cloud of the shell model by using a Point Cloud Library (PCL), namely a coordinate Point set with uniform distribution on the appearance surface of the intelligent voice robot;
(2) Since the situation of coordinate point shift occurs when PCL processes 3D models, the coordinate point set needs to be processed:
considering that PCL obtains point cloud with uniform distribution, the mean values of all coordinate points on the X axis and the Y axis can be calculated, and then the mean values are subtracted from all coordinate points one by one, so that the centralization of a coordinate point set on the X axis and the Y axis can be realized;
wherein, calculating the mean value of all coordinate points on the X axis and the Y axis is:
adding all the X-axis numerical values in the coordinate point set, dividing the added X-axis numerical values by the number of coordinates in the coordinate point set to obtain the mean value of all the coordinate points on the X axis, adding all the Y-axis numerical values in the coordinate point set, and dividing the added Y-axis numerical values by the number of coordinates in the coordinate point set to obtain the mean value of all the coordinate points on the Y axis;
(3) Step 3, fitting an polynomial of the shape curved surface of the robot according to the processed coordinate points:
the method comprises the following steps of establishing a shell model coordinate system by taking the intersection point of a central axis and upper and lower boundaries of a shell model as a coordinate origin, taking the upper and lower interfaces as an XOY plane, taking the central axis of the shell model as a Z axis, upwards taking the central axis as a positive direction, using coordinate points with a Z value larger than zero for fitting an upper surface function, and using coordinate points with a Z value smaller than zero for fitting a lower surface function to obtain a shape fitting function:
because the shell model is a rotationally symmetric graph, and meanwhile, as shown in fig. 2, the shell model has obvious upper and lower dividing lines, a shell model coordinate system is established by using the upper and lower dividing lines of the shell model, the upper and lower dividing lines are used as an XOY plane, a central axis of the shell model is a Z axis, the vertical direction is a positive value of the Z axis, the vertical direction is a negative value of the Z axis, then the appearance height of the intelligent voice robot is divided, coordinate points with the Z value larger than zero are used for fitting an upper curved function, and coordinate points with the Z value smaller than zero are used for fitting a lower curved functionFitting, respectively performing polynomial fitting on the two parts of coordinate points, and utilizing R 2 The values of the values are evaluated on polynomial regression models with different degrees, and the result that the fitting effect is best when the upper surface function and the lower surface function are both sextic polynomials can be found, the effect of the function on fitting the graph is good, and R is the effect of the function on fitting the graph 2 Values of 0.999785135463 and 0.971197127999, respectively, the shape fitting function is obtained as:
f up (x，y)＝W 0_0 +W 0_1 x+W 0_2 y+W 0_3 x 2 +W 0_4 x*y+W 0_5 y 2 +W 0_6 x 3 +W 0_7 y*x 2 +W 0_8 x*y 2 +W 0_9 y 3 +W 0_10 x 4 +W 0_11 x 3 *y+W 0_12 x 2 *y 2 +W 0_13 x*y 3 +W 0_14 y 4 +W 0_15 x 5 +W 0_16 x 4 *y+W 0_17 x 3 *y 2 +W 0_18 x 2 *y 3 +W 0_19 x*y 4 +W 0_20 y 5 +W 0_21 x 6 +W 0_22 x 5 *y+W 0_23 x 4 *y 2 +W 0_24 x 3 *y 3 +W 0_25 x 2 *y 4 +W 0_26 x*y 5 +W 0_27 y 6
f down (x，y)＝W 1_0 +W 1_1 x+W 1_2 y+W 1_3 x 2 +W 1_4 x*y+W 1_5 y 2 +W 1_6 x 3 +W 1_7 y*x 2 +W 1_8 x*y 2 +W 1_9 y 3 +W 1_10 x 4 +W 1_11 x 3 *y+W 1_12 x 2 *y 2 +W 1_13 x*y 3 +W 1_14 y 4 +W 1_15 x 5 +W 1_16 x 4 *y+W 1_17 x 3 *y 2 +W 1_18 x 2 *y 3 +W 1_ 19 x*y 4 +W 1_20 y 5 +W 1_21 x 6 +W 1_22 x 5 *y+W 1_23 x 4 *y 2 +W 1_24 x 3 *y 3 +W 1_25 x 2 *y 4 +W 1_26 x*y 5 +W 1_27 y 6
in the formula, f up (x, y) is the upper surface function, W 0_0 、W 0_1 、W 0_2 、W 0_3 …W 0_21 、W 0_22 、W 0_23 …W 0_27 Equation coefficients, f, which are all functions of the upper surface down (x, y) is a lower surface function, W 1_0 、W 1_1 、W 1_2 、W 1_3 …W 1_21 、…W 1_27 Are all the equation coefficients of the lower surface function.
2. Carrying out parametric representation on a part assembly model of the intelligent voice robot:
considering that the position and the shape of the part are fixed and do not need to be changed like the appearance shell of the intelligent voice robot, a plurality of parts in the part assembly can be directly projected to the Z axis in the shell model coordinate system, and the projection height of the part on the Z axis is taken as the parameterized representation of the part;
if a plurality of parts are crossed in the same projection range and the projection of the parts is on the Z-axis positive direction, taking the maximum value of the height of the parts as the parameterized representation of the parts; and if the parts intersect in the same projection range and the projection of the parts is in the negative direction of the Z axis, taking the minimum value of the height of the parts as the parameterized representation of the parts.
Step three, constructing an appearance optimization model:
the method is characterized in that the minimum volume of a shell model of the intelligent voice robot is taken as an optimization target, equation coefficients in an upper curved surface function and a lower curved surface function which are fitted by utilizing three-dimensional coordinate points on the appearance surface of the intelligent voice robot are taken as optimization variables, and a part assembly model is taken as a constraint condition, namely, the condition that whether parts penetrate through the upper curved surface or the lower curved surface (the shell model) exists in the projection range of a shell model coordinate system is judged.
Wherein, the invention adopts the thought of the infinitesimal method to carry out approximate calculation on the volume of the intelligent voice robot,that is, the area enclosed by the curved surface (upper curved surface and lower curved surface) and the plane (bottom surface) is divided into a plurality of small cubes, and the bottom area of each small cube is marked as Deltax i And recording the maximum value of the bottom area as delta and the corresponding height of the curved surface as h, then obtaining the volume of each small cube by means of local 'curve substitution by straight', and adding the volumes of each cube to approximate to the actual curved surface volume.
In the invention, the range of the bottom surface is set as [ -12,12], then the bottom surface is divided into small squares of 1mm x 1mm, the coordinate of the central point of each small square is obtained in a shell model coordinate system, the coordinate of the central point of each small square is respectively substituted into an appearance fitting function to obtain an upper curved surface function value and a lower curved surface function value, if the difference value of subtracting the lower curved surface function value from the upper curved surface function value is more than 0, the value is reserved as the height value of the cube, and then the volumes of all cubes are calculated; and if the difference value obtained by subtracting the lower curved surface function value from the upper curved surface function value is not more than 0, the lower curved surface of the cube is higher than the upper curved surface, obviously unreasonable, the volume of the corresponding cube is 0, and then the volumes of all cubes are calculated, so that the volume of the shell is obtained.
The constraint condition of the part assembling body model is as follows:
the values of x and y can be known to be in the range of [ -12,12] through the shape fitting function, therefore, only whether the optimization variable meets the part constraint needs to be judged in the range of [ -12,12], that is, under a certain coordinate point, if the height difference between the upper curved surface function and the upper half part or the height difference between the lower curved surface function and the lower half part is less than 0, it means that the upper half part penetrates through the upper curved surface or the lower half part penetrates through the lower curved surface, and only if both the difference values are greater than 0, the obtained solution is a feasible solution of the optimization problem.
The optimization variables satisfy:
the shape of the intelligent voice robot is divided into an upper part and a lower part, a function equation of the upper curved surface and a function equation of the lower curved surface are respectively obtained by utilizing polynomial function fitting, the coefficient of the function equation is a parameter needing to be adjusted in the optimization process of solving the minimum volume of the voice recognition robot, namely, an optimization variable, each optimization variable has a respective value range, the maximum value and the minimum value of each variable value are determined by taking the constraint of exceeding parts as the limit through multiple times of experimental adjustment on each variable value, then the optimal value is searched by randomly adjusting the value of the optimization variable in the value range, wherein the value range of the optimization variable is { -8e-10,30}.
Optimizing the shape optimization model through a particle swarm optimization algorithm, and searching the optimal solution of the current shape optimization model, wherein the method specifically comprises the following steps of:
step 1, initializing a particle swarm:
in the particle swarm optimization algorithm, optimization variables are used for representing particles, the algorithm is initialized to 200 particles, the position and the speed of each particle are initialized randomly, the iteration times are set to 900 generations, and the particle dimension is the number of the optimization variables, namely 56.
Step 2, determining the fitness function of the particles and ensuring that the fitness function of the particles is larger and better:
Figure BDA0003870772400000091
where fitness is the fitness function, v 0 Is the volume of the original shape, v is the volume corresponding to the new individual, fit is the fitness factor;
when the projection height value of the upper half part is higher than the height of the upper half part curved surface corresponding to the individual in the projection range of the part or the projection height value of the lower half part is lower than the height of the lower half part curved surface corresponding to the individual, the part penetrates out of the shell, and then fit =1, otherwise fit =0.
Step 3, reducing the calculation times of the target function through a support vector regression algorithm:
(1) Establishing and maintaining a local training set:
the scale of the local training set is 200, the local training set is empty initially, when the particle swarm optimization algorithm is iterated for the first time, the particle swarm is added into the local training set until the local training set reaches the scale, at the beginning of the second iteration of the particle swarm optimization algorithm, an average sampling selection strategy is adopted, a mode of randomly selecting 1 particle from every 4 particles is adopted, the sample selected in the mode is representative, 1/4 particles of the particle swarm which is iterated each time are selected to be added into the local training set, the FIFO principle is executed, the number of the particles is eliminated, the scale of the training set is maintained, the newly generated population is greatly different from the initial population due to continuous iteration updating, the training set can be ensured to be similar to the latest population by using the FIFO principle, the current particle distribution can be better described, and the regression result can be more accurate.
(2) Constructing a support vector regression algorithm and training an SVR (support vector regression) model through a local training set along with iteration of a particle swarm optimization algorithm, wherein the size of the local training set is 200, and an acceleration constant is 2;
the essence of the support vector regression algorithm is to solve the regression problem using the idea of support vector machine, i.e. D = { (x) in the given training data 1 ,y 1 ),(x 2 ,y 2 ),...,(x m ,y m ) In the case of (f) = w, f (x) = w is desirable T The solution obtained by x + b makes f (x) and y equal as much as possible, w and b are model parameters, a buffer threshold epsilon, also called a relaxation variable, is designed for the calculation process by the support vector regression algorithm, and the loss is calculated only when the absolute value of the difference between the model output value f (x) and the true value y is larger than epsilon.
The SVR model of the invention adopts a Gaussian radial basis kernel function:
Figure BDA0003870772400000101
as a kernel function in the support vector regression model, the penalty factor is set to 300, the gamma value is set to 0.003,
inputting an updated local training set into the support vector regression algorithm during each iteration of the particle swarm optimization algorithm, and performing fitness function calculation on particles added into the local training set from the second iteration of the particle swarm optimization algorithm so as to train the support vector regression algorithm;
(3) And calling an interface of the SVR model by using a getmodel method to predict the fitness of the remaining 3/4 particles of the particle swarm in each iteration according to the SVR model, thereby reducing the fitness calculation times of the particle swarm algorithm and shortening the operation time of the algorithm.
And 4, updating the individual optimal position pbest and the global optimal position gbest through the fitness value of each particle.
And 5, judging whether the termination condition is met, if the termination condition is not met, recalculating the fitness function value of each particle in the population, repeating the operation, and if the termination condition is met, determining the optimal particle as the optimal solution of the current appearance optimization model according to the fitness function value of each particle in the current population.
The invention designs and develops a shape optimization method of a voice robot based on SVR-PSO, which combines a support vector regression algorithm with a particle swarm optimization algorithm, establishes a new fitness evaluation strategy by utilizing the support vector regression algorithm, predicts the fitness of an individual to be evaluated through a trained SVR model, reduces the calculation times of the fitness in the particle swarm optimization algorithm, can shorten the optimization iteration time, improves the calculation precision and provides a referable thought for the processing of the modern high-dimensional structure optimization problem.
While embodiments of the invention have been described above, it is not limited to the applications set forth in the description and the embodiments, which are fully applicable to various fields of endeavor for which the invention may be embodied with additional modifications as would be readily apparent to those skilled in the art, and the invention is therefore not limited to the details given herein and to the embodiments shown and described without departing from the generic concept as defined by the claims and their equivalents.

Claims (10)

1. A shape optimization method of a voice robot based on SVR-PSO is characterized by comprising the following steps:
the method comprises the following steps that firstly, modeling is carried out on an appearance shell and internal parts of the intelligent voice robot to obtain a shell model and a part assembly model;
carrying out parametric representation on the shell model to obtain a shape fitting function, and carrying out parametric representation on the assembling body model to obtain the projection heights of a plurality of parts;
step three, constructing an appearance optimization model:
taking the minimum volume of the shell model as an optimization target, taking an equation coefficient in an appearance fitting function as an optimization variable, and taking a part assembly model as a constraint condition;
optimizing the shape optimization model through a particle swarm optimization algorithm, wherein after the iteration times are reached, the maximum fitness function of the particles is the optimal solution of the shape optimization model, and the shape of the intelligent voice robot is adjusted to the minimum volume;
the operation time of the particle swarm optimization algorithm is calculated through a support vector regression algorithm;
the fitness function of the particle is:
Figure FDA0003870772390000011
where fitness is the fitness function, v 0 Is the volume of the original shape, v is the volume corresponding to the new individual, and fit is the fitness factor.
2. The method for optimizing the shape of an SVR-PSO-based speech robot according to claim 1, wherein said parameterizing the shell model comprises the steps of:
step 1, deriving a point cloud of a shell model;
step 2, performing centralization treatment on the point cloud:
calculating the mean values of all coordinate points in the point cloud on the X axis and the Y axis, and subtracting the mean values from all the coordinate points one by one to obtain processed coordinate points;
step 3, fitting an polynomial of the shape curved surface of the robot according to the processed coordinate points:
the method comprises the following steps of establishing a shell model coordinate system by taking the intersection point of a central axis and upper and lower boundaries of a shell model as a coordinate origin, taking the upper and lower interfaces as an XOY plane, taking the central axis of the shell model as a Z axis, upwards taking the central axis as a positive direction, using coordinate points with a Z value larger than zero for fitting an upper surface function, and using coordinate points with a Z value smaller than zero for fitting a lower surface function to obtain a shape fitting function:
f up (x，y)＝W 0_0 +W 0_1 x+W 0_2 y+W 0_3 x 2 +W 0_4 x*y+W 0_5 y 2 +W 0_6 x 3 +W 0_7 y*x 2 +W 0_8 x*y 2 +W 0_9 y 3 +W 0_10 x 4 +W 0_11 x 3 *y+W 0_12 x 2 *y 2 +W 0_13 x*y 3 +W 0_14 y 4 +W 0_15 x 5 +W 0_16 x 4 *y+W 0_17 x 3 *y 2 +W 0_18 x 2 *y 3 +W 0_19 x*y 4 +W 0_20 y 5 +W 0_21 x 6 +W 0_22 x 5 *y+W 0_23 x 4 *y 2 +W 0_24 x 3 *y 3 +W 0_25 x 2 *y 4 +W 0_26 x*y 5 +W 0_27 y 6
f down (x，y)＝W 1_0 +W 1_1 x+W 1_2 y+W 1_3 x 2 +W 1_4 x*y+W 1_5 y 2 +W 1_6 x 3 +W 1_7 y*x 2 +W 1_8 x*y 2 +W 1_9 y 3 +W 1_ 10 x 4 +W 1_11 x 3 *y+W 1_12 x 2 *y 2 +W 1_13 x*y 3 +W 1_14 y 4 +W 1_15 x 5 +W 1_16 x 4 *y+W 1_17 x 3 *y 2 +W 1_18 x 2 *y 3 +W 1_19 x*y 4 +W 1_20 y 5 +W 1_21 x 6 +W 1_22 x 5 *y+W 1_23 x 4 *y 2 +W 1_24 x 3 *y 3 +W 1_25 x 2 *y 4 +W 1_26 x*y 5 +W 1_27 y 6
in the formula (f) up (x, y) is an upper surface function, W 0_0 、W 0_1 、W 0_2 、W 0_3 …W 0_21 、W 0_22 、W 0_23 …W 0_27 Equation coefficients, f, which are all functions of the upper surface down (x, y) is the lower surface function, W 1_0 、W 1_1 、W 1_2 、W 1_3 …W 1_21 、…W 1_27 Are all equation coefficients of the lower surface function.
3. The method of claim 2, wherein the X-axis and Y-axis values of the shell model in the shell model coordinate system are both [ -12,12] as determined by the shape fitting function.
4. The method for optimizing the shape of an SVR-PSO-based speech robot of claim 3, wherein said parameterizing said rigging body model comprises:
projecting a plurality of parts in the assembly body to a Z axis in the shell model coordinate system to obtain the projection height of the parts on the Z axis, and if the plurality of parts are intersected in the same projection range and the projection of the parts is on the positive direction of the Z axis, taking the maximum value of the height of the parts as the parameterized representation of the parts; and if the parts intersect in the same projection range and the projection of the parts is in the negative direction of the Z axis, taking the minimum value of the height of the parts as the parameterized representation of the parts.
5. The method for optimizing the shape of an SVR-PSO-based voice robot according to claim 4, wherein said calculating the volume of said shell model comprises:
dividing the bottom surface of the shell model into small squares of 1mm x 1mm, obtaining the coordinate of the central point of each small square in a shell model coordinate system, respectively substituting the coordinate of the central point of each small square into an appearance fitting function to obtain an upper curved surface function value and a lower curved surface function value, if the difference value of subtracting the lower curved surface function value from the upper curved surface function value is greater than 0, the difference value is the height value of the corresponding cube, and then calculating the volumes of all cubes to further obtain the volume of the shell model; and if the difference value obtained by subtracting the lower curved surface function value from the upper curved surface function value is not more than 0, the volume of the corresponding cube is 0, and then the volumes of all the small squares are calculated to further obtain the volume of the shell model.
6. The method of optimizing the shape of an SVR-PSO-based speech robot of claim 5, wherein the constraints comprise:
when the x value of the part assembly model is within the value range of the x value of the shell model and the y value of the part assembly model is within the value range of the y value of the shell model:
if the shape-optimized coordinate point is in the Z-axis positive direction in the shell model coordinate system, calculating a curved surface function value, and if the difference value between the shape-optimized coordinate point and the Z-axis positive part parameterization is greater than 0, the shape-optimized coordinate point is a feasible solution;
and if the coordinate point after the shape optimization is in the Z-axis negative direction in the shell model coordinate system, calculating a lower curved surface function value, and if the difference value of the Z-axis negative direction part parameterization and the Z-axis negative direction part parameterization is more than 0, determining the coordinate point after the shape optimization as a feasible solution.
7. The method of claim 6, wherein the optimization variable is selected from the group consisting of-8 e-10, 30.
8. The method of claim 7, wherein the particle swarm optimization algorithm is initially 200 particles, iterates 900 generations, and has a particle dimension of 56.
9. The method of claim 8, wherein the support vector regression algorithm comprises the following steps:
step 1, establishing a local training set:
the scale of the local training set is 200, the local training set is empty initially, when the particle swarm optimization algorithm is iterated for the first time, the particle swarm is added into the local training set until the local training set reaches the scale, after the particle swarm optimization algorithm is iterated for the second time, 1/4 particles of the particle swarm are selected and added into the local training set each time according to the mode that 1 particle is randomly selected from every 4 particles, and the FIFO principle is executed to maintain the scale of the training set;
step 2, constructing a support vector regression algorithm and updating along with iteration of the particle swarm optimization algorithm;
the penalty factor is set to be 300, the gamma value is set to be 0.003, the updated local training set is input into the support vector regression algorithm during each iteration of the particle swarm optimization algorithm, and the fitness function calculation is carried out on the particles added into the local training set from the second iteration of the particle swarm optimization algorithm, so that the support vector regression algorithm is trained;
and 3, predicting the fitness of the remaining 3/4 particles of the particle swarm in each iteration by using a support vector regression algorithm.
10. The method of claim 9, wherein the fitness factor takes on a value that satisfies the following:
when the parameterization representation of any part exceeds the shape fitting function value under a certain shape optimization scheme, fit =0;
fit =1 when, under a certain profile optimization scheme, the parameterized representations of all parts are within the profile fit function values.
CN202211197359.3A 2022-09-29 2022-09-29 SVR-PSO-based voice robot shape optimization method Pending CN115470593A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211197359.3A CN115470593A (en) 2022-09-29 2022-09-29 SVR-PSO-based voice robot shape optimization method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211197359.3A CN115470593A (en) 2022-09-29 2022-09-29 SVR-PSO-based voice robot shape optimization method

Publications (1)

Publication Number Publication Date
CN115470593A true CN115470593A (en) 2022-12-13

Family

ID=84335449

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211197359.3A Pending CN115470593A (en) 2022-09-29 2022-09-29 SVR-PSO-based voice robot shape optimization method

Country Status (1)

Country Link
CN (1) CN115470593A (en)

Similar Documents

Publication Publication Date Title
Rameshbabu et al. Hybrid feature recognition method for setup planning from STEP AP-203
CN113297650B (en) BIM technology-based unit type glass curtain wall construction method and system
CN110543727B (en) Improved particle swarm algorithm-based omnidirectional mobile intelligent wheelchair robot parameter identification method
CN112508895B (en) Propeller blade quality assessment method based on curved surface registration
CN113276130B (en) Free-form surface spraying path planning method and system based on point cloud slice
CN113221478A (en) Numerical analysis method and system for actively controlling rotor wing
CN114676522B (en) Pneumatic shape optimization design method, system and equipment integrating GAN and migration learning
CN116503575A (en) Method for graphically generating high-quality building model and electronic equipment
CN113593043B (en) Point cloud three-dimensional reconstruction method and system based on generation countermeasure network
CN114721402A (en) Parking path planning method and system based on improved ant colony algorithm
CN115470593A (en) SVR-PSO-based voice robot shape optimization method
JP2881389B2 (en) Automatic mesh generation method and system
Yu et al. A hybrid point cloud alignment method combining particle swarm optimization and iterative closest point method
EP4092557A1 (en) Parameterization of digital irregular freeform geometries
CN115130340A (en) Pipeline modeling method based on fractional Brownian motion
CN106909730B (en) Building three-dimensional model simulation method and system based on homotopy mapping algorithm
JP2023084699A (en) Computer aided generative design with feature thickness control to facilitate manufacturing and structural performance
CN116776401A (en) Method for modeling an object affected by boundary conditions from a three-dimensional object model
CN114841079A (en) Continuous collision detection method based on deep neural network optimization
CN111222262B (en) Air floatation vibration isolation platform performance optimization design method based on mass ratio influence
CN114218845A (en) SPH calculation optimization method and system based on particle splitting
Winter et al. NURBS-based shape and parameter optimization of structural components with an adaptive amount of control points
CN114019911A (en) Curve fitting method based on speed planning
Gu et al. Application research of mechanical mold design based on computer aided technology
Nie Efficient cutter-workpiece engagement determination in multi-axis milling by voxel modeling

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